WEBVTT
00:00:01.919 --> 00:00:08.240
Right now, half of all consumers question the authenticity of almost everything they see online.
00:00:09.039 --> 00:00:11.679
And that segment is growing fast.
00:00:11.839 --> 00:00:16.879
And it's not just some of the things they see online, it's almost everything.
00:00:17.120 --> 00:00:20.719
And this is the world that your company is deploying AI into.
00:00:21.039 --> 00:00:27.839
And most companies have no system in place to prove their AI outputs are real, audited, or accountable.
00:00:28.079 --> 00:00:35.840
Actually, studies say that only 30% of organizations have even reached a mature level of AI governance.
00:00:36.079 --> 00:00:37.280
So, what's all that mean?
00:00:37.439 --> 00:00:46.960
Well, the majority of companies are shoving AI down their employees and their consumers' throats, but they have no way to verify almost anything that they create.
00:00:47.200 --> 00:00:53.280
All that AI text, the AI graphics, the AI apps.
00:00:54.640 --> 00:00:55.520
Is it real?
00:00:55.759 --> 00:00:59.200
And for consumers, that's why it's turned into a trust problem.
00:00:59.439 --> 00:01:02.960
But for companies, it's a responsible AI nightmare.
00:01:03.119 --> 00:01:04.719
And this isn't a future issue.
00:01:04.879 --> 00:01:06.959
This is a right now issue.
00:01:07.120 --> 00:01:13.599
And today we're going to give you the responsible AI playbook to actually deal with it on everyday AI.
00:01:13.840 --> 00:01:14.400
Welcome.
00:01:14.640 --> 00:01:15.599
What's going on?
00:01:15.840 --> 00:01:18.719
Here is the big picture for today's show.
00:01:18.959 --> 00:01:30.159
And as we tackle responsible AI and give you the playbook, tell you what it means and the five uh moves to ensure your strategy survives.
00:01:30.319 --> 00:01:31.280
So here's the big picture.
00:01:31.439 --> 00:01:37.920
Well, responsible AI now determines whether companies can scale AI or just if they're gonna stay stuck.
00:01:38.079 --> 00:01:50.400
And we're talking, it's uh kind of a point now where trust lawsuits and regulation and consumer confidence are just shaping AI's real business value.
00:01:50.480 --> 00:01:56.000
And most companies know the risks that exist, but they don't have systems in place to deal with them.
00:01:56.159 --> 00:01:58.159
It's almost like everyone's just kind of ignoring them.
00:01:58.239 --> 00:02:07.840
And well, that's why many companies are treating responsible AI as, well, just a checkbox that they may or may not check.
00:02:07.920 --> 00:02:16.560
But those that do and that are treating it properly with the uh respect and time that it deserves, well, they're pulling ahead right now.
00:02:16.719 --> 00:02:20.879
So on today's show, stick around for the next 20 minutes, and here's what you're gonna learn.
00:02:21.120 --> 00:02:25.680
So you're gonna learn the real reason most companies can't get past the AI pilot stage.
00:02:25.840 --> 00:02:33.199
You're gonna see why 2026 became the year that AI trust, billion-dollar lawsuits, and regulation collided.
00:02:33.360 --> 00:02:40.639
And I'm gonna give you at the end a five-move responsible AI playbook that separates the leaders from everyone else.
00:02:41.120 --> 00:02:42.560
Let it let's get into it.
00:02:42.719 --> 00:02:46.000
This is the Start Here series on Everyday AI.
00:02:46.159 --> 00:02:52.159
So if you're new here, after 700 plus episodes, everyone's asking, Jordan, where do I start on this journey?
00:02:52.240 --> 00:02:53.280
You have so many podcasts.
00:02:53.360 --> 00:02:57.680
Well, you start with the Start Here series, preferably if you can do it in order.
00:02:57.759 --> 00:02:59.680
Uh, this is now what are we on?
00:02:59.840 --> 00:03:00.879
Volume 17.
00:03:00.960 --> 00:03:07.520
But uh Start Here Series, it is the essential podcast series to both learn the AI basics and to double down on your AI knowledge.
00:03:07.680 --> 00:03:10.639
So if you haven't already, please go to startheseries.com.
00:03:10.800 --> 00:03:15.120
That's going to give you free access to our exclusive inner circle community.
00:03:15.199 --> 00:03:18.000
You literally can't get access right now any other way.
00:03:18.080 --> 00:03:27.520
Uh, and then in the Start Here Series space, there inside our community, you can go listen and watch to uh watch every single uh Start Here series.
00:03:27.599 --> 00:03:33.120
They're fairly quick, it's all in there, and then you can connect with other business leaders who are trying to do the same.
00:03:33.280 --> 00:03:37.599
So if you missed last episode, that was volume 16 of our series.
00:03:37.840 --> 00:03:43.360
We went from chatbots to super agents and talked about the 11 AI categories explained.
00:03:43.439 --> 00:03:44.879
So make sure you go check that one out.
00:03:45.120 --> 00:03:52.400
But today we're talking about the responsible AI playbook, what it means, and five moves to ensure your AI strategy survives.
00:03:52.479 --> 00:04:00.080
Uh, and if you are really new here, uh, this isn't an AI version of me, or if you listen all the time, I just allergies now, right?
00:04:00.240 --> 00:04:04.719
I was sick for uh months and now I sound like this.
00:04:04.879 --> 00:04:05.520
So yeah.
00:04:05.680 --> 00:04:24.079
Uh, anyways, I before we get into responsible AI, I want to talk a little bit on some definitions because it can actually get confusing because uh a lot of times people uh mistake uh or kind of uh cross over between ethical AI and responsible AI.
00:04:24.160 --> 00:04:40.160
So today we're gonna be tackling responsible AI, which is different than ethical AI, because uh I don't think there's really a blanket um kind of statement that you can apply ethical AI because well, ethical AI defines the moral principles of what is right, fairness, safety, and privacy.
00:04:40.319 --> 00:04:43.439
And depending on well, what country you're in, that's a big one.
00:04:43.519 --> 00:04:45.120
We have listeners from all over the world.
00:04:45.199 --> 00:04:53.439
Uh, but also a lot of well, if C-suites are being honest, they look at AI as a headcount thing, right?
00:04:53.519 --> 00:04:56.959
They're like, okay, well, if we deploy AI everywhere, we can reduce headcount by 30%.
00:04:57.279 --> 00:05:00.879
And for them, maybe that's an ethical decision, even though I would argue it's not.
00:05:01.040 --> 00:05:05.839
So we're not gonna really tackle ethical AI today, but we're gonna be talking about responsible AI.
00:05:05.920 --> 00:05:11.199
And that's the operational framework underneath ethical AI that puts those ethical principles into practice.
00:05:11.360 --> 00:05:19.600
So, whatever, whatever those ethical principles are, responsible AI is kind of the framework that you put into place to make sure that those principles work.
00:05:20.000 --> 00:05:24.000
So, ethics essentially is the question of should we do this?
00:05:24.160 --> 00:05:27.439
And responsible AI asks, well, how do we do it right?
00:05:28.079 --> 00:05:30.560
And then if you're like, wait, did you already cover this?
00:05:30.800 --> 00:05:34.160
No, we covered uh AI governance.
