070 Why Copilot Gives Bad Answers: The Information Problem No One Wants To Admit

WEBVTT
00:00:07.599 --> 00:00:11.279
Hello and welcome back to Information Matters with AI.
00:00:11.759 --> 00:00:26.960
I'm Jacqueline Stockwell, CEO of Leadership Through Data, an author of the international best-selling book, Blooming Good Information Management, and somebody who probably says the family's garbage in, garbage about too many times each day, so many times each week.
00:00:27.039 --> 00:00:29.760
But I'm going to keep saying it because it matters.
00:00:30.000 --> 00:00:34.880
So today's episode is called Why Copilot Gives Bad Answers.
00:00:35.119 --> 00:00:38.000
The information problem no one wants to admit.
00:00:38.159 --> 00:00:44.640
And I love this topic because I love copilot, because it's exactly where everybody is talking around AI.
00:00:44.880 --> 00:00:49.840
But it meets the messy reality of how organizations actually manage information.
00:00:50.320 --> 00:00:57.520
But let's be honest, when copilot gives a poor answer, the first reaction is that AI is absolutely rubbish.
00:00:57.759 --> 00:00:58.560
But is it?
00:00:58.719 --> 00:01:07.599
Or is it copilot simply reflecting the quality, structure, and permissions and all that chaos that's sitting underneath it?
00:01:07.760 --> 00:01:19.120
And this is the conversation that I once had today with my amazing friendal Kankare, who brings a really practical Microsoft 365 and co-pilot lens to this conversation.
00:01:29.439 --> 00:01:33.439
Thank you so much, Jackie, and really glad to be on this podcast with you.
00:01:33.599 --> 00:01:38.719
And Al Kantaria, author of Paras Insula, joining you from Chennai, India.
00:01:38.879 --> 00:01:47.359
I am an independent consultant, and you know, the simplest way to describe what I do is I walk into organizations where you know information is chaos.
00:01:47.840 --> 00:01:48.159
Right?
00:01:48.480 --> 00:01:52.000
Like what Jack mentioned, and I help them get it under control.
00:01:52.319 --> 00:02:05.760
So majority of what I do is within Microsoft 365 purview, retention, records management, sensitivity labels, DLB, e-discovery, DSPM, basically anything and everything that you see at Purview is what I've been doing for years now.
00:02:06.719 --> 00:02:10.400
And what I love is that something new comes up, and I'm like, Canal, do you know this?
00:02:10.479 --> 00:02:11.840
And you're like, yeah, I know that.
00:02:11.919 --> 00:02:16.159
And I'm like, how do you fit all this amazing stuff into your brain?
00:02:16.240 --> 00:02:19.599
And I'm so excited we're gonna be doing some piece of podcasts together.
00:02:19.680 --> 00:02:22.000
So this won't be the first time that you're on the show.
00:02:22.159 --> 00:02:23.919
So let's jump straight into it.
00:02:24.080 --> 00:02:32.319
So when leaders say, Copilot is giving me bad answers, what do you think is usually going on underneath the surface?
00:02:32.639 --> 00:02:34.560
Let's start us with a story, right?
00:02:34.719 --> 00:02:37.840
Because it is easier to tell this via a real story.
00:02:41.919 --> 00:02:48.479
So in I think it was in about 2022, a man in Canada lost an animal.
00:02:49.280 --> 00:02:57.439
And same day, he goes on an airlines website and you know asks that chat bot about the revement fares.
00:02:57.759 --> 00:03:05.280
The bot tells him to book the flight at the full price, and then later on, you know, he can claim that discount back afterwards.
00:03:05.439 --> 00:03:12.000
So he does that, he books the flight, he goes on and comes back and applies for that claim.
00:03:12.080 --> 00:03:12.240
Okay.
00:03:12.639 --> 00:03:14.240
Now here's the part that I love.
00:03:14.560 --> 00:03:19.199
Um the airline says, no, you should have done it while you got the air, the ticket.
