WEBVTT
00:00:23.730 --> 00:00:27.489
Thank you for joining Community IT for this podcast, part one.
00:00:27.489 --> 00:00:34.769
Subscribe wherever you listen to podcasts and leave us a rating to help others find this leadership resource for nonprofits.
00:00:34.769 --> 00:00:37.969
Listen for part two in your podcast feed.
00:00:47.409 --> 00:01:07.170
If you don't know the difference between using a freemium tool like ChatGPT or logging on to a more private enterprise tool at your organization, like Copilot if you're using Microsoft or Gemini if you're using Google Workspace, this webinar is really going to clarify that for you.
00:01:07.170 --> 00:01:16.289
We hear a lot about make sure not to share your sensitive data with the AI learning models, but how do you know that you're using them safely?
00:01:16.289 --> 00:01:18.530
How do you check the terms and conditions?
00:01:18.530 --> 00:01:19.810
Where do you begin?
00:01:19.810 --> 00:01:35.649
Today, our cybersecurity expert Matt Eshleman is going to demystify the enterprise AI and share some tips on how to create an organizational AI policy, how to share knowledge, how to do staff training.
00:01:35.649 --> 00:01:40.530
So all you and all of your colleagues are using AI in the most secure way possible.
00:01:41.170 --> 00:01:42.929
My name is Carolyn Woodard.
00:01:42.929 --> 00:01:45.250
I'm the outreach director for Community IT.
00:01:45.250 --> 00:01:46.929
I'm going to be the moderator today.
00:01:46.929 --> 00:01:50.449
But first, I'm going to go over our learning objectives.
00:01:50.689 --> 00:02:06.609
So by the end of the session today, we hope that you will learn the difference between enterprise or subscription AI and the more freemium, the free AI tools like ChatGPT, if you just Google it and go to the website.
00:02:06.609 --> 00:02:14.770
We're going to go over accessing Microsoft Copilot and Google Workspace Gemini tools at the organizational level.
00:02:14.770 --> 00:02:21.009
We're going to review IT policy guidelines and provide some training and knowledge sharing tips.
00:02:21.009 --> 00:02:28.449
If you have some tips, something that you're doing at your organization that's working, I'm going to have a chance for you to share that with us later.
00:02:28.930 --> 00:02:40.930
If you're looking for more information on AI topics, we just started a midweek nonprofit AI podcast where we give you 10 to 15 minutes of news and resources weekly.
00:02:40.930 --> 00:02:52.449
If you subscribe to our regular podcast, the Technology Topics Podcast, you'll get that in your feed on Tuesdays and our regular Friday podcast on Fridays.
00:02:52.449 --> 00:02:55.889
On Friday, we have lots of different nonprofit IT topics.
00:02:55.889 --> 00:02:57.729
We chat with guests.
00:02:57.729 --> 00:03:09.250
We also have some other recorded webinars which you can access on our site, communityite.com, covering things like creating an ethical AI framework for your organization, AI governance.
00:03:09.250 --> 00:03:16.049
We did an amazing webinar last summer with Brenda Foster on how to use AI kind of in general.
00:03:16.049 --> 00:03:23.729
We have a downloadable AI acceptable use policy template that you can use if you're working on your policy for your organization.
00:03:23.969 --> 00:03:29.650
So we have a lot of those kind of bigger questions, ethical questions in other resources.
00:03:29.650 --> 00:03:38.049
Today we're really going to focus on this specific question that we get a lot, which is how do you know if you're using your AI tools safely?
00:03:38.049 --> 00:03:41.490
And with that, Matt, would you like to introduce yourself?
00:03:42.449 --> 00:03:42.689
Great.
00:03:42.689 --> 00:03:44.769
Well, thanks for that introduction, Carolyn.
00:03:44.769 --> 00:03:46.049
It's great to be here with you.
00:03:46.049 --> 00:03:56.210
And I'm looking forward to talking about adopting AI in your organization and kind of sharing some of those tips and tricks and getting under the hood a little bit.
00:03:56.210 --> 00:04:01.169
As Carolyn mentioned, I'm the chief technology officer here at Community IT.
00:04:01.169 --> 00:04:09.409
And I am pleased that I've just celebrated 24 years officially full-time with Community IT.
00:04:09.409 --> 00:04:16.530
And so get to play a, do a lot of different things, play a lot of different roles, and I'm really excited about this topic in particular.
00:04:17.810 --> 00:04:19.410
And you have so much experience.
00:04:19.410 --> 00:04:30.290
Before we get started with Matt, if you're not familiar with community IT, I want to tell you just a little bit about us.
00:04:30.290 --> 00:04:35.090
We are a 100% employee-owned managed services provider.
00:04:35.090 --> 00:04:37.650
We provide outsourced IT support.
00:04:37.650 --> 00:04:47.409
We work exclusively with nonprofit organizations, and our mission is to help nonprofits accomplish their missions through the effective use of technology.
00:04:47.409 --> 00:04:51.810
We are big fans of what well-managed IT can do for your nonprofit.
00:04:51.810 --> 00:04:54.610
We serve nonprofits across the United States.
00:04:54.610 --> 00:04:56.610
We've been doing this for 25 years.
00:04:56.610 --> 00:04:59.009
This is our 25-year anniversary.
00:04:59.009 --> 00:05:09.009
We are technology experts and are consistently given an MSP 501 recognition for being a top MSP, which is an honor we received again in 2025.
00:05:09.009 --> 00:05:14.529
And we believe that we're the only MSP on the list that serves nonprofits exclusively.
00:05:14.769 --> 00:05:19.730
I want to remind everyone that for these presentations, community IT is vendor agnostic.
00:05:19.730 --> 00:05:24.689
We only make recommendations to our clients and only based on their specific business needs.
00:05:24.689 --> 00:05:30.370
We never try to get a client into a product because we get an incentive or a benefit from it.
00:05:30.370 --> 00:05:34.050
We do consider ourselves a best of breed IT provider.
00:05:34.050 --> 00:05:39.810
So it's our job to know the landscape, what tools are available, reputable, and widely used.
00:05:39.810 --> 00:05:46.689
And we make recommendations on that basis for our clients based on their business needs, priorities, and budget.
00:05:46.689 --> 00:05:56.449
We're going to talk about two big uh nonprofit stack, uh IT stacks that everyone uses, not just nonprofits, Microsoft and Google today.
00:05:56.449 --> 00:05:59.490
Um, but that is because a lot of nonprofits are using them.
00:05:59.730 --> 00:06:02.209
We've got a lot of good questions at registration.
00:06:02.209 --> 00:06:04.610
So we're gonna try and answer as many of those as we can.
00:06:04.610 --> 00:06:17.009
But anything we can't get to, uh, please join us and Matt today on our in our community on Reddit at r slash nonprofit IT management after the webinar for about 30 minutes.
00:06:17.009 --> 00:06:23.969
Uh and he also, Matt also pops in once a week or so to answer any other questions that come in so you can take advantage of that.
00:06:24.209 --> 00:06:25.730
And a little bit more about us.
00:06:25.730 --> 00:06:31.250
Our mission, as I said, is to create value for the nonprofit sector through well-managed IT.
00:06:31.250 --> 00:06:39.730
We also identify four key values as employee owners that define our company: trust, knowledge, service, and balance.
00:06:39.730 --> 00:06:43.329
We seek always to treat people with respect and fairness.
00:06:43.329 --> 00:06:58.689
We seek to empower our staff, clients, and sector to understand and use technology effectively, to be helpful with our talents, and we recognize that the health of our communities is vital to our well-being and that work is only a part of our lives.
00:06:59.089 --> 00:07:03.649
All right, as we usually do, I'm gonna start out with a poll.
00:07:03.649 --> 00:07:07.649
We just want to get a feel for your comfort level with AI tools.
00:07:07.649 --> 00:07:14.850
So your options are one, completely uncomfortable or unfamiliar with most tools.
00:07:14.850 --> 00:07:19.649
Two, somewhat uncomfortable, we use a few popular tools occasionally.
00:07:19.649 --> 00:07:24.370
Three is the neutral, neither uncomfortable nor comfortable kind of average use.
00:07:24.370 --> 00:07:26.610
Four is somewhat comfortable.
00:07:26.610 --> 00:07:29.089
I use a few AI tools daily.
00:07:29.089 --> 00:07:31.329
And five is completely comfortable.
00:07:31.329 --> 00:07:33.569
I use a lot of these tools a lot of the time.
00:07:33.569 --> 00:07:37.649
My colleagues ask me how to, you know, how to use them to teach them how.
00:07:37.649 --> 00:07:41.889
And the six option is uh not applicable or other.
00:07:41.889 --> 00:07:44.769
And Matt, can you see the results?
00:07:45.089 --> 00:07:45.649
Yes, I can.
00:07:45.649 --> 00:07:47.569
I always like it's like a big a big reveal.
00:07:47.569 --> 00:07:50.529
You know, you got the drum roll because I can't see it as it's coming in.
00:07:50.610 --> 00:07:53.329
So all right, so can you go ahead and share it with us?
00:07:53.490 --> 00:07:55.009
How uh yes.
00:07:55.009 --> 00:08:05.329
So, in terms of uh the folks that are um responding today, uh about 12% are completely uh uncomfortable, unfamiliar, haven't really done much at all.
00:08:05.329 --> 00:08:15.730
Um, which maybe on the flip side, right, about 9% of the respondents are completely comfortable and use it a lot, and they're really the resource that people in their organization go to.