00:05:34.240 --> 00:05:36.879
So make sure to go check out volume 14 of the series.
00:05:36.959 --> 00:05:39.279
That was episode 737.
00:05:39.439 --> 00:05:45.680
But essentially, ethics gives the values, tells you what's fair when it comes to AI safe and aligned.
00:05:45.759 --> 00:05:53.439
Uh, and then responsible AI, which we're going to be tackling today, that's how you turn those values into real decisions about use cases, oversight, and data.
00:05:53.519 --> 00:06:01.519
And then governance, which we already covered, that provides the roles, rules, and review systems that enforce responsible AI over time.
00:06:01.680 --> 00:06:10.000
So make sure after you listen to this uh episode, if you haven't already, make sure to go listen to volume 14, that's episode 737 on governance.
00:06:10.160 --> 00:06:17.759
So we're technically zooming out here uh because we already gave you the uh the governance playbook and the five governance rules.
00:06:17.920 --> 00:06:21.839
And this uh episode kind of explains why those rules exist.
00:06:23.680 --> 00:06:26.879
All right, we're gonna start with five and then we're gonna end with five.
00:06:27.040 --> 00:06:36.800
So, first, uh just to get this framework really in place, I want to first go over the five core pillars that hold up every responsible AI decision.
00:06:36.959 --> 00:06:45.199
So that's fairness, transparency or explainability, uh, accountability, privacy or security, and then safety or reliability.
00:06:45.360 --> 00:06:47.040
So, number one, fairness.
00:06:47.120 --> 00:06:54.560
That's how you actively, how your company actively identifies and mitigates any algorithm bias for equitable outcomes, right?
00:06:54.720 --> 00:07:00.720
We have to understand that most large language models that we use are biased.
00:07:00.800 --> 00:07:02.879
And we have to start tackling that issue.
00:07:03.040 --> 00:07:18.959
What systems do you have in place to make sure that your outcomes, whatever those are, those are text uh-based outcomes that you put on your website, maybe they're uh they play into your hiring practices, whatever it is, you need to understand that by default, AI is extremely biased.
00:07:19.120 --> 00:07:25.600
So, uh, how can you make sure that you are putting those systems in place to make sure that AI is equitable and fair?
00:07:25.759 --> 00:07:28.160
Number two, transparency or explainability.
00:07:28.240 --> 00:07:31.519
You need to understand how and why AI reaches its decisions.
00:07:31.680 --> 00:07:41.839
So if you've already listened to the Start Here series, you already know that because we've tackled everything from here's what a large language model is, how it works, you know, cutting through the buzzword and jargon.
00:07:41.920 --> 00:07:44.079
So go back and listen to that one, it'll teach you.
00:07:44.240 --> 00:07:45.680
Then three, accountability.
00:07:45.759 --> 00:07:49.759
That's how you establish clear human responsibility for AI actions and outcomes.
00:07:49.839 --> 00:08:08.079
And we did tackle that one a little bit more in governance, but essentially it's this if the 10-second test, if something goes terribly wrong, if you have an agent crash in your organization and an agent goes and you know does something absolutely terrible within 10 seconds, if someone says, Who's like who's responsible for this?
00:08:08.319 --> 00:08:14.240
Can you with 100% certainty identify the one human being that is ultimately responsible for that?
00:08:14.399 --> 00:08:15.040
Yes or no?
00:08:15.120 --> 00:08:27.279
You need to understand the accountability chain, just as it's as important as it is uh you know to have agentic observability and traceability, you also have to have that on the accountability side with your humans.
00:08:27.439 --> 00:08:29.600
Number four, privacy or security.
00:08:29.680 --> 00:08:34.320
And that's protecting personal data from misuse and adversarial attacks.
00:08:34.399 --> 00:08:38.240
And this becomes even more important as we get into agentic AI, right?
00:08:38.320 --> 00:08:45.600
So as we give AI agents the keys to the castle to make proactive decisions for us, right?
00:08:45.840 --> 00:08:49.679
Um, large language models are no longer reactive read-only.
00:08:49.759 --> 00:08:52.000
They are now proactive read write.
00:08:52.159 --> 00:08:58.320
They are making decisions on our behalf, oftentimes without expert guardrails, right?
00:08:58.480 --> 00:09:07.120
You might have a catch-all safety human in the loop that may or may not have any understanding of how to properly protect personal data.
00:09:07.279 --> 00:09:10.639
Um and then last but not least, number five, safety and reliability.
00:09:10.799 --> 00:09:16.799
You need to ensure in your company that AI performs as intended without causing harm.
00:09:17.039 --> 00:09:17.440
All right.
00:09:17.600 --> 00:09:20.720
And you might say, okay, well, how can AI cause harm?
00:09:20.879 --> 00:09:31.600
Well, if you're only thinking of, you know, using AI, like, oh, I just go to copilot and you know make my uh, you know, make my email sound better, that's that's not what I'm talking about here, right?
00:09:31.759 --> 00:09:33.840
We are in the year 2026.
00:09:34.000 --> 00:09:41.840
And hopefully, if you're an avid listener of this show, you are already deploying right agentic societies uh throughout your organization.
00:09:42.000 --> 00:09:52.480
So I think at that point you will understand uh how AI, how it's important to make sure that your AI is making safe decisions that don't cause humans harm ultimately, right?
00:09:52.639 --> 00:09:57.120
Uh, because as we give more and more agency away, right?
00:09:57.279 --> 00:10:09.759
And that's ultimately uh kind of this this uh I think rough uh 18-month transition that we're gonna be going through mid-2026 to the end of 2027, is human experts giving away their agency, uh, right?
00:10:09.840 --> 00:10:11.600
So you're like, okay, well, what do we do now?
00:10:11.759 --> 00:10:16.639
Well, you have to make sure that the systems that you deploy are safe and reliable.
00:10:17.360 --> 00:10:20.720
So, why is responsible AI even important?
00:10:21.200 --> 00:10:22.639
Why, well, we talked about it.
00:10:22.720 --> 00:10:24.559
It is the trust crisis.
00:10:24.799 --> 00:10:33.360
So, right now, about half of consumers, according to an iProve study, about half of consumers question the authenticity of almost everything they encounter online.
00:10:33.759 --> 00:10:35.120
And I agree, right?
00:10:35.279 --> 00:10:37.519
We actually had a start here series.
00:10:37.600 --> 00:10:41.360
Let me go ahead and and bring it up and tell you uh what episode it was.
00:10:41.519 --> 00:10:48.399
So this was episode 740, it's volume 15 of the Start Here series, called Everything Is Fake, right?
00:10:48.559 --> 00:10:53.519
And how your company can leverage human expertise to fight AI work slop.
00:10:53.759 --> 00:10:58.960
So, consumers, I don't think most people see or understand this yet.
00:10:59.120 --> 00:11:09.519
And as the technology gets better, right, as AI video gets better, uh, as AI images get better, as AI audio in deepfakes, unfortunately, get better.
00:11:09.840 --> 00:11:14.159
That percentage of people questioning the authenticity is only going to rise, right?
00:11:14.240 --> 00:11:21.200
So synthetic media, uh, deepfake fraud, all of those things just erode the baseline trust that people rely on, right?