00:03:19.759 --> 00:03:23.680
But the chatbot's answer was not the same, right?
00:03:24.000 --> 00:03:35.439
Now, what I assume what might have happened was there were two versions of the same bereavement fare policy sitting on the web sign, and the chatbot actually met the wrong one.
00:03:35.680 --> 00:03:43.120
He sued the airline, and the airline's defense was that the chat model is basically its own thing, and it's not their responsibility.
00:03:43.360 --> 00:03:46.719
And of course, long story short, then they lost that case.
00:03:46.879 --> 00:03:49.439
So this is the whole issue with one story.
00:03:49.680 --> 00:03:51.280
Nobody hacked anything, right?
00:03:51.360 --> 00:03:53.840
There were two versions of a policy.
00:03:54.080 --> 00:03:58.479
The AI ran the wrong one, and the organization had to own the consequences.
00:03:58.639 --> 00:04:05.199
So AI did not make anything up, it ran answers on your whole website, and you had two answers to it.
00:04:05.360 --> 00:04:06.800
So, what would the AI now?
00:04:07.039 --> 00:04:09.919
That's the problem that many have been facing with AI.
00:04:10.000 --> 00:04:13.439
So it's it's not, you know, blame the AI, blame your information.
00:04:13.680 --> 00:04:14.080
Quality.
00:04:14.560 --> 00:04:24.319
The AI is working on the wrong information, it's picked it up, it's given a response, the customer's gone away, and then the airline said, well, actually, it's using that old ad date.
00:04:24.399 --> 00:04:25.279
Let's talk about copilot.
00:04:25.360 --> 00:04:29.519
There's a wide range of AI tools, but we're talking about copilot today.
00:04:29.600 --> 00:04:37.839
So if copilots working on the wrong information, are leaders expecting AI to fix the problems that they've ignored for years?
00:04:38.240 --> 00:04:42.160
So fixing the problem is not copiloting its own thing, right?
00:04:42.319 --> 00:04:49.439
Copilot is looking at the information that it is being presented with, and then you won't trying to understand, make sense out of it.
00:04:49.600 --> 00:04:51.759
And you talk about the gates, right?
00:04:52.000 --> 00:04:55.759
That's exactly what I've seen in a lot of organizations.
00:04:56.319 --> 00:05:05.680
Stale information sitting there and co-pilot, you know, four versions of the same file, and co-pilot does not really know which one is the right one, or which one should I pick.
00:05:05.839 --> 00:05:08.959
And you know, most more often than not what happens.
00:05:09.120 --> 00:05:13.120
This co-pilot tries to pick it's not just co-pilot, right?
00:05:13.279 --> 00:05:14.240
We get EAI.
00:05:16.319 --> 00:05:23.439
It tries to pick the wrong one, and you know, it tries to make sense that okay, this world has been structured in a better way.
00:05:23.759 --> 00:05:24.959
Let's pick this one.
00:05:25.120 --> 00:05:31.920
And it pits the old one because you know, back then we have the time to write things very neatly in a very decisive way.
00:05:32.000 --> 00:05:33.519
So that's that's your problem, right?
00:05:33.600 --> 00:05:38.160
Duplicates are causing that issue, and copilot really does not know which the final question is the final.
00:05:38.639 --> 00:05:44.560
So not a problem create and buy co-pilot, and copilot, not a problem that copilot needs to fix, right?
00:05:44.639 --> 00:05:45.360
It is agreed.
00:05:45.680 --> 00:05:50.399
It is up to the information managers, you know, to pick that up and and start fixing it.
00:05:51.120 --> 00:05:58.800
100% as I said, oh AI is rubbish, it's not, it's actually the information leadership problem, not the technology problem.
00:05:59.040 --> 00:06:07.759
So if the quality of the information is one part of the problem, the next question is what co-pilot can access can actually access, right?