00:08:15.730 --> 00:08:18.129
So we kind of have the those those ends.
00:08:18.129 --> 00:08:23.170
And then in the middle, uh we have about 29% of folks are somewhat uncomfortable.
00:08:23.170 --> 00:08:25.730
Maybe they use some of the tools occasionally.
00:08:25.730 --> 00:08:30.529
16% are neither uncomfortable, like right in the middle.
00:08:30.529 --> 00:08:37.490
Uh, and then we have a bigger bump there uh on the folks that are somewhat comfortable use you you using those AI tools um daily.
00:08:37.490 --> 00:08:42.769
So kind of an interesting distribution um in terms of uh the respondents here.
00:08:42.769 --> 00:08:45.490
So we've got kind of bar, what is it, bar barbells?
00:08:45.490 --> 00:08:50.769
We got uh folks on on kind of both ends of the spectrum uh related to their AI usage.
00:08:51.170 --> 00:09:04.450
Yeah, no, that was interesting because often we get that uh bell curve, but uh here we have people who are not as many people who are really neutral of like they're putting themselves either in I use it sometimes or I really don't use it very often.
00:09:04.450 --> 00:09:06.689
So that was interesting.
00:09:06.689 --> 00:09:07.649
Um, all right.
00:09:07.809 --> 00:09:12.289
Well, we heard me say a word a little while ago, freemium AI.
00:09:12.289 --> 00:09:15.169
And Matt, you were gonna talk about what is that.
00:09:16.210 --> 00:09:33.889
Yeah, so I think freemium, and I use that that term intentionally as we were developing this um presentation to really describe how a lot of us maybe have come to you know using AI tools, uh, you know, probably the one being ChatGPT, which we'll take a look at in terms of their overall growth here in the next slide.
00:09:33.889 --> 00:09:35.009
Um, but really, you know,
00:09:35.009 --> 00:09:40.129
Freemium is a it's a business model uh for getting users into the platform.
00:09:40.129 --> 00:09:43.649
Um AI is very expensive to deploy.
00:09:43.649 --> 00:09:47.649
Um, you know, I was looking uh again, the background research, right?
00:09:47.649 --> 00:10:00.129
So I think OpenAI, which is kind of the entity that owns Chat GP GPT, is kind of scheduled to invest like $1.15 trillion in building up their capacity, right?
00:10:00.129 --> 00:10:02.370
So this is enormously expensive.
00:10:02.370 --> 00:10:11.649
Um and so you know, they're they're giving away or starting to give away um access to their tool in the hopes of converting everyone into paid customers later on.
00:10:11.809 --> 00:10:29.490
And I would distinguish that from you know, maybe a public AM model, which would be more of a utility service, maybe something owned by the government, um, right, but as a available as a public good as opposed to a privately held uh AI solution that uh you know is is fundamentally there to kind of make profit.
00:10:29.649 --> 00:10:41.009
So um, you know, all the AI models, I'm you know, and again, this is we're not gonna get into a lot of the um maybe the philosophy or the the you know kind of ethical issues around here.
00:10:41.009 --> 00:10:48.689
Um, but you know, a lot of the these AI um tools were trained using um publicly available data, um, you know, not always with consent.
00:10:48.689 --> 00:10:54.769
Um, you know, it was with uh, you know, so that the you know, some of those things are currently winding their way through the court system.
00:10:54.769 --> 00:10:57.009
Um and you know, again,
00:10:57.009 --> 00:11:08.289
I think with that freemium model, uh, you know, it's maybe a little bit um flippant, but right, if you're not paying for something, right, then then kind of you're the product, you know, could apply in the in this case as well, right?
00:11:08.289 --> 00:11:28.450
So if you don't pay for an AI tool, the content that you're putting into it, the questions you're asking, how you're interacting with it, uh typically is going to go back in to feed the model to provide you know ever increasing or escalating um information about you know about the usage of the tool and what people are are are asking it and learning about.
00:11:28.450 --> 00:11:38.850
So um again, so I think that's a helpful model to keep in mind, uh, that freemium model largely focused on uh you know converting users to paid customers at some point.
00:11:39.090 --> 00:11:45.490
And as I mentioned, um, you know, ChatGPT is I think the common example, and this one um gets bandied about a lot, right?
00:11:45.490 --> 00:11:51.490
So chat GPT was released, I think, officially back in November of 2022.
00:11:51.490 --> 00:12:00.529
Um and I think it had the fastest uh adoption for um, you know, a million active users, you know, measured in weeks.
00:12:00.529 --> 00:12:14.370
And so, you know, just in the last um, you know, several years, right, to two and a half years, you know, now they're up to over 800 million uh monthly active users or weekly active users in terms of people that are interacting with the system, right?
00:12:14.370 --> 00:12:26.210
So it's been a phenomenally um impressive growth in terms of people going to using and interacting with these tools because they are so compelling, I think.
00:12:26.210 --> 00:12:33.090
And so, you know, the use and adoption is really driven by the fact that people are finding it uh helpful.
00:12:33.090 --> 00:12:38.929
And uh I think it really is a you know the disruptive technology change of this time.
00:12:41.009 --> 00:12:47.889
So that covers the free versions, which we are putting in air quotes because they're not free.
00:12:47.889 --> 00:12:52.850
Um can you talk a little bit more about enterprise AI?
00:12:53.809 --> 00:13:17.570
Yeah, so um, you know, a lot of these free, you know, kind of or these um AI tools or uh large language language models that have been developed by these companies or enterprises, again, they typically have uh the model that they are providing, and then they're either you know, there's a free way to access that, in which case, you know, the you know, there there's some, you know, right?
00:13:17.570 --> 00:13:34.210
You may be giving up some privacy or you know, the data you're putting in the system uh gets used to improve the model, and then you know, distinguish that to enterprise AI, where uh that same back-end model maybe has an intermediary layer that protects the information in there.
00:13:34.210 --> 00:13:34.529
And
00:13:34.529 --> 00:13:42.210
I think this is incredibly important for those enterprise customers to ensure that the data that they are putting in the system is private.
00:13:42.210 --> 00:13:46.529
So in the Microsoft world, um, that's called Copilot.
00:13:46.529 --> 00:14:02.689
So um co-pilot is kind of Microsoft's uh kind of business integration or business um intermediary that uh kind of helps to protect the information that is interrogated using the ChatGPT language model.
00:14:02.689 --> 00:14:05.649
Um the same thing for um Gemini, right?
00:14:05.730 --> 00:14:06.610
So there are free versions.
00:14:06.610 --> 00:14:27.250
If you just go to you know kind of Gemini or um Copilot right now in your web browser, you will likely go to a kind of a free version or uh kind of a consumer version where you don't really have those protections and the information that you put in and the system um spits back, right, is kind of recorded and incorporated and using to train the models moving forward.
00:14:27.250 --> 00:14:28.049
Um, right.
00:14:28.129 --> 00:14:43.409
And so uh the enterprise um companies of Google and Microsoft really want to provide a paid version where you can get those protections uh and you know, in turn, you know, have more and more people use the service.
00:14:43.409 --> 00:14:47.809
Um there are also you know lots of other AI um tools out there.
00:14:47.809 --> 00:14:58.689
Again, this isn't an exhaustive survey and an evaluation of kind of what's out there and their relative worth, but again, there are um dedicated solutions built specifically for AI um for nonprofits.
00:14:58.689 --> 00:15:01.809
So change agent AI is one um such tool.
00:15:01.809 --> 00:15:09.009
And then there's other uh tools, again, you know, innumerable uh that are available for um enterprise subscription.
00:15:09.009 --> 00:15:10.610
And there's a couple listed here.
00:15:10.610 --> 00:15:17.169
And again, if you have some of the favorites that you use that you really find valuable, um you can go ahead and put that into the chat.
00:15:17.169 --> 00:15:17.889
And again,
00:15:17.889 --> 00:15:35.809
I think the important thing with all of these, um, you know, and this is really the shift that we've seen going from on-prem server infrastructure to uh kind of these cloud services, is right, it's much more important, I think, for individuals to understand and organizations to understand, right, what's in the terms and conditions, right?
00:15:35.809 --> 00:15:40.929
How can we use this software uh because we don't really control it at all, right?
00:15:40.929 --> 00:15:46.850
It's not in the server room down the hall where we can, you know, kind of see it and touch it and kind of get under the hood, right?
00:15:46.850 --> 00:15:50.769
It's often some far-flung um data center somewhere.
00:15:50.769 --> 00:15:52.210
And uh, you know,
00:15:52.210 --> 00:16:04.689
apart from the terms and conditions that govern that use, um, you know, it's it's really important to investigate do these terms and conditions align with how I expect uh a system to use my information.
00:16:04.689 --> 00:16:06.289
Um, and again, I'll go back, you know,
00:16:06.289 --> 00:16:11.250
I think change agent AI, for example, if you go to their website and look at their terms and conditions, it is very clear.
00:16:11.250 --> 00:16:16.289
Like I think they do a good job of saying how they use your data, what rights you have to it, right?
00:16:16.289 --> 00:16:26.529
And so that's a little bit clearer to review and analyze than looking at the terms and conditions of some of these other big enterprise um players that are a lot more uh obtuse.
00:16:27.090 --> 00:16:27.490
Yeah.
00:16:27.490 --> 00:16:45.970
Yeah, and I I know that I have a friend who uses AI to help understand what the terms and conditions mean and like summarize them for them because it's all you know can be long documents and trying to pull out what it actually says about um how they use your information.