00:11:21.279 --> 00:11:28.879
It used to be something like, you know, oh, you go on the internet as an example, you go to Google page one, you probably have a lot of trust in that, right?
00:11:28.960 --> 00:11:37.200
If you went to page 20 and it's some random blog from some dude in the basement, you might look at that and be like, not sure about that, right?
00:11:38.639 --> 00:11:43.440
That page 20, take that times 10, right?
00:11:43.600 --> 00:11:48.240
And that's gonna be the commonplace level of distrust that we are gonna see every single day.
00:11:48.320 --> 00:11:55.200
And that's what your right, whatever sector you're in, if you haven't started to experience yet, it is gonna slap you, right?
00:11:55.279 --> 00:11:56.000
It's gonna slap you.
00:11:56.080 --> 00:11:57.759
I think it's starting with social media, right?
00:11:57.840 --> 00:12:02.399
You're seeing more and more things, and you know, things that happen in real life, people are like, oh, that's fake, right?
00:12:02.480 --> 00:12:06.159
And then things that didn't happen in real life, people are like, oh, that's real, right?
00:12:06.320 --> 00:12:08.399
You're starting to see it play out on social media.
00:12:08.559 --> 00:12:10.879
It is going to play out in our everyday lives.
00:12:11.039 --> 00:12:22.879
And I think even things that people start to see with their own eyes, they're gonna start to second guess those things as well, because it is gonna be a come a default to assume that everything you see is fake or AI generated, right?
00:12:23.120 --> 00:12:29.360
Whether it's on a screen or whether you hear uh you know something on the on the radio, uh, right?
00:12:29.679 --> 00:12:31.600
Apparently I'm 90 talking about the radio.
00:12:31.679 --> 00:12:33.519
I still love AM radio, by the way.
00:12:33.600 --> 00:12:44.240
Um, but customers are already shifting toward brands that can prove their outputs are authentic, and that's what you need to understand, and that's why responsible AI is so important.
00:12:44.720 --> 00:12:51.279
And well, aside from just consumer trust, well, you got to pay attention to the regulatory side as well.
00:12:51.600 --> 00:13:01.279
Because uh current there's been a lot of lawsuits, but the Mobly versus Workday certified that as a collective action for AI hiring discrimination.
00:13:01.440 --> 00:13:06.559
So uh a federal court just said the algorithm did it is not a defense anymore.
00:13:06.720 --> 00:13:12.159
So your company, that's why you need to start paying attention to what you produce, right?
00:13:12.320 --> 00:13:26.000
So in this case, right, there's been a lot of different lawsuits, and I think the first ones that uh, you know, kind of rose to prominence are ones that where algorithms made decisions that discriminated against people for whatever reason, uh, right?
00:13:26.159 --> 00:13:30.879
Whether it's hiring, getting a loan, uh, you know, healthcare coverage, et cetera.
00:13:30.960 --> 00:13:39.120
And that's not new, that's been around for technically decades, but it's becoming more and more prevalent as more and more companies that are not necessarily machine learning companies, right?
00:13:39.200 --> 00:13:47.759
But they're just, you know, wrapping their data lake house around uh, you know, a large language model company, and then they're putting that into production.
00:13:47.919 --> 00:13:54.399
That doing that without having the proper safeguards, that is bypassing responsible AI.
00:13:54.639 --> 00:13:58.799
Because now courts are saying that, well, no, you're on the hook for it.
00:13:58.879 --> 00:14:00.399
It's not the AI's fault.
00:14:00.480 --> 00:14:06.720
Uh, you know, so courts are rejecting any distinction between software decision makers and humans make.
00:14:06.879 --> 00:14:13.519
So, right, there's no difference between if I go and make a decision versus if the AI we're using makes a decision.
00:14:13.600 --> 00:14:21.360
And there's now state laws, as an example, in California, Illinois, New York, and Colorado that regulate how you use AI in hiring.
00:14:21.600 --> 00:14:34.799
So we'll see because we obviously know uh here in the US at least, there's uh some federal momentum to try to uh not allow individual states to mandate AI, but there's already state laws.
00:14:34.960 --> 00:14:41.759
They are in place, they are real, and we're gonna talk here in a minute about how it's actually getting bigger than that.
00:14:43.200 --> 00:14:45.679
And yeah, it can be costly.
00:14:45.919 --> 00:14:53.759
It's not just about maybe getting it wrong and maybe not winning back uh, you know, trust or winning trust uh from a potential customer.
00:14:53.919 --> 00:14:55.200
It is real money, right?
00:14:55.360 --> 00:15:00.080
Anthropic, obviously, one of the uh now one of the largest companies in the world.
00:15:00.159 --> 00:15:13.039
Uh Anthropic paid a$1.5 billion uh suit to settle a copyright case with authors, and courts treat AI outputs as relevant to whether they compete with copyrighted works.
00:15:13.200 --> 00:15:18.799
So right now, enterprise users are facing real IP exposure if their vendors trade on unlicensed data.
00:15:18.960 --> 00:15:21.679
So yeah, you need to understand, well, who are you using?
00:15:21.840 --> 00:15:26.879
OpenAI, Google Gemini, Microsoft Copilot, and are you on the hook?
00:15:27.120 --> 00:15:27.440
Right?
00:15:27.600 --> 00:15:30.399
Are you on the hook if something that you create, right?
00:15:30.480 --> 00:15:35.840
Uh let's say a uh, I don't know, a new, a new piece of software, right?
00:15:36.080 --> 00:15:44.879
And uh it's it's something related to health, and you're using uh one of these enterprise tools, but you're using AI to help you build this, right?
00:15:45.279 --> 00:15:48.240
If you get in trouble, do you have protection?
00:15:48.480 --> 00:15:49.200
Do you know?
00:15:49.440 --> 00:15:51.120
These are questions that you have to have.
00:15:51.279 --> 00:15:52.399
It starts, right?
00:15:52.559 --> 00:15:58.879
Starts in the boardroom, but then it it has to trickle down to every single major decision maker in your organization.
00:15:58.960 --> 00:16:00.879
You have to be aware of those things.
00:16:01.120 --> 00:16:05.120
And I think we may start to see some of this hit the fan in August.
00:16:05.279 --> 00:16:09.679
Now, the rollout of the EU AI Act could get pushed back again.
00:16:09.759 --> 00:16:23.200
It's already got pushed back a couple of times, but bookmark it August 2026, that is when the EU AI Act enforcement begins for high-risk AI uh in hiring credits and biometrics.
00:16:23.360 --> 00:16:26.799
So essentially, if your company and you're like, okay, why does this matter?
00:16:26.879 --> 00:16:42.240
Well, if your company does business in the EU, if they're offering business in the EU and you're using AI in that process, which is probably a good majority of our listeners, this applies to you.
00:16:42.399 --> 00:16:52.879
And non-compliance with these high-risk AI rules in the EUI and the EU AI Act carries fines of up to 35 million euros.
00:16:53.120 --> 00:16:55.840
Or here's here's a wild one, right?
00:16:55.919 --> 00:16:59.600
If you're a large company, you're like, okay, 35 million euros, right?
00:16:59.679 --> 00:17:04.799
That's our catering costs, or 7% of global revenue.
00:17:05.279 --> 00:17:07.680
Oh, now it's got your attention, right?