00:06:07.920 --> 00:06:12.480
And this is where permissions and duplicated content comes really, really important.
00:06:12.959 --> 00:06:18.399
How do poor permissions and duplicated content distore co-pilots' results?
00:06:18.720 --> 00:06:21.199
I think access is something that we still need to discuss.
00:06:21.360 --> 00:06:25.600
You know, think of co-pilot as a new starter who arrives from Deva.
00:06:26.000 --> 00:06:31.040
You know, it already holds every key you've ever had it on.
00:06:31.439 --> 00:06:35.040
And unlike the rest of us, co-pilot will never get tired of looking.
00:06:35.279 --> 00:06:35.439
Right?
00:06:35.600 --> 00:06:39.199
It can't open anything you couldn't already open.
00:06:39.360 --> 00:06:44.639
So it just actually goes and looks, which I don't give them, right?
00:06:44.720 --> 00:06:52.079
So for example, let's say I've always had access to a sensitive file that I was given as back in let's say 2014.
00:06:52.240 --> 00:06:55.279
Okay, and I don't know that I have access to a sensitive file.
00:06:55.360 --> 00:07:02.000
It contains a lot of sensitive information, uh, which if it goes in the wrong hands could create a lot of problems for the organization.
00:07:02.480 --> 00:07:10.560
Now, if let's say you're not in the AI ETA yet, so what I would do normally is a simple search for information, right?
00:07:10.639 --> 00:07:13.360
A normal search, and that search wouldn't give me resonance.
00:07:13.680 --> 00:07:15.920
Would be four pages, five pages, six pages.
00:07:16.079 --> 00:07:16.319
Okay.
00:07:16.959 --> 00:07:24.079
And maybe that document that I've always had access to is buried in the fifth page or the sixth page, but I never get there, right?
00:07:24.160 --> 00:07:27.759
I find my answer in the first one or two pages.
00:07:28.079 --> 00:07:32.720
So although I had access, the problem was never magnified.
00:07:32.879 --> 00:07:35.920
Okay, and I never knew I had access to sensitive file.
00:07:36.160 --> 00:07:47.920
Now, what copilot is instead doing or the way AI you know retrieves information, it does not do so by a number of pages that you know and does not use the result as number of pages that you need to look at.
00:07:48.160 --> 00:08:01.439
But it might surface that information right in front of you if co-pilot thinks you know it is the right resource and it's the to to provide the answer to the uh uh person who's written the problem.
00:08:02.160 --> 00:08:13.439
So that problem is now amnified because now suddenly you are seeing answers from a document that was, you know, you had access to in 2014.
00:08:13.759 --> 00:08:21.600
So hidden accesses, accesses that you did not know you have, you know, with given are surfacing right now because of co-finite.
00:08:21.680 --> 00:08:22.800
Now that's the real problem.
00:08:22.879 --> 00:08:26.000
You've always had access, but you never knew you had access.
00:08:26.160 --> 00:08:26.480
Right?
00:08:26.560 --> 00:08:28.399
And now those things are surfacing.
00:08:28.480 --> 00:08:32.080
Yeah, of course, block permissions and everything, that still holds through.
00:08:32.320 --> 00:08:40.240
Everyone's been talking about a lot of blocks, but you know why this is a major problem is because it is surfacing right in front of you now.
00:08:40.399 --> 00:08:42.320
Yeah, it's not buried on the search research.
00:08:43.200 --> 00:08:46.960
Yeah, and it's shining that massive spotlight on something that already exists.
00:08:47.120 --> 00:08:56.320
And if information is overshared, like poorly labelled or stored in the wrong places, like copilot is making it just visible, isn't it?
00:08:56.559 --> 00:08:56.799
Yes.
00:08:57.039 --> 00:08:59.759
And that can be uncomfortable, increase risks.
00:08:59.919 --> 00:09:04.080
But is it something that the board can finally listen to?