00:16:46.210 --> 00:16:59.409
So um so you also were telling me a little bit about um these kind of this gradation between the different AI tools and what they do and how autonomous they can be.
00:16:59.409 --> 00:17:01.730
So can you talk a little bit about that?
00:17:02.370 --> 00:17:10.850
Yeah, and I think this is really an exciting, you know, evolution of kind of the AI tool set in terms of how we um think of them.
00:17:10.850 --> 00:17:13.490
So, you know, kind of from the timeline perspective, right?
00:17:13.490 --> 00:17:21.570
So we're talking about you know, November uh, you know, kind of 2022, you know, you could ask, you know, Chat GPT some questions, right?
00:17:21.570 --> 00:17:25.090
And maybe augmented or supplanted your internet search, right?
00:17:25.090 --> 00:17:32.850
You wanted to find some information, you're asking questions, you're you know, maybe getting it to review your your uh document, right?
00:17:33.090 --> 00:17:41.730
So that is generative AI um in terms of it being able to kind of uh analyze and kind of create, create quote unquote new things.
00:17:41.730 --> 00:17:42.850
Um and so
00:17:42.850 --> 00:17:53.810
now we kind of see that evolving into maybe these three um categories, and there maybe could be others, but I think for the purposes of our discussion, right, these are the most typical use cases that we see right now.
00:17:53.810 --> 00:17:54.530
So um
00:17:54.530 --> 00:18:01.570
the first category that kind of exists on this continuum of how AI solutions are being used is really the assistive technology model.
00:18:01.570 --> 00:18:05.730
And so um, you know, that would be again, copilot, right?
00:18:05.730 --> 00:18:09.250
Microsoft is as, you know, that's their branding term, right?
00:18:09.250 --> 00:18:15.490
Where uh it is an acts as an assistant to you, the human who remains fully in control, right?
00:18:15.490 --> 00:18:21.890
You're asking it a question, um, it's providing suggestions or drafts or explanations.
00:18:21.890 --> 00:18:29.170
Um, you know, you're giving it a prompt, it's reproviding providing um a response, and then it's done, right?
00:18:29.170 --> 00:18:31.090
It maybe keeps a history, right?
00:18:31.090 --> 00:18:43.090
So you can have a conversation that uh occurs maybe over a few hours or something, but um you know it's you can't go back and ask it, you know, the analysis that it gave you maybe a couple of days ago, right?
00:18:43.090 --> 00:18:44.130
So uh
00:18:44.130 --> 00:18:53.330
in the assistive world, uh the human's kind of in the top in charge, and it's maybe helpful to think of that as maybe like a smart intern sitting next to you.
00:18:53.330 --> 00:18:57.170
So uh, you know, those are the examples that are pretty common.
00:18:57.170 --> 00:19:06.050
Um, and I think we see uh most folks you know taking their their dipping their toe in using that assistive um model, right?
00:19:06.050 --> 00:19:09.330
It's it's relatively low risk, it's easy to adopt.
00:19:09.330 --> 00:19:21.090
Um you know, the the end user is is kind of at really at the end of the day kind of making a decision about yes, do I want to accept this information or um am I going to reject it, right?
00:19:21.090 --> 00:19:25.730
And and kind of as the final arbiter of those uh of those decisions.
00:19:25.730 --> 00:19:28.850
Um so again, that's a great, a great place to get started.
00:19:28.850 --> 00:19:29.970
Um
00:19:29.970 --> 00:19:35.570
in the middle um here is more of like a kind of a scenario-based or workflow.
00:19:35.570 --> 00:19:43.410
Um and again, I think we're seeing these now be um uh embedded maybe into the tools that you are using.
00:19:43.410 --> 00:19:47.970
Um, you know, so uh, you know, there could be some predefined scenarios.
00:19:47.970 --> 00:19:50.370
So it's triggered by an event.
00:19:50.370 --> 00:19:55.970
Um, you know, there's like certain parameters around when that workflow uh would would trigger.
00:19:55.970 --> 00:20:01.810
And um, you know, the mental models as opposed to like an intern sitting next to you, you're asking questions, right?
00:20:01.810 --> 00:20:07.250
It's it's maybe kind of a smart automation with some judgment about how to make uh some decisions, right?
00:20:07.330 --> 00:20:15.090
So for us, uh you know, we're seeing those tools, you know, kind of crop up in uh, for example, like IT ticketing systems, right?
00:20:15.090 --> 00:20:18.770
The AI can look at this, say, oh, like we've seen this before.
00:20:18.770 --> 00:20:24.610
It looks like the solution might be here and and kind of point to um to some examples, right?
00:20:24.610 --> 00:20:34.850
So it's built into a process that exists um already and you know can help speed up uh that uh analysis and uh reporting.
00:20:34.850 --> 00:20:48.530
So again, you know, it could be those customer support or kind of case management um kind of support triage, uh could be maybe some onboarding or offboarding processes, um, could be some compliance reviews, right?
00:20:48.530 --> 00:20:57.970
So it's um really designed to kind of speed uh maybe some of those routine decisions for regularly occurring processes.
00:20:57.970 --> 00:20:59.090
Uh and then
00:20:59.090 --> 00:21:04.130
the final option or the final example here is the uh agentic model.
00:21:04.130 --> 00:21:10.690
And I guess this might be the new term that you know you have to you can't have a webinar without talking about um agentic AI.
00:21:10.690 --> 00:21:16.370
Um, but that is you know more of like an autonomous or semi-autonomous agent, right?
00:21:16.370 --> 00:21:21.570
That is given a goal, and then the agent kind of figures out how to kind of solve that.
00:21:21.570 --> 00:21:23.810
And again, you may have good examples of that, right?
00:21:23.890 --> 00:21:34.450
If you're interacting with customer support, uh those chats, um, right, those are those are almost, I would say, exclusively agentic AI agents at the start, right?
00:21:34.450 --> 00:21:37.490
And and we're seeing you know that that initial interaction, right?
00:21:37.490 --> 00:21:53.330
Where uh you can ask it questions, it's gonna find a response, it's gonna, you know, it has a kind of a body of knowledge that it's drawing on, that it knows how to um to kind of interpret information and um you know respond.
00:21:53.330 --> 00:22:01.650
And you know, it's it represents real automation and real efficiency for um for organizations to you know to kind of take advantage of.
00:22:01.650 --> 00:22:08.610
And so uh, you know, that that kind of exists right as kind of a again, it's operating on its own.
00:22:08.610 --> 00:22:11.250
It's not um assistive, right?
00:22:11.250 --> 00:22:21.890
It's kind of that co pilot model where it's like somebody sitting next to you, you can kind of set up this agenc model uh and it you know has some instructions, it has a library of information to to draw from.
00:22:21.890 --> 00:22:28.530
It has some boundaries and it can you know basically make decisions and provide answers in and of itself.
00:22:28.530 --> 00:22:30.370
So again, um, you know,
00:22:30.370 --> 00:22:35.090
that can be a good solution for um you know that repetitive um operational work.
00:22:35.090 --> 00:22:37.490
Like if you always have questions you always need to answer, right?
00:22:37.490 --> 00:22:41.730
That can be a great way to to incorporate that into your organization.
00:22:41.730 --> 00:22:57.810
Um, you know, whenever you have pretty clear boundaries um over that, and uh you know, you've maybe you've invested already in those um good processes, good documentations, you know, good information to help feed into that to make good um to make good decisions.
00:22:58.290 --> 00:23:03.090
And so what can you wait, what can you expect from what you pay?
00:23:03.090 --> 00:23:06.050
Yeah, nonprofits always want to know.
00:23:06.370 --> 00:23:11.970
Yeah, so again, um, you know, if you're looking at this and be like, hey, like where do I get started, right?
00:23:11.970 --> 00:23:18.530
I think everybody, and we have some slides coming up here, right, should be using um the free enterprise protected tier.
00:23:18.690 --> 00:23:33.730
So again, as opposed to going to kind of free copilot uh or free chat GPT, right, where you might now be seeing ads or uh you know the the conversations that you're having with the agent are are incorporated back into the model.
00:23:33.970 --> 00:23:43.410
If you go to you know copilot.microsoft.com or or gemini.google.com and sign in with your organizational ID, that becomes protected, right?
00:23:43.410 --> 00:23:49.890
So now you get the enterprise terms of service as opposed to the kind of premium or consumer um terms of service.
00:23:49.890 --> 00:23:58.370
Uh and you can you know basically have those conversations and and kind of you know the the everything is it should be should be protected there, right?
00:23:58.370 --> 00:24:00.130
It's uh QA, right?
00:24:00.130 --> 00:24:05.970
Enhanced search, help build, you know, help you build this policy, review this email, right?
00:24:05.970 --> 00:24:12.050
Those kinds of things that you maybe would have um looked up, you know, used a Google search, right?
00:24:12.050 --> 00:24:20.210
Now you're using uh, you know, that that kind of uh natural language queries for the for the AI.
00:24:20.210 --> 00:24:20.690
Um
00:24:20.690 --> 00:24:26.930
kind of stepping up from there, right, where we see most organizations moving next would be that co-pilot model.
00:24:26.930 --> 00:24:31.650
Again, so that's gonna be about $20 to $30 a month per user.
00:24:31.650 --> 00:24:35.170
Um, some of these do have uh annual terms, right?
00:24:35.170 --> 00:24:39.490
So you may have to commit to pay for a year to get that, um, to get that pricing.