00:17:08.240 --> 00:17:15.759
So if you have completely bypassed responsible AI and governance, right?
00:17:15.839 --> 00:17:17.680
But that's later down the line, right?
00:17:17.759 --> 00:17:23.359
If you've bypassed responsible AI, the clock is quite literally ticking.
00:17:23.920 --> 00:17:24.240
Right?
00:17:24.400 --> 00:17:27.359
We'll see if anything like this ever catches on in the US.
00:17:27.519 --> 00:17:35.920
I think maybe at the state level it will, but then, like I've talked about on the show, if a federal judge throws it out, well, that will be the deciding factor, right?
00:17:36.000 --> 00:17:39.039
It's probably gonna get uh litigated at some point.
00:17:39.359 --> 00:17:47.039
But if you do business in Europe, which is many, many companies, and you're using AI, you could be on the hook.
00:17:47.119 --> 00:17:53.119
If you go, if you are found non-compliant with any of those high-risk AI rules.
00:17:54.000 --> 00:18:02.559
It's not just about enforcement, it's not just about doing uh, well, we better be responsible uh with AI so we don't get in trouble.
00:18:02.799 --> 00:18:11.039
No, being responsible, yes, it creates trust, but responsible AI leads to higher ROI on AI.
00:18:11.200 --> 00:18:21.440
So, according to McKinsey, companies that are investing heavily in responsible AI report over a 5% profit impact versus those companies that aren't.
00:18:21.599 --> 00:18:27.680
And senior leadership involvement in AI governance drives significantly greater business value.
00:18:27.839 --> 00:18:41.119
So if your business leaders are heavily involved in the process and they understand the basics of responsible AI and they're making those decisions from top to bottom, it's not just about not getting in trouble, right?
00:18:41.599 --> 00:18:49.920
This is foundational for growing your organization because it unlocks real scale and real returns.
00:18:50.640 --> 00:18:53.440
So let's start to wrap with this.
00:18:53.599 --> 00:18:58.720
We started with the five pillars that kind of helped define what responsible AI is.
00:18:58.880 --> 00:19:07.519
Now, here is the playbook, the five moves to ensure your responsible AI strategy survives.
00:19:08.319 --> 00:19:14.000
And again, make sure you go uh check out the governance episode because a lot of these things are aligned.
00:19:14.079 --> 00:19:14.799
But here we go.
00:19:14.960 --> 00:19:17.119
Uh, rule number one, know what you have.
00:19:17.200 --> 00:19:21.519
You need to audit every AI system and classify each one by risk level.
00:19:21.680 --> 00:19:25.519
The cool thing is the EU AI Act actually put those risk levels out.
00:19:25.680 --> 00:19:28.079
Number two, assign real ownership.
00:19:28.160 --> 00:19:32.079
You need to be able to name the individual with the authority, budget, and accountability.
00:19:32.160 --> 00:19:36.000
That 10-second rule when someone goes wrong, who is the human responsible?
00:19:36.160 --> 00:19:39.039
You have to know with the utmost certainty.
00:19:39.279 --> 00:19:43.440
Number three, you need to audit for bias before the courts do.
00:19:43.599 --> 00:19:57.680
You need to test your AI tools against your actual data and to see, well, is there any leaks that could cause legal troubles, whether it's against the uh the EU high risk or uh any of the state laws here in the US?
00:19:57.839 --> 00:20:01.440
Number four, you need to build expert-driven oversight.
00:20:01.519 --> 00:20:04.319
If you listen to the show, you know I hate human in the loop.
00:20:04.480 --> 00:20:15.039
You need to have expert-driven oversight, you need to have proactively your domain experts driving your AI, uh responsible AI policy.
00:20:15.200 --> 00:20:23.039
So that is domain professionals reviewing outputs, not just well, bill an IT rubber stamping your new agentic workforce.
00:20:23.200 --> 00:20:28.319
And then number five, you need to trade treat transparency as a competitive advantage.
00:20:28.480 --> 00:20:33.359
You should be disclosing when and how AI is involved, uh, right?
00:20:33.519 --> 00:20:41.759
Not just to your consumers, but to your employees, to your stakeholders, both internal and external, right?
00:20:42.000 --> 00:20:50.640
When we saw that uh consumers are starting to make decisions on those companies that are transparent in AI, right?
00:20:50.880 --> 00:20:52.240
Like you kind of think back.
00:20:52.319 --> 00:20:54.480
I think that there's been this kind of boom.
00:20:54.640 --> 00:20:56.880
Maybe it's a generational thing, I don't know.
00:20:56.960 --> 00:20:57.519
Uh, right.
00:20:57.599 --> 00:21:02.160
But when you think of like organic and you know, food with clean ingredients, right?
00:21:02.319 --> 00:21:13.200
You can see over the last 10 years how these types of brands have exploded because they're being transparent about what you are consuming and they're saying what's real, what's fake.
00:21:13.519 --> 00:21:15.200
Think of the same thing, right?
00:21:15.440 --> 00:21:20.000
With what you are putting out there to the world, with what your company is putting out there to the world.
00:21:20.160 --> 00:21:31.920
Trust me, the trust, the trust crisis has barely started, and it is a much more significant issue than much than most people realize.
00:21:32.079 --> 00:21:41.119
And it's something that if you are like, uh, well, we'll we'll worry about this when we know it's a problem, when we hear rumblings of it, it's gonna be too late at that point, right?
00:21:41.200 --> 00:21:43.440
Because your competitors are already gonna be doing this.
00:21:43.519 --> 00:21:47.039
So I think the fifth one here uh is huge.
00:21:47.200 --> 00:21:52.319
You have to become overly transparent now as a competitive advantage.
00:21:52.480 --> 00:22:04.880
And again, it's not just uh how you communicate these things externally or to you know external stakeholders, to uh current clients, potential customers, etc.
00:22:05.200 --> 00:22:09.279
It's also how you communicate these things in your own organization, right?
00:22:09.839 --> 00:22:12.400
How do you deploy AI agents?
00:22:13.440 --> 00:22:16.000
Do all of your staff know?
00:22:16.160 --> 00:22:17.279
Do all of your leaders know?
00:22:17.440 --> 00:22:22.000
Do frontline workers know how these decisions are ultimately being made, right?
00:22:22.079 --> 00:22:24.720
When they go click that, you know, agent run?
00:22:25.279 --> 00:22:26.000
Probably not.
00:22:26.559 --> 00:22:34.319
Transparency starts at home and then it is going to expand uh right to your neighborhood or your potential clients.
00:22:35.279 --> 00:22:38.640
So as we wrap here, here's what I want to leave you with.
00:22:39.759 --> 00:22:42.000
The same thing with governance, right?
00:22:42.079 --> 00:22:46.160
The governance are the rules in which you carry this out, but responsible AI, right?
00:22:46.240 --> 00:22:48.799
When you think about ethics, what's right, what's wrong.
00:22:48.960 --> 00:22:55.279
And then you think about how we are actually uh, you know, the foundational rules that help set this.
00:22:56.240 --> 00:23:03.440
Remember that responsible AI turns your company's AI values into real decisions.
00:23:03.599 --> 00:23:07.279
It is the foundation on how you look at use cases, data, and oversight.