00:09:04.240 --> 00:09:07.679
Is it something that we can utilize to say, well, copilot's doing this?
00:09:07.840 --> 00:09:11.519
Listen, this is where we need to get our information management sorted.
00:09:11.679 --> 00:09:20.240
So before an organization goes from a small pilot to a full pilot co-pilot rollout, what should they actually be testing?
00:09:20.799 --> 00:09:27.039
From a testing perspective, I mean, again, it is not like a software that you test, right?
00:09:27.519 --> 00:09:33.039
And here's my most organizations don't just test copilot, they circle it.
00:09:33.200 --> 00:09:44.879
And you like you get a copilot license, you ask it three questions, and it goes test something in the background, gives some gives you something very clever, everyone claps, and you know, that becomes a business case.
00:09:45.120 --> 00:09:47.360
Now, that's not exactly what it should do.
00:09:47.600 --> 00:09:48.960
You need to test it properly.
00:09:49.200 --> 00:09:55.360
Like, you know, there is a testing methodology in the software development lifecycle or the STLC.
00:09:55.600 --> 00:10:00.799
Even Copilot has, you know, kind of steps that you can take to test it out hopefully.
00:10:01.120 --> 00:10:04.879
The first thing that you probably need to do is for knowledge testing.
00:10:05.200 --> 00:10:07.440
Does it actually know your business?
00:10:07.759 --> 00:10:11.919
Take 20 questions that where you know you are confident of the RSS.
00:10:12.240 --> 00:10:18.399
And for example, you could ask what's a defense window, what's the approval limit, uh, what's the current pricing information.
00:10:18.559 --> 00:10:21.120
And out of those answers, score two things.
00:10:21.360 --> 00:10:22.399
Was the answer right?
00:10:22.559 --> 00:10:26.240
And did co-pilot cite the right document?
00:10:26.320 --> 00:10:26.559
Right?
00:10:26.639 --> 00:10:28.080
That's supposed to be the official one.
00:10:28.240 --> 00:10:35.519
So there you do that knowledge testing, and then you kind of know exactly, you know, whether co-pilot is able to get the right information.
00:10:36.559 --> 00:10:39.840
Second, I would do is you know regressive testing.
00:10:40.000 --> 00:10:45.120
Again, this is something that nobody does, and probably the one that hurts the most.
00:10:45.440 --> 00:10:47.679
Your information changes constantly, right?
00:10:47.840 --> 00:10:49.200
You obtain the policy.
00:10:49.840 --> 00:10:52.879
Archivocide, you know, we share a folder.
00:10:53.279 --> 00:11:01.600
So what you need to do is a couple of months later, take those 20 questions again, ask them again co-pilot, and see what results come back.
00:11:01.840 --> 00:11:02.000
Okay.
00:11:02.320 --> 00:11:04.720
If they are still the same, you're good, right?
00:11:04.879 --> 00:11:09.279
And if the policy is updated, are you actually getting the updated results or on?
00:11:09.519 --> 00:11:17.679
So that's you know, you have regression testing where uh your information has undergone changes, and now you're testing co-pilot if co-connect is able to pick up those changes.
00:11:17.919 --> 00:11:30.559
So that's your knowledge, your regression, and then you do something called as you know, edge case testing or adversary testing, where you ask some awkward questions, the one where two departments might disagree, right?
00:11:30.799 --> 00:11:33.360
For example, is this customer profitable or not?
00:11:33.519 --> 00:11:36.080
Or or maybe ask some big questions, right?
00:11:36.320 --> 00:11:39.120
And there is no good answer, right?
00:11:39.200 --> 00:11:42.240
So so so let's say, for example, let me turn this around.
00:11:42.320 --> 00:11:46.960
So so let's say I ask a question and I know the answer does not exist.
00:11:47.120 --> 00:11:55.120
Okay, I want copilot to tell me I can't find an answer to it, and not give me a confident advantage of its own.