00:24:39.490 --> 00:24:42.850
And again, I think, you know, kind of coming back to the level of investment, right?
00:24:42.850 --> 00:24:46.210
This is a very expensive endeavor.
00:24:46.210 --> 00:24:49.250
And the big players are are still burnt.
00:24:49.250 --> 00:24:51.810
I think, you know, this is a money loser at this point, right?
00:24:51.810 --> 00:24:56.930
They are spending more to build the platform than they are getting in revenue from it at this point.
00:24:56.930 --> 00:24:58.610
Obviously, that's gonna change.
00:24:58.610 --> 00:25:01.090
Um, so again, you know,
00:25:01.090 --> 00:25:07.730
Microsoft, right, they discount 75% for some of the Microsoft um 365 SKUs.
00:25:07.730 --> 00:25:12.130
That discounting does not extend into Copilot to the same degree.
00:25:12.130 --> 00:25:17.890
I think you know, Microsoft is trying to do some of that, but it again, it's it's not discounted, right?
00:25:17.890 --> 00:25:19.810
So you have to pay the full the full price.
00:25:19.970 --> 00:25:41.010
And the benefit you get in these areas would be um now instead of just having kind of better web search or kind of interactive web search or analysis of things that you provide it, um now Copilot or uh the Google Gemini can actually analyze documents and information that are within your cloud environment, right?
00:25:41.010 --> 00:25:47.410
So that it has protected access and it can access everything you can as a user within those platforms.
00:25:47.410 --> 00:26:13.730
And so now uh you can whatever analyze spreadsheets that you have in your organization, you can review policies, uh, you can have it, or you know, like I did, say, hey, I'm getting ready to fill out my APR, analyze all my email that I sent, analyze all the documents I've created over the past year, um, provide a summary of uh some themes of things that I've been a part of, right?
00:26:13.730 --> 00:26:19.410
That's the kind of thing that you can do with those assistive co-pilot um services.
00:26:19.410 --> 00:26:20.690
Um, right.
00:26:20.770 --> 00:26:31.570
And then the next level up, right, would be kind of in the several hundred dollar a month category where you're looking at kind of copilot studio, uh the Gemini AI elite, right?
00:26:31.570 --> 00:26:39.730
And that you know is more of the use cases of you're building um agents, um, you know, probably more extensive code development.
00:26:39.730 --> 00:26:44.610
Uh we see um video generation show up in a lot of these more premium tiers.
00:26:44.610 --> 00:26:56.450
Um again, I don't have very much experience in the um uh in the in the video generation, but I know that again, like you know, video could be an effective method of communicating information, right?
00:26:56.450 --> 00:26:58.450
Everybody's not reading their emails, right?
00:26:58.450 --> 00:27:01.490
So maybe video or these short clips are a way to do that.
00:27:01.490 --> 00:27:11.810
Um and so some of these services really help lower the barrier uh to entry or the sophistication that you need to have to you know really develop compelling online um content.
00:27:12.050 --> 00:27:29.170
So so those are kind of the the you know from a uh yeah basic level, um kind of where we see uh kind of the pricing tiers break down of and I would say, right, this is not something uh, you know, from a licensing perspective, you don't have to license everybody all at once.
00:27:29.170 --> 00:27:39.810
Um, you know, you can kind of just license uh a few people, maybe in a in a working group, and we talk about that in terms of how to do kind of effective AI adoption at an organization.
00:27:40.610 --> 00:27:49.810
We have a quick question in the chat, Matt, about the nominal level there, which is that medium $20 to $30 per user per month.
00:27:49.810 --> 00:27:53.970
Uh, and is that protecting, is that secure for sensitive data?
00:27:53.970 --> 00:27:57.330
So then your data is not used to train the model.
00:27:57.330 --> 00:28:03.650
Your inputs, you can, you know, input like you know, client data if you had it at your nonprofit.
00:28:04.450 --> 00:28:10.530
Yes, that data is you know kind of protected from you know, kind of the enterprise um perspective, right?
00:28:10.530 --> 00:28:17.170
So it's it's you know, the system has access to everything that you do as a as an end user.
00:28:17.170 --> 00:28:24.610
Um and so again, organizations from a policy perspective say, hey, like we don't want to put personally identifiable information into the system, right?
00:28:24.610 --> 00:28:28.050
You can that could be a policy guardrail that you decide.
00:28:28.050 --> 00:28:38.530
Um, but again, the the you know, the enterprise agreements will say this data is yours and you have privacy and control over that, and we're not gonna disclose it.
00:28:38.850 --> 00:28:52.770
And usually when you're logged in, I've noticed in both Copilot and Gemini, there's a little bar across the bottom that will say you're using your enterprise version and this is not being used for the model training.
00:28:52.770 --> 00:28:54.850
So you can also look for that.
00:28:54.850 --> 00:28:56.050
Um,
00:28:56.050 --> 00:28:57.970
but you know, you get what you pay for.
00:28:57.970 --> 00:29:13.810
So can you talk just a little bit more about because I know in the registration questions somebody said, How do I convince my organization to pay for licenses for people instead of just asking people to go to the, you know, go out to ChatGPT and ask their questions.
00:29:13.810 --> 00:29:15.890
So can you talk a little bit about that?
00:29:17.250 --> 00:29:21.650
Um, yeah, I mean, I think if you're asking those questions, like I don't, yeah, I don't know.
00:29:21.650 --> 00:29:29.650
If if you uh, you know, that may be a tough uh sounds like maybe some tough organizational challenges, but I think in general, right.
00:29:29.650 --> 00:29:31.330
I mean, and I think this is I mean,
00:29:31.330 --> 00:29:33.970
this is the equation that that these companies are making, right?
00:29:33.970 --> 00:29:47.410
That you will find, you know, $30 a month worth of value or $200 worth a month of value to pay, you know, to kind of pay for the service because you're gonna get that um back in efficiency, uh, you know, kind of many times over.
00:29:47.410 --> 00:30:25.170
Um so again, uh the key difference that these enterprise um agreements will will give you, right, is the assurance that um the data that is provided into the system um, you know, is not going to be uh then kind of disclosed or kind of made available to other users, right, who may find ways to query and uh elicit information about other organizations or other sensitive topics um through you know through some other kind of sophisticated attacks against the AI, you know, kind of engines themselves.
00:30:25.330 --> 00:30:38.930
And so, you know, making sure that you've got good kind of policies in place that govern how you're using it as an organization, and then providing ongoing training and um engagement to say, yeah, this is the right way to use it.
00:30:38.930 --> 00:30:44.210
Here's how we can use it as an organization, um, and uh, you know, are all important.
00:30:44.210 --> 00:30:51.650
And I would say, you know, the other thing, you know, the asterisk here, right, is you know, uh, you know, I think the you know,
00:30:51.650 --> 00:30:56.370
Asking the question like, what's the most secure AI tool for me to use, right?
00:30:56.370 --> 00:30:58.290
Like maybe isn't the right question.
00:30:58.290 --> 00:31:03.890
Because I think every tool, you know, could be could be used insecurely, right?
00:31:03.890 --> 00:31:23.490
And so I think um you know, you know, you again, like I said, you you may want to uh you know prevent staff from uploading um person identifiable information um into the AI, even if you have you know kind of the enterprise license, right?
00:31:23.490 --> 00:31:28.050
Because you want to have that extra layer of protection around the data that you have.
00:31:28.050 --> 00:31:28.770
Um and so
00:31:28.770 --> 00:31:57.730
I think being intentional about what that means um uh you know for your organization and those guardrails means that your organization needs to, again, understand the data it has, where it lives, who has access to it, um, you know, how it may be uh you know accessed by these other tools, um, you know, just to kind of provide that level of protection and um privacy that you know, maybe your your stakeholders um would would expect.
00:31:58.050 --> 00:32:06.210
Yeah, and you have to like when you have a policy, you do have to be constantly reiterating that to staff and understanding what the policy is.
00:32:06.210 --> 00:32:14.770
Because like Chat GPT has as part of the terms and conditions of using the free one, that you're not allowed to make misleading information.
00:32:14.770 --> 00:32:20.850
And everyone knows that that's like what half of the people are using ChatGPT for all day long, basically.
00:32:20.850 --> 00:32:25.410
So it goes against their terms and conditions, but you know, it's possible for you to do it.
00:32:25.490 --> 00:32:33.810
And so for your staff also, like there may be something that they can do that as an organization goes against your values or against your policies.
00:32:33.810 --> 00:32:44.530
So just making sure that people know what the policies are, and you kind of have to do it over and over because the uh the tools are changing so quickly, and someone will say, Oh, I can use it to do this now.
00:32:44.530 --> 00:32:47.970
And you have to constantly say, but that goes against our policies.
00:32:47.970 --> 00:32:50.210
So please don't use it to do that.
00:32:50.210 --> 00:32:50.770
Um, and
00:32:50.770 --> 00:32:59.090
it is very, very hard to put in place automatic restrictions that keep people from doing the things that you don't want them to be doing.
00:32:59.090 --> 00:33:03.730
So doing a lot of education, I think, is a better way to do it.
00:33:05.410 --> 00:33:09.090
Thank you for joining Community IT for this podcast, part one.
00:33:09.090 --> 00:33:16.450
Subscribe wherever you listen to podcasts and leave us a rating to help others find this leadership resource for nonprofits.
00:33:16.450 --> 00:33:19.250
Listen for part two in your podcast feed.