00:23:07.359 --> 00:23:11.599
And it also sets the pace for how you will ultimately govern AI.
00:23:11.839 --> 00:23:16.079
But responsible AI helps you technically go faster.
00:23:16.240 --> 00:23:20.960
In the same way people think governance slows your AI efforts down.
00:23:21.200 --> 00:23:26.720
Well, responsible AI doesn't slow your efforts down either because it builds trust.
00:23:26.880 --> 00:23:32.240
And trust uh builds the roads, it builds the highway for speed, right?
00:23:32.400 --> 00:23:39.839
Without those lanes uh of responsible AI, you're just gonna have your cars going all over the place, right?
00:23:40.000 --> 00:23:42.079
So trust is that infrastructure.
00:23:42.160 --> 00:23:48.160
The ethics provides the values, responsible AIs provide the decisions, and governance enforces both.
00:23:48.240 --> 00:23:57.519
And the companies that are earning trust now, while distrust is on the rise, will own the competitive advantage in 2027.
00:23:57.839 --> 00:23:59.279
All right, that's a wrap, y'all.
00:23:59.359 --> 00:24:03.519
Thanks for sticking with me as I can barely uh you know talk.
00:24:03.599 --> 00:24:05.200
It feels like my nose is plugged.
00:24:05.279 --> 00:24:08.720
Hopefully, at one point here, I'm not gonna be sick or have crazy allergies.
00:24:08.799 --> 00:24:09.839
So you can all actually hear me.
00:24:09.920 --> 00:24:10.880
But thank you for sticking around.
00:24:10.960 --> 00:24:16.160
And I hope this was helpful as we tackled the responsible AI playbook.
00:24:16.240 --> 00:24:20.480
Uh, and if this was helpful, make sure you go to starthearseries.com.
00:24:20.559 --> 00:24:25.039
That is gonna give you free access to our exclusive inner circle community.
00:24:25.119 --> 00:24:26.880
You're not gonna find it anywhere else.
00:24:26.960 --> 00:24:33.599
Uh, and if you haven't already, please do me a favor go subscribe to the podcast on Spotify or Apple.
00:24:33.759 --> 00:24:34.880
Leave us a rating if you could.
00:24:34.960 --> 00:24:35.839
I'd appreciate that.
00:24:36.000 --> 00:24:36.960
Thank you for tuning in.
00:24:37.039 --> 00:24:39.599
Hope to see you back tomorrow and every day for more.
00:24:39.680 --> 00:24:40.720
Every day, a lot.
00:24:40.799 --> 00:24:41.839
Thanks, y'all.
00:00:01.919 --> 00:00:08.240
Right now, half of all consumers question the authenticity of almost everything they see online.
00:00:09.039 --> 00:00:11.679
And that segment is growing fast.
00:00:11.839 --> 00:00:16.879
And it's not just some of the things they see online, it's almost everything.
00:00:17.120 --> 00:00:20.719
And this is the world that your company is deploying AI into.
00:00:21.039 --> 00:00:27.839
And most companies have no system in place to prove their AI outputs are real, audited, or accountable.
00:00:28.079 --> 00:00:35.840
Actually, studies say that only 30% of organizations have even reached a mature level of AI governance.
00:00:36.079 --> 00:00:37.280
So, what's all that mean?
00:00:37.439 --> 00:00:46.960
Well, the majority of companies are shoving AI down their employees and their consumers' throats, but they have no way to verify almost anything that they create.
00:00:47.200 --> 00:00:53.280
All that AI text, the AI graphics, the AI apps.
00:00:54.640 --> 00:00:55.520
Is it real?
00:00:55.759 --> 00:00:59.200
And for consumers, that's why it's turned into a trust problem.
00:00:59.439 --> 00:01:02.960
But for companies, it's a responsible AI nightmare.
00:01:03.119 --> 00:01:04.719
And this isn't a future issue.
00:01:04.879 --> 00:01:06.959
This is a right now issue.
00:01:07.120 --> 00:01:13.599
And today we're going to give you the responsible AI playbook to actually deal with it on everyday AI.
00:01:13.840 --> 00:01:14.400
Welcome.
00:01:14.640 --> 00:01:15.599
What's going on?
00:01:15.840 --> 00:01:18.719
Here is the big picture for today's show.
00:01:18.959 --> 00:01:30.159
And as we tackle responsible AI and give you the playbook, tell you what it means and the five uh moves to ensure your strategy survives.
00:01:30.319 --> 00:01:31.280
So here's the big picture.
00:01:31.439 --> 00:01:37.920
Well, responsible AI now determines whether companies can scale AI or just if they're gonna stay stuck.
00:01:38.079 --> 00:01:50.400
And we're talking, it's uh kind of a point now where trust lawsuits and regulation and consumer confidence are just shaping AI's real business value.
00:01:50.480 --> 00:01:56.000
And most companies know the risks that exist, but they don't have systems in place to deal with them.
00:01:56.159 --> 00:01:58.159
It's almost like everyone's just kind of ignoring them.
00:01:58.239 --> 00:02:07.840
And well, that's why many companies are treating responsible AI as, well, just a checkbox that they may or may not check.
00:02:07.920 --> 00:02:16.560
But those that do and that are treating it properly with the uh respect and time that it deserves, well, they're pulling ahead right now.
00:02:16.719 --> 00:02:20.879
So on today's show, stick around for the next 20 minutes, and here's what you're gonna learn.
00:02:21.120 --> 00:02:25.680
So you're gonna learn the real reason most companies can't get past the AI pilot stage.
00:02:25.840 --> 00:02:33.199
You're gonna see why 2026 became the year that AI trust, billion-dollar lawsuits, and regulation collided.
00:02:33.360 --> 00:02:40.639
And I'm gonna give you at the end a five-move responsible AI playbook that separates the leaders from everyone else.
00:02:41.120 --> 00:02:42.560
Let it let's get into it.
00:02:42.719 --> 00:02:46.000
This is the Start Here series on Everyday AI.
00:02:46.159 --> 00:02:52.159
So if you're new here, after 700 plus episodes, everyone's asking, Jordan, where do I start on this journey?
00:02:52.240 --> 00:02:53.280
You have so many podcasts.
00:02:53.360 --> 00:02:57.680
Well, you start with the Start Here series, preferably if you can do it in order.
00:02:57.759 --> 00:02:59.680
Uh, this is now what are we on?
00:02:59.840 --> 00:03:00.879
Volume 17.
00:03:00.960 --> 00:03:07.520
But uh Start Here Series, it is the essential podcast series to both learn the AI basics and to double down on your AI knowledge.
00:03:07.680 --> 00:03:10.639
So if you haven't already, please go to startheseries.com.
00:03:10.800 --> 00:03:15.120
That's going to give you free access to our exclusive inner circle community.
00:03:15.199 --> 00:03:18.000
You literally can't get access right now any other way.
00:03:18.080 --> 00:03:27.520
Uh, and then in the Start Here Series space, there inside our community, you can go listen and watch to uh watch every single uh Start Here series.
00:03:27.599 --> 00:03:33.120
They're fairly quick, it's all in there, and then you can connect with other business leaders who are trying to do the same.
00:03:33.280 --> 00:03:37.599
So if you missed last episode, that was volume 16 of our series.