00:11:55.600 --> 00:12:05.600
If copilot can do that, it's crazy because I've asked the wrong question, and I know an answer does not exist, copilent should actually tell the guard instead of saying no, this is the answer.
00:12:06.080 --> 00:12:12.480
So that's another thing that you can do, and then of course the security and compliance testing that you have to do anyway, right?
00:12:14.720 --> 00:12:15.840
We'll go deep into that.
00:12:15.919 --> 00:12:17.519
That's the technical part of it.
00:12:17.679 --> 00:12:23.440
But yeah, knowledge, integration, HPS testing, like I mentioned, and security and commands.
00:12:23.519 --> 00:12:27.279
So there is a method by which you can test your co-pilot outcomes.
00:12:27.360 --> 00:12:27.440
Yeah.
00:12:27.679 --> 00:12:28.559
Love that, love that.
00:12:28.639 --> 00:12:30.799
And this gives us uh that testing mindset.
00:12:30.960 --> 00:12:35.120
So now let's look at something leaders can actually do this month, right?
00:12:35.200 --> 00:12:38.480
I'm all for practical things, so what can we do?
00:12:38.639 --> 00:12:46.720
So if we're advising a leadership team that wants a practical 30-day co-pilot readiness check, for example, what would that look like?
00:12:47.600 --> 00:13:02.080
For a leadership team, a 30-day readiness checklist would be, you know, focus on your permissions, try to get those numbers out there, your number of anonymous links, number of files that are pushed.
00:13:02.480 --> 00:13:11.200
Now there are multiple, you know, tools or ways for you, especially in Microsoft 365, to get the uh reports out.
00:13:11.600 --> 00:13:12.799
So act on that.
00:13:13.759 --> 00:13:17.679
Second, I think from a readiness perspective is is copilot ready.
00:13:17.840 --> 00:13:19.679
That's the check that you need, right?
00:13:19.759 --> 00:13:27.360
The ones that we talked about already, whether it is, you know, still surfacing information from stale old documents.
00:13:27.519 --> 00:13:35.519
And have you actually I mean one thing that we always talk about when in terms of uh you know co-pilot readiness is always about data security.
00:13:35.919 --> 00:13:40.080
But what matters even more is retention strategy, right?
00:13:40.240 --> 00:13:44.559
And when I talk about retention strategy, it is also about deletion of information.
00:13:44.879 --> 00:13:45.039
Right?
00:13:45.360 --> 00:13:58.639
If you do not delete information, you have something stale out there, something that does not have any business value, and copilot is actually feeding information of it and giving you so so are you have you done that?
00:13:58.799 --> 00:14:00.080
Have you done those exercises?
00:14:00.159 --> 00:14:01.519
Are your sensitivity labels ready?
00:14:01.679 --> 00:14:03.440
Are your retention labels ready?
00:14:04.399 --> 00:14:22.000
If you are asking a question that you know copilot is, you know, giving or looking at the document that's confidential and answering based on that, test the copilot response also flag that this is a confidential response and all of that.
00:14:22.080 --> 00:14:32.960
And if you have protected documents that co-pilot shouldn't look at, we also need to figure out so these are the steps that you know an organization needs to take to ensure that we are copilot ready.
00:14:33.519 --> 00:14:42.000
So again, a list of steps, I think there is a checklist out there somewhere, but every organization has to make their own because every organization has to be.
00:14:42.159 --> 00:14:52.639
I love that because that really moves the conversation from fear into action, and I think that was um exceptionally good advice for listeners, and it gives them something practical to do.
00:14:52.799 --> 00:15:09.440
It's not a 300-page um strategy that sits in a drawer, it's really focused, it's honest, uh, and really identification as that, yes, you can have something that you know we could suggest or you could find, but you need to tailor tailor it to your own organisation.
00:15:09.600 --> 00:15:14.320
And I think this is where the conversation becomes bigger than IT, right?