00:00:23.730 --> 00:00:27.489
Thank you for joining Community IT for this podcast, part one.
00:00:27.489 --> 00:00:34.769
Subscribe wherever you listen to podcasts and leave us a rating to help others find this leadership resource for nonprofits.
00:00:34.769 --> 00:00:37.969
Listen for part two in your podcast feed.
00:00:47.409 --> 00:01:07.170
If you don't know the difference between using a freemium tool like ChatGPT or logging on to a more private enterprise tool at your organization, like Copilot if you're using Microsoft or Gemini if you're using Google Workspace, this webinar is really going to clarify that for you.
00:01:07.170 --> 00:01:16.289
We hear a lot about make sure not to share your sensitive data with the AI learning models, but how do you know that you're using them safely?
00:01:16.289 --> 00:01:18.530
How do you check the terms and conditions?
00:01:18.530 --> 00:01:19.810
Where do you begin?
00:01:19.810 --> 00:01:35.649
Today, our cybersecurity expert Matt Eshleman is going to demystify the enterprise AI and share some tips on how to create an organizational AI policy, how to share knowledge, how to do staff training.
00:01:35.649 --> 00:01:40.530
So all you and all of your colleagues are using AI in the most secure way possible.
00:01:41.170 --> 00:01:42.929
My name is Carolyn Woodard.
00:01:42.929 --> 00:01:45.250
I'm the outreach director for Community IT.
00:01:45.250 --> 00:01:46.929
I'm going to be the moderator today.
00:01:46.929 --> 00:01:50.449
But first, I'm going to go over our learning objectives.
00:01:50.689 --> 00:02:06.609
So by the end of the session today, we hope that you will learn the difference between enterprise or subscription AI and the more freemium, the free AI tools like ChatGPT, if you just Google it and go to the website.
00:02:06.609 --> 00:02:14.770
We're going to go over accessing Microsoft Copilot and Google Workspace Gemini tools at the organizational level.
00:02:14.770 --> 00:02:21.009
We're going to review IT policy guidelines and provide some training and knowledge sharing tips.
00:02:21.009 --> 00:02:28.449
If you have some tips, something that you're doing at your organization that's working, I'm going to have a chance for you to share that with us later.
00:02:28.930 --> 00:02:40.930
If you're looking for more information on AI topics, we just started a midweek nonprofit AI podcast where we give you 10 to 15 minutes of news and resources weekly.
00:02:40.930 --> 00:02:52.449
If you subscribe to our regular podcast, the Technology Topics Podcast, you'll get that in your feed on Tuesdays and our regular Friday podcast on Fridays.
00:02:52.449 --> 00:02:55.889
On Friday, we have lots of different nonprofit IT topics.
00:02:55.889 --> 00:02:57.729
We chat with guests.
00:02:57.729 --> 00:03:09.250
We also have some other recorded webinars which you can access on our site, communityite.com, covering things like creating an ethical AI framework for your organization, AI governance.
00:03:09.250 --> 00:03:16.049
We did an amazing webinar last summer with Brenda Foster on how to use AI kind of in general.
00:03:16.049 --> 00:03:23.729
We have a downloadable AI acceptable use policy template that you can use if you're working on your policy for your organization.
00:03:23.969 --> 00:03:29.650
So we have a lot of those kind of bigger questions, ethical questions in other resources.
00:03:29.650 --> 00:03:38.049
Today we're really going to focus on this specific question that we get a lot, which is how do you know if you're using your AI tools safely?
00:03:38.049 --> 00:03:41.490
And with that, Matt, would you like to introduce yourself?
00:03:42.449 --> 00:03:42.689
Great.
00:03:42.689 --> 00:03:44.769
Well, thanks for that introduction, Carolyn.
00:03:44.769 --> 00:03:46.049
It's great to be here with you.
00:03:46.049 --> 00:03:56.210
And I'm looking forward to talking about adopting AI in your organization and kind of sharing some of those tips and tricks and getting under the hood a little bit.
00:03:56.210 --> 00:04:01.169
As Carolyn mentioned, I'm the chief technology officer here at Community IT.
00:04:01.169 --> 00:04:09.409
And I am pleased that I've just celebrated 24 years officially full-time with Community IT.
00:04:09.409 --> 00:04:16.530
And so get to play a, do a lot of different things, play a lot of different roles, and I'm really excited about this topic in particular.
00:04:17.810 --> 00:04:19.410
And you have so much experience.
00:04:19.410 --> 00:04:30.290
Before we get started with Matt, if you're not familiar with community IT, I want to tell you just a little bit about us.
00:04:30.290 --> 00:04:35.090
We are a 100% employee-owned managed services provider.
00:04:35.090 --> 00:04:37.650
We provide outsourced IT support.
00:04:37.650 --> 00:04:47.409
We work exclusively with nonprofit organizations, and our mission is to help nonprofits accomplish their missions through the effective use of technology.
00:04:47.409 --> 00:04:51.810
We are big fans of what well-managed IT can do for your nonprofit.
00:04:51.810 --> 00:04:54.610
We serve nonprofits across the United States.
00:04:54.610 --> 00:04:56.610
We've been doing this for 25 years.
00:04:56.610 --> 00:04:59.009
This is our 25-year anniversary.
00:04:59.009 --> 00:05:09.009
We are technology experts and are consistently given an MSP 501 recognition for being a top MSP, which is an honor we received again in 2025.
00:05:09.009 --> 00:05:14.529
And we believe that we're the only MSP on the list that serves nonprofits exclusively.
00:05:14.769 --> 00:05:19.730
I want to remind everyone that for these presentations, community IT is vendor agnostic.
00:05:19.730 --> 00:05:24.689
We only make recommendations to our clients and only based on their specific business needs.
00:05:24.689 --> 00:05:30.370
We never try to get a client into a product because we get an incentive or a benefit from it.
00:05:30.370 --> 00:05:34.050
We do consider ourselves a best of breed IT provider.
00:05:34.050 --> 00:05:39.810
So it's our job to know the landscape, what tools are available, reputable, and widely used.
00:05:39.810 --> 00:05:46.689
And we make recommendations on that basis for our clients based on their business needs, priorities, and budget.
00:05:46.689 --> 00:05:56.449
We're going to talk about two big uh nonprofit stack, uh IT stacks that everyone uses, not just nonprofits, Microsoft and Google today.
00:05:56.449 --> 00:05:59.490
Um, but that is because a lot of nonprofits are using them.
00:05:59.730 --> 00:06:02.209
We've got a lot of good questions at registration.
00:06:02.209 --> 00:06:04.610
So we're gonna try and answer as many of those as we can.
00:06:04.610 --> 00:06:17.009
But anything we can't get to, uh, please join us and Matt today on our in our community on Reddit at r slash nonprofit IT management after the webinar for about 30 minutes.
00:06:17.009 --> 00:06:23.969
Uh and he also, Matt also pops in once a week or so to answer any other questions that come in so you can take advantage of that.
00:06:24.209 --> 00:06:25.730
And a little bit more about us.
00:06:25.730 --> 00:06:31.250
Our mission, as I said, is to create value for the nonprofit sector through well-managed IT.
00:06:31.250 --> 00:06:39.730
We also identify four key values as employee owners that define our company: trust, knowledge, service, and balance.
00:06:39.730 --> 00:06:43.329
We seek always to treat people with respect and fairness.
00:06:43.329 --> 00:06:58.689
We seek to empower our staff, clients, and sector to understand and use technology effectively, to be helpful with our talents, and we recognize that the health of our communities is vital to our well-being and that work is only a part of our lives.
00:06:59.089 --> 00:07:03.649
All right, as we usually do, I'm gonna start out with a poll.
00:07:03.649 --> 00:07:07.649
We just want to get a feel for your comfort level with AI tools.
00:07:07.649 --> 00:07:14.850
So your options are one, completely uncomfortable or unfamiliar with most tools.
00:07:14.850 --> 00:07:19.649
Two, somewhat uncomfortable, we use a few popular tools occasionally.
00:07:19.649 --> 00:07:24.370
Three is the neutral, neither uncomfortable nor comfortable kind of average use.
00:07:24.370 --> 00:07:26.610
Four is somewhat comfortable.
00:07:26.610 --> 00:07:29.089
I use a few AI tools daily.
00:07:29.089 --> 00:07:31.329
And five is completely comfortable.
00:07:31.329 --> 00:07:33.569
I use a lot of these tools a lot of the time.
00:07:33.569 --> 00:07:37.649
My colleagues ask me how to, you know, how to use them to teach them how.
00:07:37.649 --> 00:07:41.889
And the six option is uh not applicable or other.
00:07:41.889 --> 00:07:44.769
And Matt, can you see the results?
00:07:45.089 --> 00:07:45.649
Yes, I can.
00:07:45.649 --> 00:07:47.569
I always like it's like a big a big reveal.
00:07:47.569 --> 00:07:50.529
You know, you got the drum roll because I can't see it as it's coming in.
00:07:50.610 --> 00:07:53.329
So all right, so can you go ahead and share it with us?
00:07:53.490 --> 00:07:55.009
How uh yes.
00:07:55.009 --> 00:08:05.329
So, in terms of uh the folks that are um responding today, uh about 12% are completely uh uncomfortable, unfamiliar, haven't really done much at all.
00:08:05.329 --> 00:08:15.730
Um, which maybe on the flip side, right, about 9% of the respondents are completely comfortable and use it a lot, and they're really the resource that people in their organization go to.