00:03:37.840 --> 00:03:43.360
We went from chatbots to super agents and talked about the 11 AI categories explained.
00:03:43.439 --> 00:03:44.879
So make sure you go check that one out.
00:03:45.120 --> 00:03:52.400
But today we're talking about the responsible AI playbook, what it means, and five moves to ensure your AI strategy survives.
00:03:52.479 --> 00:04:00.080
Uh, and if you are really new here, uh, this isn't an AI version of me, or if you listen all the time, I just allergies now, right?
00:04:00.240 --> 00:04:04.719
I was sick for uh months and now I sound like this.
00:04:04.879 --> 00:04:05.520
So yeah.
00:04:05.680 --> 00:04:24.079
Uh, anyways, I before we get into responsible AI, I want to talk a little bit on some definitions because it can actually get confusing because uh a lot of times people uh mistake uh or kind of uh cross over between ethical AI and responsible AI.
00:04:24.160 --> 00:04:40.160
So today we're gonna be tackling responsible AI, which is different than ethical AI, because uh I don't think there's really a blanket um kind of statement that you can apply ethical AI because well, ethical AI defines the moral principles of what is right, fairness, safety, and privacy.
00:04:40.319 --> 00:04:43.439
And depending on well, what country you're in, that's a big one.
00:04:43.519 --> 00:04:45.120
We have listeners from all over the world.
00:04:45.199 --> 00:04:53.439
Uh, but also a lot of well, if C-suites are being honest, they look at AI as a headcount thing, right?
00:04:53.519 --> 00:04:56.959
They're like, okay, well, if we deploy AI everywhere, we can reduce headcount by 30%.
00:04:57.279 --> 00:05:00.879
And for them, maybe that's an ethical decision, even though I would argue it's not.
00:05:01.040 --> 00:05:05.839
So we're not gonna really tackle ethical AI today, but we're gonna be talking about responsible AI.
00:05:05.920 --> 00:05:11.199
And that's the operational framework underneath ethical AI that puts those ethical principles into practice.
00:05:11.360 --> 00:05:19.600
So, whatever, whatever those ethical principles are, responsible AI is kind of the framework that you put into place to make sure that those principles work.
00:05:20.000 --> 00:05:24.000
So, ethics essentially is the question of should we do this?
00:05:24.160 --> 00:05:27.439
And responsible AI asks, well, how do we do it right?
00:05:28.079 --> 00:05:30.560
And then if you're like, wait, did you already cover this?
00:05:30.800 --> 00:05:34.160
No, we covered uh AI governance.
00:05:34.240 --> 00:05:36.879
So make sure to go check out volume 14 of the series.
00:05:36.959 --> 00:05:39.279
That was episode 737.
00:05:39.439 --> 00:05:45.680
But essentially, ethics gives the values, tells you what's fair when it comes to AI safe and aligned.
00:05:45.759 --> 00:05:53.439
Uh, and then responsible AI, which we're going to be tackling today, that's how you turn those values into real decisions about use cases, oversight, and data.
00:05:53.519 --> 00:06:01.519
And then governance, which we already covered, that provides the roles, rules, and review systems that enforce responsible AI over time.
00:06:01.680 --> 00:06:10.000
So make sure after you listen to this uh episode, if you haven't already, make sure to go listen to volume 14, that's episode 737 on governance.
00:06:10.160 --> 00:06:17.759
So we're technically zooming out here uh because we already gave you the uh the governance playbook and the five governance rules.
00:06:17.920 --> 00:06:21.839
And this uh episode kind of explains why those rules exist.
00:06:23.680 --> 00:06:26.879
All right, we're gonna start with five and then we're gonna end with five.
00:06:27.040 --> 00:06:36.800
So, first, uh just to get this framework really in place, I want to first go over the five core pillars that hold up every responsible AI decision.
00:06:36.959 --> 00:06:45.199
So that's fairness, transparency or explainability, uh, accountability, privacy or security, and then safety or reliability.
00:06:45.360 --> 00:06:47.040
So, number one, fairness.
00:06:47.120 --> 00:06:54.560
That's how you actively, how your company actively identifies and mitigates any algorithm bias for equitable outcomes, right?
00:06:54.720 --> 00:07:00.720
We have to understand that most large language models that we use are biased.
00:07:00.800 --> 00:07:02.879
And we have to start tackling that issue.
00:07:03.040 --> 00:07:18.959
What systems do you have in place to make sure that your outcomes, whatever those are, those are text uh-based outcomes that you put on your website, maybe they're uh they play into your hiring practices, whatever it is, you need to understand that by default, AI is extremely biased.
00:07:19.120 --> 00:07:25.600
So, uh, how can you make sure that you are putting those systems in place to make sure that AI is equitable and fair?
00:07:25.759 --> 00:07:28.160
Number two, transparency or explainability.
00:07:28.240 --> 00:07:31.519
You need to understand how and why AI reaches its decisions.
00:07:31.680 --> 00:07:41.839
So if you've already listened to the Start Here series, you already know that because we've tackled everything from here's what a large language model is, how it works, you know, cutting through the buzzword and jargon.
00:07:41.920 --> 00:07:44.079
So go back and listen to that one, it'll teach you.
00:07:44.240 --> 00:07:45.680
Then three, accountability.
00:07:45.759 --> 00:07:49.759
That's how you establish clear human responsibility for AI actions and outcomes.
00:07:49.839 --> 00:08:08.079
And we did tackle that one a little bit more in governance, but essentially it's this if the 10-second test, if something goes terribly wrong, if you have an agent crash in your organization and an agent goes and you know does something absolutely terrible within 10 seconds, if someone says, Who's like who's responsible for this?
00:08:08.319 --> 00:08:14.240
Can you with 100% certainty identify the one human being that is ultimately responsible for that?
00:08:14.399 --> 00:08:15.040
Yes or no?
00:08:15.120 --> 00:08:27.279
You need to understand the accountability chain, just as it's as important as it is uh you know to have agentic observability and traceability, you also have to have that on the accountability side with your humans.
00:08:27.439 --> 00:08:29.600
Number four, privacy or security.
00:08:29.680 --> 00:08:34.320
And that's protecting personal data from misuse and adversarial attacks.
00:08:34.399 --> 00:08:38.240
And this becomes even more important as we get into agentic AI, right?
00:08:38.320 --> 00:08:45.600
So as we give AI agents the keys to the castle to make proactive decisions for us, right?
00:08:45.840 --> 00:08:49.679
Um, large language models are no longer reactive read-only.
00:08:49.759 --> 00:08:52.000
They are now proactive read write.
00:08:52.159 --> 00:08:58.320
They are making decisions on our behalf, oftentimes without expert guardrails, right?
00:08:58.480 --> 00:09:07.120
You might have a catch-all safety human in the loop that may or may not have any understanding of how to properly protect personal data.
00:09:07.279 --> 00:09:10.639
Um and then last but not least, number five, safety and reliability.
00:09:10.799 --> 00:09:16.799
You need to ensure in your company that AI performs as intended without causing harm.
00:09:17.039 --> 00:09:17.440
All right.
00:09:17.600 --> 00:09:20.720
And you might say, okay, well, how can AI cause harm?