00:15:14.639 --> 00:15:20.879
Because copilot readiness is is around leadership, it's around risk and board level conversations.
00:15:21.120 --> 00:15:36.159
So, how can information leaders or anybody using copilot gain board level visibility rather than being seen as the people that say no or slow things down, no, you can't have AI, or talk about records management when everybody else wants AI?
00:15:36.480 --> 00:15:46.080
Okay, that's a great one because you know uh that's why an information manager would probably get about 400 sites, say, at four million, right?
00:15:46.240 --> 00:15:59.600
And if you go to the slide that says, you know, 62% of our files are labeled and 340 sites are remediated, the poor the board would just, you know, nod politely and they say, Yes, thank you, when they move on.
00:15:59.759 --> 00:16:05.519
Nothing really changes because nobody in that room knows what those numbers mean for the business, right?
00:16:05.600 --> 00:16:07.759
So the board is not really interested in that.
00:16:07.919 --> 00:16:11.440
So what you need to focus is on things like exposure, right?
00:16:11.600 --> 00:16:13.919
Not the total numbers, but the direction.
00:16:14.159 --> 00:16:23.440
You can say things like even 12 months ago we had 400 sites that were open to everyone in the company, including two sites where there are code papers there.
00:16:23.600 --> 00:16:27.360
And today it's 90, and those two sites are also recreated.
00:16:27.600 --> 00:16:30.240
So think in trials, right?
00:16:30.480 --> 00:16:32.080
So give them a trailer.
00:16:32.159 --> 00:16:38.639
And if you want them to lean in, add the honest line that will, I mean, will will not IBC.
00:16:39.279 --> 00:16:44.320
Like things like you know, we could not have told you that number a year ago, but now we can.
00:16:44.480 --> 00:16:47.440
Because you're now, you know, trying to get combined ready.
00:16:47.600 --> 00:16:49.519
So so that's your first thing.
00:16:49.679 --> 00:16:52.320
Talk about exposure, talk about readiness.
00:16:52.559 --> 00:16:59.679
And in in terms of readiness, it's like if you're asking those 20 questions that I talked about earlier, right?
00:17:00.080 --> 00:17:03.039
A bunch of people, are you getting the right answers?
00:17:03.200 --> 00:17:10.000
If you're asking those same 20 questions to different departments, different people, are you getting the right answers?
00:17:10.319 --> 00:17:20.240
And then you know, put that test through copilot and see what access it generates, and put them in your slide deck that you have with your code pilot, right?
00:17:20.400 --> 00:17:24.880
Because that will keep you honest, and that will tell you yes, whether we are copilot ready or not.
00:17:24.960 --> 00:17:28.720
Maybe we're co-pilot got let's say 7 out of 20, right?
00:17:28.880 --> 00:17:29.359
It's okay.
00:17:29.839 --> 00:17:30.559
Put it there.
00:17:30.799 --> 00:17:35.039
Maybe after a month you get another chance, you say that now we've got 14 out of 20, right?
00:17:35.359 --> 00:17:37.519
Which is again a trend that we have.
00:17:37.599 --> 00:17:45.119
So show that readiness metric, show that exposure metric, and third one, of course, what the board loves is value, right?
00:17:45.759 --> 00:17:50.319
Be ruthlessly honest about what value will actually have to pilot.
00:17:50.880 --> 00:17:57.200
Like there have been organizations that have just got the licenses and they're just hoping everything will work out.
00:17:57.599 --> 00:17:59.279
Does not work that way, right?
00:18:00.000 --> 00:18:04.240
Just buying licenses will not make your employees more protective.
00:18:04.720 --> 00:18:12.960
In fact, there was, I think, one study done in Australia by uh I don't know whom, but I read it again a couple of uh months ago uh there.
00:18:13.759 --> 00:18:20.079
This organization got co-pilot licenses, and the whatever summaries that copilot will do, for example.