00:08:15.730 --> 00:08:18.129
So we kind of have the those those ends.
00:08:18.129 --> 00:08:23.170
And then in the middle, uh we have about 29% of folks are somewhat uncomfortable.
00:08:23.170 --> 00:08:25.730
Maybe they use some of the tools occasionally.
00:08:25.730 --> 00:08:30.529
16% are neither uncomfortable, like right in the middle.
00:08:30.529 --> 00:08:37.490
Uh, and then we have a bigger bump there uh on the folks that are somewhat comfortable use you you using those AI tools um daily.
00:08:37.490 --> 00:08:42.769
So kind of an interesting distribution um in terms of uh the respondents here.
00:08:42.769 --> 00:08:45.490
So we've got kind of bar, what is it, bar barbells?
00:08:45.490 --> 00:08:50.769
We got uh folks on on kind of both ends of the spectrum uh related to their AI usage.
00:08:51.170 --> 00:09:04.450
Yeah, no, that was interesting because often we get that uh bell curve, but uh here we have people who are not as many people who are really neutral of like they're putting themselves either in I use it sometimes or I really don't use it very often.
00:09:04.450 --> 00:09:06.689
So that was interesting.
00:09:06.689 --> 00:09:07.649
Um, all right.
00:09:07.809 --> 00:09:12.289
Well, we heard me say a word a little while ago, freemium AI.
00:09:12.289 --> 00:09:15.169
And Matt, you were gonna talk about what is that.
00:09:16.210 --> 00:09:33.889
Yeah, so I think freemium, and I use that that term intentionally as we were developing this um presentation to really describe how a lot of us maybe have come to you know using AI tools, uh, you know, probably the one being ChatGPT, which we'll take a look at in terms of their overall growth here in the next slide.
00:09:33.889 --> 00:09:35.009
Um, but really, you know,
00:09:35.009 --> 00:09:40.129
Freemium is a it's a business model uh for getting users into the platform.
00:09:40.129 --> 00:09:43.649
Um AI is very expensive to deploy.
00:09:43.649 --> 00:09:47.649
Um, you know, I was looking uh again, the background research, right?
00:09:47.649 --> 00:10:00.129
So I think OpenAI, which is kind of the entity that owns Chat GP GPT, is kind of scheduled to invest like $1.15 trillion in building up their capacity, right?
00:10:00.129 --> 00:10:02.370
So this is enormously expensive.
00:10:02.370 --> 00:10:11.649
Um and so you know, they're they're giving away or starting to give away um access to their tool in the hopes of converting everyone into paid customers later on.
00:10:11.809 --> 00:10:29.490
And I would distinguish that from you know, maybe a public AM model, which would be more of a utility service, maybe something owned by the government, um, right, but as a available as a public good as opposed to a privately held uh AI solution that uh you know is is fundamentally there to kind of make profit.
00:10:29.649 --> 00:10:41.009
So um, you know, all the AI models, I'm you know, and again, this is we're not gonna get into a lot of the um maybe the philosophy or the the you know kind of ethical issues around here.
00:10:41.009 --> 00:10:48.689
Um, but you know, a lot of the these AI um tools were trained using um publicly available data, um, you know, not always with consent.
00:10:48.689 --> 00:10:54.769
Um, you know, it was with uh, you know, so that the you know, some of those things are currently winding their way through the court system.
00:10:54.769 --> 00:10:57.009
Um and you know, again,
00:10:57.009 --> 00:11:08.289
I think with that freemium model, uh, you know, it's maybe a little bit um flippant, but right, if you're not paying for something, right, then then kind of you're the product, you know, could apply in the in this case as well, right?
00:11:08.289 --> 00:11:28.450
So if you don't pay for an AI tool, the content that you're putting into it, the questions you're asking, how you're interacting with it, uh typically is going to go back in to feed the model to provide you know ever increasing or escalating um information about you know about the usage of the tool and what people are are are asking it and learning about.
00:11:28.450 --> 00:11:38.850
So um again, so I think that's a helpful model to keep in mind, uh, that freemium model largely focused on uh you know converting users to paid customers at some point.
00:11:39.090 --> 00:11:45.490
And as I mentioned, um, you know, ChatGPT is I think the common example, and this one um gets bandied about a lot, right?
00:11:45.490 --> 00:11:51.490
So chat GPT was released, I think, officially back in November of 2022.
00:11:51.490 --> 00:12:00.529
Um and I think it had the fastest uh adoption for um, you know, a million active users, you know, measured in weeks.
00:12:00.529 --> 00:12:14.370
And so, you know, just in the last um, you know, several years, right, to two and a half years, you know, now they're up to over 800 million uh monthly active users or weekly active users in terms of people that are interacting with the system, right?
00:12:14.370 --> 00:12:26.210
So it's been a phenomenally um impressive growth in terms of people going to using and interacting with these tools because they are so compelling, I think.
00:12:26.210 --> 00:12:33.090
And so, you know, the use and adoption is really driven by the fact that people are finding it uh helpful.
00:12:33.090 --> 00:12:38.929
And uh I think it really is a you know the disruptive technology change of this time.
00:12:41.009 --> 00:12:47.889
So that covers the free versions, which we are putting in air quotes because they're not free.
00:12:47.889 --> 00:12:52.850
Um can you talk a little bit more about enterprise AI?
00:12:53.809 --> 00:13:17.570
Yeah, so um, you know, a lot of these free, you know, kind of or these um AI tools or uh large language language models that have been developed by these companies or enterprises, again, they typically have uh the model that they are providing, and then they're either you know, there's a free way to access that, in which case, you know, the you know, there there's some, you know, right?
00:13:17.570 --> 00:13:34.210
You may be giving up some privacy or you know, the data you're putting in the system uh gets used to improve the model, and then you know, distinguish that to enterprise AI, where uh that same back-end model maybe has an intermediary layer that protects the information in there.
00:13:34.210 --> 00:13:34.529
And
00:13:34.529 --> 00:13:42.210
I think this is incredibly important for those enterprise customers to ensure that the data that they are putting in the system is private.
00:13:42.210 --> 00:13:46.529
So in the Microsoft world, um, that's called Copilot.
00:13:46.529 --> 00:14:02.689
So um co-pilot is kind of Microsoft's uh kind of business integration or business um intermediary that uh kind of helps to protect the information that is interrogated using the ChatGPT language model.
00:14:02.689 --> 00:14:05.649
Um the same thing for um Gemini, right?
00:14:05.730 --> 00:14:06.610
So there are free versions.
00:14:06.610 --> 00:14:27.250
If you just go to you know kind of Gemini or um Copilot right now in your web browser, you will likely go to a kind of a free version or uh kind of a consumer version where you don't really have those protections and the information that you put in and the system um spits back, right, is kind of recorded and incorporated and using to train the models moving forward.
00:14:27.250 --> 00:14:28.049
Um, right.
00:14:28.129 --> 00:14:43.409
And so uh the enterprise um companies of Google and Microsoft really want to provide a paid version where you can get those protections uh and you know, in turn, you know, have more and more people use the service.
00:14:43.409 --> 00:14:47.809
Um there are also you know lots of other AI um tools out there.
00:14:47.809 --> 00:14:58.689
Again, this isn't an exhaustive survey and an evaluation of kind of what's out there and their relative worth, but again, there are um dedicated solutions built specifically for AI um for nonprofits.
00:14:58.689 --> 00:15:01.809
So change agent AI is one um such tool.
00:15:01.809 --> 00:15:09.009
And then there's other uh tools, again, you know, innumerable uh that are available for um enterprise subscription.
00:15:09.009 --> 00:15:10.610
And there's a couple listed here.
00:15:10.610 --> 00:15:17.169
And again, if you have some of the favorites that you use that you really find valuable, um you can go ahead and put that into the chat.
00:15:17.169 --> 00:15:17.889
And again,
00:15:17.889 --> 00:15:35.809
I think the important thing with all of these, um, you know, and this is really the shift that we've seen going from on-prem server infrastructure to uh kind of these cloud services, is right, it's much more important, I think, for individuals to understand and organizations to understand, right, what's in the terms and conditions, right?
00:15:35.809 --> 00:15:40.929
How can we use this software uh because we don't really control it at all, right?
00:15:40.929 --> 00:15:46.850
It's not in the server room down the hall where we can, you know, kind of see it and touch it and kind of get under the hood, right?
00:15:46.850 --> 00:15:50.769
It's often some far-flung um data center somewhere.
00:15:50.769 --> 00:15:52.210
And uh, you know,
00:15:52.210 --> 00:16:04.689
apart from the terms and conditions that govern that use, um, you know, it's it's really important to investigate do these terms and conditions align with how I expect uh a system to use my information.
00:16:04.689 --> 00:16:06.289
Um, and again, I'll go back, you know,
00:16:06.289 --> 00:16:11.250
I think change agent AI, for example, if you go to their website and look at their terms and conditions, it is very clear.
00:16:11.250 --> 00:16:16.289
Like I think they do a good job of saying how they use your data, what rights you have to it, right?
00:16:16.289 --> 00:16:26.529
And so that's a little bit clearer to review and analyze than looking at the terms and conditions of some of these other big enterprise um players that are a lot more uh obtuse.
00:16:27.090 --> 00:16:27.490
Yeah.
00:16:27.490 --> 00:16:45.970
Yeah, and I I know that I have a friend who uses AI to help understand what the terms and conditions mean and like summarize them for them because it's all you know can be long documents and trying to pull out what it actually says about um how they use your information.