00:09:20.879 --> 00:09:31.600
Well, if you're only thinking of, you know, using AI, like, oh, I just go to copilot and you know make my uh, you know, make my email sound better, that's that's not what I'm talking about here, right?
00:09:31.759 --> 00:09:33.840
We are in the year 2026.
00:09:34.000 --> 00:09:41.840
And hopefully, if you're an avid listener of this show, you are already deploying right agentic societies uh throughout your organization.
00:09:42.000 --> 00:09:52.480
So I think at that point you will understand uh how AI, how it's important to make sure that your AI is making safe decisions that don't cause humans harm ultimately, right?
00:09:52.639 --> 00:09:57.120
Uh, because as we give more and more agency away, right?
00:09:57.279 --> 00:10:09.759
And that's ultimately uh kind of this this uh I think rough uh 18-month transition that we're gonna be going through mid-2026 to the end of 2027, is human experts giving away their agency, uh, right?
00:10:09.840 --> 00:10:11.600
So you're like, okay, well, what do we do now?
00:10:11.759 --> 00:10:16.639
Well, you have to make sure that the systems that you deploy are safe and reliable.
00:10:17.360 --> 00:10:20.720
So, why is responsible AI even important?
00:10:21.200 --> 00:10:22.639
Why, well, we talked about it.
00:10:22.720 --> 00:10:24.559
It is the trust crisis.
00:10:24.799 --> 00:10:33.360
So, right now, about half of consumers, according to an iProve study, about half of consumers question the authenticity of almost everything they encounter online.
00:10:33.759 --> 00:10:35.120
And I agree, right?
00:10:35.279 --> 00:10:37.519
We actually had a start here series.
00:10:37.600 --> 00:10:41.360
Let me go ahead and and bring it up and tell you uh what episode it was.
00:10:41.519 --> 00:10:48.399
So this was episode 740, it's volume 15 of the Start Here series, called Everything Is Fake, right?
00:10:48.559 --> 00:10:53.519
And how your company can leverage human expertise to fight AI work slop.
00:10:53.759 --> 00:10:58.960
So, consumers, I don't think most people see or understand this yet.
00:10:59.120 --> 00:11:09.519
And as the technology gets better, right, as AI video gets better, uh, as AI images get better, as AI audio in deepfakes, unfortunately, get better.
00:11:09.840 --> 00:11:14.159
That percentage of people questioning the authenticity is only going to rise, right?
00:11:14.240 --> 00:11:21.200
So synthetic media, uh, deepfake fraud, all of those things just erode the baseline trust that people rely on, right?
00:11:21.279 --> 00:11:28.879
It used to be something like, you know, oh, you go on the internet as an example, you go to Google page one, you probably have a lot of trust in that, right?
00:11:28.960 --> 00:11:37.200
If you went to page 20 and it's some random blog from some dude in the basement, you might look at that and be like, not sure about that, right?
00:11:38.639 --> 00:11:43.440
That page 20, take that times 10, right?
00:11:43.600 --> 00:11:48.240
And that's gonna be the commonplace level of distrust that we are gonna see every single day.
00:11:48.320 --> 00:11:55.200
And that's what your right, whatever sector you're in, if you haven't started to experience yet, it is gonna slap you, right?
00:11:55.279 --> 00:11:56.000
It's gonna slap you.
00:11:56.080 --> 00:11:57.759
I think it's starting with social media, right?
00:11:57.840 --> 00:12:02.399
You're seeing more and more things, and you know, things that happen in real life, people are like, oh, that's fake, right?
00:12:02.480 --> 00:12:06.159
And then things that didn't happen in real life, people are like, oh, that's real, right?
00:12:06.320 --> 00:12:08.399
You're starting to see it play out on social media.
00:12:08.559 --> 00:12:10.879
It is going to play out in our everyday lives.
00:12:11.039 --> 00:12:22.879
And I think even things that people start to see with their own eyes, they're gonna start to second guess those things as well, because it is gonna be a come a default to assume that everything you see is fake or AI generated, right?
00:12:23.120 --> 00:12:29.360
Whether it's on a screen or whether you hear uh you know something on the on the radio, uh, right?
00:12:29.679 --> 00:12:31.600
Apparently I'm 90 talking about the radio.
00:12:31.679 --> 00:12:33.519
I still love AM radio, by the way.
00:12:33.600 --> 00:12:44.240
Um, but customers are already shifting toward brands that can prove their outputs are authentic, and that's what you need to understand, and that's why responsible AI is so important.
00:12:44.720 --> 00:12:51.279
And well, aside from just consumer trust, well, you got to pay attention to the regulatory side as well.
00:12:51.600 --> 00:13:01.279
Because uh current there's been a lot of lawsuits, but the Mobly versus Workday certified that as a collective action for AI hiring discrimination.
00:13:01.440 --> 00:13:06.559
So uh a federal court just said the algorithm did it is not a defense anymore.
00:13:06.720 --> 00:13:12.159
So your company, that's why you need to start paying attention to what you produce, right?
00:13:12.320 --> 00:13:26.000
So in this case, right, there's been a lot of different lawsuits, and I think the first ones that uh, you know, kind of rose to prominence are ones that where algorithms made decisions that discriminated against people for whatever reason, uh, right?
00:13:26.159 --> 00:13:30.879
Whether it's hiring, getting a loan, uh, you know, healthcare coverage, et cetera.
00:13:30.960 --> 00:13:39.120
And that's not new, that's been around for technically decades, but it's becoming more and more prevalent as more and more companies that are not necessarily machine learning companies, right?
00:13:39.200 --> 00:13:47.759
But they're just, you know, wrapping their data lake house around uh, you know, a large language model company, and then they're putting that into production.
00:13:47.919 --> 00:13:54.399
That doing that without having the proper safeguards, that is bypassing responsible AI.
00:13:54.639 --> 00:13:58.799
Because now courts are saying that, well, no, you're on the hook for it.
00:13:58.879 --> 00:14:00.399
It's not the AI's fault.
00:14:00.480 --> 00:14:06.720
Uh, you know, so courts are rejecting any distinction between software decision makers and humans make.
00:14:06.879 --> 00:14:13.519
So, right, there's no difference between if I go and make a decision versus if the AI we're using makes a decision.
00:14:13.600 --> 00:14:21.360
And there's now state laws, as an example, in California, Illinois, New York, and Colorado that regulate how you use AI in hiring.
00:14:21.600 --> 00:14:34.799
So we'll see because we obviously know uh here in the US at least, there's uh some federal momentum to try to uh not allow individual states to mandate AI, but there's already state laws.
00:14:34.960 --> 00:14:41.759
They are in place, they are real, and we're gonna talk here in a minute about how it's actually getting bigger than that.
00:14:43.200 --> 00:14:45.679
And yeah, it can be costly.
00:14:45.919 --> 00:14:53.759
It's not just about maybe getting it wrong and maybe not winning back uh, you know, trust or winning trust uh from a potential customer.
00:14:53.919 --> 00:14:55.200
It is real money, right?
00:14:55.360 --> 00:15:00.080
Anthropic, obviously, one of the uh now one of the largest companies in the world.
00:15:00.159 --> 00:15:13.039
Uh Anthropic paid a$1.5 billion uh suit to settle a copyright case with authors, and courts treat AI outputs as relevant to whether they compete with copyrighted works.