00:18:20.480 --> 00:18:26.480
It was not a copilot disc, it was some other AI uh tool out there, not a copilot to be honest, but something else.
00:18:26.559 --> 00:18:28.880
But anyways, the the result matters.
00:18:29.119 --> 00:18:34.240
What happened was they took a bunch of documents and asked co-pilot to make summaries out of it.
00:18:34.640 --> 00:18:51.200
Okay, they asked humans to do the same, and there were independent reviewers who would do you know analysis of both the results and what the reviewers on that humans were 87% correct, and AI was 4% in their summaries.
00:18:51.839 --> 00:18:57.839
So somewhere you know that and then they were like AI is of no use to us.
00:18:58.240 --> 00:19:01.279
So unless you can prove value, what will never be?
00:19:01.440 --> 00:19:02.640
So you need to prove that value.
00:19:02.720 --> 00:19:08.960
So these three metrics I will say one is your exposure, one is your readiness, and one is value out of your co-pilot.
00:19:10.000 --> 00:19:16.400
It's gently, you know, suddenly, and you've that four minutes are more than enough for you to convey that message.
00:19:16.640 --> 00:19:41.680
I love that because that really gives us such an opportunity to move information leaders to the forefront, saying AI is here, and it's giving us that final visibility to move from back office function to front office function, and changing the language, and we should now be saying here is how we do this safely, confidently, and commercially from an AI perspective.
00:19:42.079 --> 00:19:50.559
So we've talked about information quality, permissions, testing, practical readiness, and leadership visibility.
00:19:50.880 --> 00:19:58.720
My last question to you is I want to finish with the main conversation and putting you in front of a CEO.
00:19:59.039 --> 00:20:00.400
So big question.
00:20:00.640 --> 00:20:06.799
If you had one thing you would tell a CEO before they approved a co-pilot rollout, what would it be?
00:20:07.119 --> 00:20:08.640
Good, good one.
00:20:10.000 --> 00:20:20.640
So so I will tell the CEO back, you know, before we switch it on, you always or or probably the bay you switch it off, and you fall into trouble.
00:20:20.799 --> 00:20:21.920
Don't say it with the no.
00:20:22.799 --> 00:20:24.880
And and it becomes really hard, right?
00:20:25.200 --> 00:20:29.839
Co-pilot or fix your information, it will read it out aloud in front of you.
00:20:30.720 --> 00:20:33.279
So decide now what you want it to say.
00:20:33.519 --> 00:20:39.039
And that's what I'll tell the CEO that you know get co-pilot ready.
00:20:39.519 --> 00:20:44.480
You should be ready for your or in fact, your organization's information should be ready for co-pilot.
00:20:44.640 --> 00:20:47.119
Yeah, that should be the organization.
00:20:47.920 --> 00:20:48.319
Love that.
00:20:48.400 --> 00:20:54.720
And I would say speak to your information manager, records manager, or anybody that manages information with inside your business.
00:20:54.880 --> 00:20:56.000
Right, quick fire around.
00:20:56.160 --> 00:20:58.720
It's a new thing for information matters with AI.
00:20:58.960 --> 00:21:00.240
Gonna ask you a bunch of questions.
00:21:00.319 --> 00:21:02.960
You just have to ask them really quickly off the top of your head.
00:21:03.119 --> 00:21:05.519
So, metadata or folder structures?
00:21:06.079 --> 00:21:07.920
Metadata, any given day.
00:21:08.240 --> 00:21:11.680
Yeah, and don't worry your files in those message folders.
00:21:12.000 --> 00:21:13.920
One answer, canal, one answer.
00:21:14.160 --> 00:21:14.400
Sorry.
00:21:14.480 --> 00:21:14.559
Yeah.
00:21:14.960 --> 00:21:16.000
Biggest co-pilot readin'.
00:21:16.319 --> 00:21:18.400
Very, very passionate about the topic, so yes.