00:16:46.210 --> 00:16:59.409
So um so you also were telling me a little bit about um these kind of this gradation between the different AI tools and what they do and how autonomous they can be.
00:16:59.409 --> 00:17:01.730
So can you talk a little bit about that?
00:17:02.370 --> 00:17:10.850
Yeah, and I think this is really an exciting, you know, evolution of kind of the AI tool set in terms of how we um think of them.
00:17:10.850 --> 00:17:13.490
So, you know, kind of from the timeline perspective, right?
00:17:13.490 --> 00:17:21.570
So we're talking about you know, November uh, you know, kind of 2022, you know, you could ask, you know, Chat GPT some questions, right?
00:17:21.570 --> 00:17:25.090
And maybe augmented or supplanted your internet search, right?
00:17:25.090 --> 00:17:32.850
You wanted to find some information, you're asking questions, you're you know, maybe getting it to review your your uh document, right?
00:17:33.090 --> 00:17:41.730
So that is generative AI um in terms of it being able to kind of uh analyze and kind of create, create quote unquote new things.
00:17:41.730 --> 00:17:42.850
Um and so
00:17:42.850 --> 00:17:53.810
now we kind of see that evolving into maybe these three um categories, and there maybe could be others, but I think for the purposes of our discussion, right, these are the most typical use cases that we see right now.
00:17:53.810 --> 00:17:54.530
So um
00:17:54.530 --> 00:18:01.570
the first category that kind of exists on this continuum of how AI solutions are being used is really the assistive technology model.
00:18:01.570 --> 00:18:05.730
And so um, you know, that would be again, copilot, right?
00:18:05.730 --> 00:18:09.250
Microsoft is as, you know, that's their branding term, right?
00:18:09.250 --> 00:18:15.490
Where uh it is an acts as an assistant to you, the human who remains fully in control, right?
00:18:15.490 --> 00:18:21.890
You're asking it a question, um, it's providing suggestions or drafts or explanations.
00:18:21.890 --> 00:18:29.170
Um, you know, you're giving it a prompt, it's reproviding providing um a response, and then it's done, right?
00:18:29.170 --> 00:18:31.090
It maybe keeps a history, right?
00:18:31.090 --> 00:18:43.090
So you can have a conversation that uh occurs maybe over a few hours or something, but um you know it's you can't go back and ask it, you know, the analysis that it gave you maybe a couple of days ago, right?
00:18:43.090 --> 00:18:44.130
So uh
00:18:44.130 --> 00:18:53.330
in the assistive world, uh the human's kind of in the top in charge, and it's maybe helpful to think of that as maybe like a smart intern sitting next to you.
00:18:53.330 --> 00:18:57.170
So uh, you know, those are the examples that are pretty common.
00:18:57.170 --> 00:19:06.050
Um, and I think we see uh most folks you know taking their their dipping their toe in using that assistive um model, right?
00:19:06.050 --> 00:19:09.330
It's it's relatively low risk, it's easy to adopt.
00:19:09.330 --> 00:19:21.090
Um you know, the the end user is is kind of at really at the end of the day kind of making a decision about yes, do I want to accept this information or um am I going to reject it, right?
00:19:21.090 --> 00:19:25.730
And and kind of as the final arbiter of those uh of those decisions.
00:19:25.730 --> 00:19:28.850
Um so again, that's a great, a great place to get started.
00:19:28.850 --> 00:19:29.970
Um
00:19:29.970 --> 00:19:35.570
in the middle um here is more of like a kind of a scenario-based or workflow.
00:19:35.570 --> 00:19:43.410
Um and again, I think we're seeing these now be um uh embedded maybe into the tools that you are using.
00:19:43.410 --> 00:19:47.970
Um, you know, so uh, you know, there could be some predefined scenarios.
00:19:47.970 --> 00:19:50.370
So it's triggered by an event.
00:19:50.370 --> 00:19:55.970
Um, you know, there's like certain parameters around when that workflow uh would would trigger.
00:19:55.970 --> 00:20:01.810
And um, you know, the mental models as opposed to like an intern sitting next to you, you're asking questions, right?
00:20:01.810 --> 00:20:07.250
It's it's maybe kind of a smart automation with some judgment about how to make uh some decisions, right?
00:20:07.330 --> 00:20:15.090
So for us, uh you know, we're seeing those tools, you know, kind of crop up in uh, for example, like IT ticketing systems, right?
00:20:15.090 --> 00:20:18.770
The AI can look at this, say, oh, like we've seen this before.
00:20:18.770 --> 00:20:24.610
It looks like the solution might be here and and kind of point to um to some examples, right?
00:20:24.610 --> 00:20:34.850
So it's built into a process that exists um already and you know can help speed up uh that uh analysis and uh reporting.
00:20:34.850 --> 00:20:48.530
So again, you know, it could be those customer support or kind of case management um kind of support triage, uh could be maybe some onboarding or offboarding processes, um, could be some compliance reviews, right?
00:20:48.530 --> 00:20:57.970
So it's um really designed to kind of speed uh maybe some of those routine decisions for regularly occurring processes.
00:20:57.970 --> 00:20:59.090
Uh and then
00:20:59.090 --> 00:21:04.130
the final option or the final example here is the uh agentic model.
00:21:04.130 --> 00:21:10.690
And I guess this might be the new term that you know you have to you can't have a webinar without talking about um agentic AI.
00:21:10.690 --> 00:21:16.370
Um, but that is you know more of like an autonomous or semi-autonomous agent, right?
00:21:16.370 --> 00:21:21.570
That is given a goal, and then the agent kind of figures out how to kind of solve that.
00:21:21.570 --> 00:21:23.810
And again, you may have good examples of that, right?
00:21:23.890 --> 00:21:34.450
If you're interacting with customer support, uh those chats, um, right, those are those are almost, I would say, exclusively agentic AI agents at the start, right?
00:21:34.450 --> 00:21:37.490
And and we're seeing you know that that initial interaction, right?
00:21:37.490 --> 00:21:53.330
Where uh you can ask it questions, it's gonna find a response, it's gonna, you know, it has a kind of a body of knowledge that it's drawing on, that it knows how to um to kind of interpret information and um you know respond.
00:21:53.330 --> 00:22:01.650
And you know, it's it represents real automation and real efficiency for um for organizations to you know to kind of take advantage of.
00:22:01.650 --> 00:22:08.610
And so uh, you know, that that kind of exists right as kind of a again, it's operating on its own.
00:22:08.610 --> 00:22:11.250
It's not um assistive, right?
00:22:11.250 --> 00:22:21.890
It's kind of that co pilot model where it's like somebody sitting next to you, you can kind of set up this agenc model uh and it you know has some instructions, it has a library of information to to draw from.
00:22:21.890 --> 00:22:28.530
It has some boundaries and it can you know basically make decisions and provide answers in and of itself.
00:22:28.530 --> 00:22:30.370
So again, um, you know,
00:22:30.370 --> 00:22:35.090
that can be a good solution for um you know that repetitive um operational work.
00:22:35.090 --> 00:22:37.490
Like if you always have questions you always need to answer, right?
00:22:37.490 --> 00:22:41.730
That can be a great way to to incorporate that into your organization.
00:22:41.730 --> 00:22:57.810
Um, you know, whenever you have pretty clear boundaries um over that, and uh you know, you've maybe you've invested already in those um good processes, good documentations, you know, good information to help feed into that to make good um to make good decisions.
00:22:58.290 --> 00:23:03.090
And so what can you wait, what can you expect from what you pay?
00:23:03.090 --> 00:23:06.050
Yeah, nonprofits always want to know.
00:23:06.370 --> 00:23:11.970
Yeah, so again, um, you know, if you're looking at this and be like, hey, like where do I get started, right?
00:23:11.970 --> 00:23:18.530
I think everybody, and we have some slides coming up here, right, should be using um the free enterprise protected tier.
00:23:18.690 --> 00:23:33.730
So again, as opposed to going to kind of free copilot uh or free chat GPT, right, where you might now be seeing ads or uh you know the the conversations that you're having with the agent are are incorporated back into the model.
00:23:33.970 --> 00:23:43.410
If you go to you know copilot.microsoft.com or or gemini.google.com and sign in with your organizational ID, that becomes protected, right?
00:23:43.410 --> 00:23:49.890
So now you get the enterprise terms of service as opposed to the kind of premium or consumer um terms of service.
00:23:49.890 --> 00:23:58.370
Uh and you can you know basically have those conversations and and kind of you know the the everything is it should be should be protected there, right?
00:23:58.370 --> 00:24:00.130
It's uh QA, right?
00:24:00.130 --> 00:24:05.970
Enhanced search, help build, you know, help you build this policy, review this email, right?
00:24:05.970 --> 00:24:12.050
Those kinds of things that you maybe would have um looked up, you know, used a Google search, right?
00:24:12.050 --> 00:24:20.210
Now you're using uh, you know, that that kind of uh natural language queries for the for the AI.
00:24:20.210 --> 00:24:20.690
Um
00:24:20.690 --> 00:24:26.930
kind of stepping up from there, right, where we see most organizations moving next would be that co-pilot model.
00:24:26.930 --> 00:24:31.650
Again, so that's gonna be about $20 to $30 a month per user.
00:24:31.650 --> 00:24:35.170
Um, some of these do have uh annual terms, right?
00:24:35.170 --> 00:24:39.490
So you may have to commit to pay for a year to get that, um, to get that pricing.
00:24:39.490 --> 00:24:42.850
And again, I think, you know, kind of coming back to the level of investment, right?
00:24:42.850 --> 00:24:46.210
This is a very expensive endeavor.
00:24:46.210 --> 00:24:49.250
And the big players are are still burnt.
00:24:49.250 --> 00:24:51.810
I think, you know, this is a money loser at this point, right?
00:24:51.810 --> 00:24:56.930
They are spending more to build the platform than they are getting in revenue from it at this point.
00:24:56.930 --> 00:24:58.610
Obviously, that's gonna change.
00:24:58.610 --> 00:25:01.090
Um, so again, you know,
00:25:01.090 --> 00:25:07.730
Microsoft, right, they discount 75% for some of the Microsoft um 365 SKUs.
00:25:07.730 --> 00:25:12.130
That discounting does not extend into Copilot to the same degree.
00:25:12.130 --> 00:25:17.890
I think you know, Microsoft is trying to do some of that, but it again, it's it's not discounted, right?
00:25:17.890 --> 00:25:19.810
So you have to pay the full the full price.
00:25:19.970 --> 00:25:41.010
And the benefit you get in these areas would be um now instead of just having kind of better web search or kind of interactive web search or analysis of things that you provide it, um now Copilot or uh the Google Gemini can actually analyze documents and information that are within your cloud environment, right?
00:25:41.010 --> 00:25:47.410
So that it has protected access and it can access everything you can as a user within those platforms.
00:25:47.410 --> 00:26:13.730
And so now uh you can whatever analyze spreadsheets that you have in your organization, you can review policies, uh, you can have it, or you know, like I did, say, hey, I'm getting ready to fill out my APR, analyze all my email that I sent, analyze all the documents I've created over the past year, um, provide a summary of uh some themes of things that I've been a part of, right?
00:26:13.730 --> 00:26:19.410
That's the kind of thing that you can do with those assistive co-pilot um services.
00:26:19.410 --> 00:26:20.690
Um, right.
00:26:20.770 --> 00:26:31.570
And then the next level up, right, would be kind of in the several hundred dollar a month category where you're looking at kind of copilot studio, uh the Gemini AI elite, right?
00:26:31.570 --> 00:26:39.730
And that you know is more of the use cases of you're building um agents, um, you know, probably more extensive code development.
00:26:39.730 --> 00:26:44.610
Uh we see um video generation show up in a lot of these more premium tiers.
00:26:44.610 --> 00:26:56.450
Um again, I don't have very much experience in the um uh in the in the video generation, but I know that again, like you know, video could be an effective method of communicating information, right?
00:26:56.450 --> 00:26:58.450
Everybody's not reading their emails, right?
00:26:58.450 --> 00:27:01.490
So maybe video or these short clips are a way to do that.
00:27:01.490 --> 00:27:11.810
Um and so some of these services really help lower the barrier uh to entry or the sophistication that you need to have to you know really develop compelling online um content.
00:27:12.050 --> 00:27:29.170
So so those are kind of the the you know from a uh yeah basic level, um kind of where we see uh kind of the pricing tiers break down of and I would say, right, this is not something uh, you know, from a licensing perspective, you don't have to license everybody all at once.
00:27:29.170 --> 00:27:39.810
Um, you know, you can kind of just license uh a few people, maybe in a in a working group, and we talk about that in terms of how to do kind of effective AI adoption at an organization.
00:27:40.610 --> 00:27:49.810
We have a quick question in the chat, Matt, about the nominal level there, which is that medium $20 to $30 per user per month.
00:27:49.810 --> 00:27:53.970
Uh, and is that protecting, is that secure for sensitive data?
00:27:53.970 --> 00:27:57.330
So then your data is not used to train the model.
00:27:57.330 --> 00:28:03.650
Your inputs, you can, you know, input like you know, client data if you had it at your nonprofit.
00:28:04.450 --> 00:28:10.530
Yes, that data is you know kind of protected from you know, kind of the enterprise um perspective, right?
00:28:10.530 --> 00:28:17.170
So it's it's you know, the system has access to everything that you do as a as an end user.
00:28:17.170 --> 00:28:24.610
Um and so again, organizations from a policy perspective say, hey, like we don't want to put personally identifiable information into the system, right?
00:28:24.610 --> 00:28:28.050
You can that could be a policy guardrail that you decide.
00:28:28.050 --> 00:28:38.530
Um, but again, the the you know, the enterprise agreements will say this data is yours and you have privacy and control over that, and we're not gonna disclose it.
00:28:38.850 --> 00:28:52.770
And usually when you're logged in, I've noticed in both Copilot and Gemini, there's a little bar across the bottom that will say you're using your enterprise version and this is not being used for the model training.
00:28:52.770 --> 00:28:54.850
So you can also look for that.
00:28:54.850 --> 00:28:56.050
Um,
00:28:56.050 --> 00:28:57.970
but you know, you get what you pay for.
00:28:57.970 --> 00:29:13.810
So can you talk just a little bit more about because I know in the registration questions somebody said, How do I convince my organization to pay for licenses for people instead of just asking people to go to the, you know, go out to ChatGPT and ask their questions.
00:29:13.810 --> 00:29:15.890
So can you talk a little bit about that?
00:29:17.250 --> 00:29:21.650
Um, yeah, I mean, I think if you're asking those questions, like I don't, yeah, I don't know.
00:29:21.650 --> 00:29:29.650
If if you uh, you know, that may be a tough uh sounds like maybe some tough organizational challenges, but I think in general, right.
00:29:29.650 --> 00:29:31.330
I mean, and I think this is I mean,
00:29:31.330 --> 00:29:33.970
this is the equation that that these companies are making, right?
00:29:33.970 --> 00:29:47.410
That you will find, you know, $30 a month worth of value or $200 worth a month of value to pay, you know, to kind of pay for the service because you're gonna get that um back in efficiency, uh, you know, kind of many times over.
00:29:47.410 --> 00:30:25.170
Um so again, uh the key difference that these enterprise um agreements will will give you, right, is the assurance that um the data that is provided into the system um, you know, is not going to be uh then kind of disclosed or kind of made available to other users, right, who may find ways to query and uh elicit information about other organizations or other sensitive topics um through you know through some other kind of sophisticated attacks against the AI, you know, kind of engines themselves.
00:30:25.330 --> 00:30:38.930
And so, you know, making sure that you've got good kind of policies in place that govern how you're using it as an organization, and then providing ongoing training and um engagement to say, yeah, this is the right way to use it.
00:30:38.930 --> 00:30:44.210
Here's how we can use it as an organization, um, and uh, you know, are all important.
00:30:44.210 --> 00:30:51.650
And I would say, you know, the other thing, you know, the asterisk here, right, is you know, uh, you know, I think the you know,
00:30:51.650 --> 00:30:56.370
Asking the question like, what's the most secure AI tool for me to use, right?
00:30:56.370 --> 00:30:58.290
Like maybe isn't the right question.
00:30:58.290 --> 00:31:03.890
Because I think every tool, you know, could be could be used insecurely, right?
00:31:03.890 --> 00:31:23.490
And so I think um you know, you know, you again, like I said, you you may want to uh you know prevent staff from uploading um person identifiable information um into the AI, even if you have you know kind of the enterprise license, right?
00:31:23.490 --> 00:31:28.050
Because you want to have that extra layer of protection around the data that you have.
00:31:28.050 --> 00:31:28.770
Um and so
00:31:28.770 --> 00:31:57.730
I think being intentional about what that means um uh you know for your organization and those guardrails means that your organization needs to, again, understand the data it has, where it lives, who has access to it, um, you know, how it may be uh you know accessed by these other tools, um, you know, just to kind of provide that level of protection and um privacy that you know, maybe your your stakeholders um would would expect.
00:31:58.050 --> 00:32:06.210
Yeah, and you have to like when you have a policy, you do have to be constantly reiterating that to staff and understanding what the policy is.
00:32:06.210 --> 00:32:14.770
Because like Chat GPT has as part of the terms and conditions of using the free one, that you're not allowed to make misleading information.
00:32:14.770 --> 00:32:20.850
And everyone knows that that's like what half of the people are using ChatGPT for all day long, basically.
00:32:20.850 --> 00:32:25.410
So it goes against their terms and conditions, but you know, it's possible for you to do it.
00:32:25.490 --> 00:32:33.810
And so for your staff also, like there may be something that they can do that as an organization goes against your values or against your policies.
00:32:33.810 --> 00:32:44.530
So just making sure that people know what the policies are, and you kind of have to do it over and over because the uh the tools are changing so quickly, and someone will say, Oh, I can use it to do this now.
00:32:44.530 --> 00:32:47.970
And you have to constantly say, but that goes against our policies.
00:32:47.970 --> 00:32:50.210
So please don't use it to do that.
00:32:50.210 --> 00:32:50.770
Um, and
00:32:50.770 --> 00:32:59.090
it is very, very hard to put in place automatic restrictions that keep people from doing the things that you don't want them to be doing.
00:32:59.090 --> 00:33:03.730
So doing a lot of education, I think, is a better way to do it.
00:33:05.410 --> 00:33:09.090
Thank you for joining Community IT for this podcast, part one.
00:33:09.090 --> 00:33:16.450
Subscribe wherever you listen to podcasts and leave us a rating to help others find this leadership resource for nonprofits.
00:33:16.450 --> 00:33:19.250
Listen for part two in your podcast feed.