00:21:18.640 --> 00:21:19.359
I know you are.
00:21:19.519 --> 00:21:21.839
Biggest co-pilot readiness myth.
00:21:22.880 --> 00:21:24.079
That we are always ready.
00:21:24.160 --> 00:21:26.480
I mean, that's the myth, right?
00:21:26.880 --> 00:21:27.839
No, you're not ready.
00:21:28.000 --> 00:21:32.559
So figure out your exposure, readiness metric, and your value, business value.
00:21:32.799 --> 00:21:36.000
And then one permission mistake you see all the time.
00:21:37.039 --> 00:21:39.119
Shared with everyone in the organization.
00:21:39.440 --> 00:21:40.079
I feel there.
00:21:40.400 --> 00:21:40.720
Yeah.
00:21:41.119 --> 00:21:44.720
One Microsoft 365 area organization should check first.
00:21:45.039 --> 00:21:48.079
Any given day, even in the context of copilot matters, yes.
00:21:48.720 --> 00:21:49.119
Love that.
00:21:49.279 --> 00:21:52.000
One word you wish leaders understood before buying AR.
00:21:52.400 --> 00:21:53.200
Wow, that's tough.
00:21:53.680 --> 00:21:54.240
Readiness.
00:21:54.400 --> 00:21:55.920
I guess that's that's about it, right?
00:21:56.079 --> 00:21:57.279
Are you copiloting?
00:21:57.440 --> 00:21:58.319
That's always the case.
00:21:58.400 --> 00:21:58.799
Yes.
00:21:59.200 --> 00:21:59.759
Love that.
00:21:59.920 --> 00:22:03.200
So last one, you've got to finish this sentence.
00:22:03.519 --> 00:22:08.799
Information matters because Information matters because it is your groundwell.
00:22:09.200 --> 00:22:09.440
Yeah.
00:22:09.680 --> 00:22:10.160
Love that.
00:22:10.319 --> 00:22:12.000
Thank you so much, Canal.
00:22:12.160 --> 00:22:13.759
It's been absolutely sensational.
00:22:13.920 --> 00:22:15.920
I think the message from today is very clear.
00:22:16.079 --> 00:22:20.240
If copilot is giving poor answers, don't just blame AI.
00:22:20.319 --> 00:22:31.119
Look underneath, look at the quality of the information, the permissions, duplication, ownership, and that lovely word that Kanal loves the most metadata and your governance.
00:22:31.200 --> 00:22:34.240
So AI does not remove the need for information management.
00:22:34.319 --> 00:22:38.160
It makes it more and more visible, more urgent and more strategic.
00:22:38.400 --> 00:22:41.519
And for every information and listening, this is your moment.
00:22:41.599 --> 00:22:42.640
You are not a blocker.
00:22:42.880 --> 00:22:44.319
You are the foundation.
00:22:44.559 --> 00:22:50.400
You are the person who can help your organization use AI with confidence and not chaos.
00:22:50.559 --> 00:23:00.079
If you've liked today's conversation, has made you wonder whether your organization is ready for co-pilot, start with the Leadership Through Data Empower AI Originus Scorecard.
00:23:00.240 --> 00:23:08.720
It will help you see where your strengths are and where your gaps are and what practical action to take next across your information journey.
00:23:08.960 --> 00:23:12.160
You'll find it in the show notes of this podcast.
00:23:12.319 --> 00:23:17.359
You don't need to have all the answers today, but you do need to start asking better questions to your organization.
00:23:17.599 --> 00:23:24.160
Join me and Canal on Wednesday for our LinkedIn Live where we're going to be talking more and more about this.
00:23:24.240 --> 00:23:30.799
And I'm really excited to say that Canal will be joining me over the next couple of weeks to talk more AI topics.
00:23:31.039 --> 00:23:32.640
Thank you so much for your time today.
00:23:32.799 --> 00:23:36.000
Be bold, be brave, and be beautiful.