00:15:13.200 --> 00:15:18.799
So right now, enterprise users are facing real IP exposure if their vendors trade on unlicensed data.
00:15:18.960 --> 00:15:21.679
So yeah, you need to understand, well, who are you using?
00:15:21.840 --> 00:15:26.879
OpenAI, Google Gemini, Microsoft Copilot, and are you on the hook?
00:15:27.120 --> 00:15:27.440
Right?
00:15:27.600 --> 00:15:30.399
Are you on the hook if something that you create, right?
00:15:30.480 --> 00:15:35.840
Uh let's say a uh, I don't know, a new, a new piece of software, right?
00:15:36.080 --> 00:15:44.879
And uh it's it's something related to health, and you're using uh one of these enterprise tools, but you're using AI to help you build this, right?
00:15:45.279 --> 00:15:48.240
If you get in trouble, do you have protection?
00:15:48.480 --> 00:15:49.200
Do you know?
00:15:49.440 --> 00:15:51.120
These are questions that you have to have.
00:15:51.279 --> 00:15:52.399
It starts, right?
00:15:52.559 --> 00:15:58.879
Starts in the boardroom, but then it it has to trickle down to every single major decision maker in your organization.
00:15:58.960 --> 00:16:00.879
You have to be aware of those things.
00:16:01.120 --> 00:16:05.120
And I think we may start to see some of this hit the fan in August.
00:16:05.279 --> 00:16:09.679
Now, the rollout of the EU AI Act could get pushed back again.
00:16:09.759 --> 00:16:23.200
It's already got pushed back a couple of times, but bookmark it August 2026, that is when the EU AI Act enforcement begins for high-risk AI uh in hiring credits and biometrics.
00:16:23.360 --> 00:16:26.799
So essentially, if your company and you're like, okay, why does this matter?
00:16:26.879 --> 00:16:42.240
Well, if your company does business in the EU, if they're offering business in the EU and you're using AI in that process, which is probably a good majority of our listeners, this applies to you.
00:16:42.399 --> 00:16:52.879
And non-compliance with these high-risk AI rules in the EUI and the EU AI Act carries fines of up to 35 million euros.
00:16:53.120 --> 00:16:55.840
Or here's here's a wild one, right?
00:16:55.919 --> 00:16:59.600
If you're a large company, you're like, okay, 35 million euros, right?
00:16:59.679 --> 00:17:04.799
That's our catering costs, or 7% of global revenue.
00:17:05.279 --> 00:17:07.680
Oh, now it's got your attention, right?
00:17:08.240 --> 00:17:15.759
So if you have completely bypassed responsible AI and governance, right?
00:17:15.839 --> 00:17:17.680
But that's later down the line, right?
00:17:17.759 --> 00:17:23.359
If you've bypassed responsible AI, the clock is quite literally ticking.
00:17:23.920 --> 00:17:24.240
Right?
00:17:24.400 --> 00:17:27.359
We'll see if anything like this ever catches on in the US.
00:17:27.519 --> 00:17:35.920
I think maybe at the state level it will, but then, like I've talked about on the show, if a federal judge throws it out, well, that will be the deciding factor, right?
00:17:36.000 --> 00:17:39.039
It's probably gonna get uh litigated at some point.
00:17:39.359 --> 00:17:47.039
But if you do business in Europe, which is many, many companies, and you're using AI, you could be on the hook.
00:17:47.119 --> 00:17:53.119
If you go, if you are found non-compliant with any of those high-risk AI rules.
00:17:54.000 --> 00:18:02.559
It's not just about enforcement, it's not just about doing uh, well, we better be responsible uh with AI so we don't get in trouble.
00:18:02.799 --> 00:18:11.039
No, being responsible, yes, it creates trust, but responsible AI leads to higher ROI on AI.
00:18:11.200 --> 00:18:21.440
So, according to McKinsey, companies that are investing heavily in responsible AI report over a 5% profit impact versus those companies that aren't.
00:18:21.599 --> 00:18:27.680
And senior leadership involvement in AI governance drives significantly greater business value.
00:18:27.839 --> 00:18:41.119
So if your business leaders are heavily involved in the process and they understand the basics of responsible AI and they're making those decisions from top to bottom, it's not just about not getting in trouble, right?
00:18:41.599 --> 00:18:49.920
This is foundational for growing your organization because it unlocks real scale and real returns.
00:18:50.640 --> 00:18:53.440
So let's start to wrap with this.
00:18:53.599 --> 00:18:58.720
We started with the five pillars that kind of helped define what responsible AI is.
00:18:58.880 --> 00:19:07.519
Now, here is the playbook, the five moves to ensure your responsible AI strategy survives.
00:19:08.319 --> 00:19:14.000
And again, make sure you go uh check out the governance episode because a lot of these things are aligned.
00:19:14.079 --> 00:19:14.799
But here we go.
00:19:14.960 --> 00:19:17.119
Uh, rule number one, know what you have.
00:19:17.200 --> 00:19:21.519
You need to audit every AI system and classify each one by risk level.
00:19:21.680 --> 00:19:25.519
The cool thing is the EU AI Act actually put those risk levels out.
00:19:25.680 --> 00:19:28.079
Number two, assign real ownership.
00:19:28.160 --> 00:19:32.079
You need to be able to name the individual with the authority, budget, and accountability.
00:19:32.160 --> 00:19:36.000
That 10-second rule when someone goes wrong, who is the human responsible?
00:19:36.160 --> 00:19:39.039
You have to know with the utmost certainty.
00:19:39.279 --> 00:19:43.440
Number three, you need to audit for bias before the courts do.
00:19:43.599 --> 00:19:57.680
You need to test your AI tools against your actual data and to see, well, is there any leaks that could cause legal troubles, whether it's against the uh the EU high risk or uh any of the state laws here in the US?
00:19:57.839 --> 00:20:01.440
Number four, you need to build expert-driven oversight.
00:20:01.519 --> 00:20:04.319
If you listen to the show, you know I hate human in the loop.
00:20:04.480 --> 00:20:15.039
You need to have expert-driven oversight, you need to have proactively your domain experts driving your AI, uh responsible AI policy.
00:20:15.200 --> 00:20:23.039
So that is domain professionals reviewing outputs, not just well, bill an IT rubber stamping your new agentic workforce.
00:20:23.200 --> 00:20:28.319
And then number five, you need to trade treat transparency as a competitive advantage.
00:20:28.480 --> 00:20:33.359
You should be disclosing when and how AI is involved, uh, right?
00:20:33.519 --> 00:20:41.759
Not just to your consumers, but to your employees, to your stakeholders, both internal and external, right?
00:20:42.000 --> 00:20:50.640
When we saw that uh consumers are starting to make decisions on those companies that are transparent in AI, right?
00:20:50.880 --> 00:20:52.240
Like you kind of think back.
00:20:52.319 --> 00:20:54.480
I think that there's been this kind of boom.
00:20:54.640 --> 00:20:56.880
Maybe it's a generational thing, I don't know.
00:20:56.960 --> 00:20:57.519
Uh, right.
00:20:57.599 --> 00:21:02.160
But when you think of like organic and you know, food with clean ingredients, right?
00:21:02.319 --> 00:21:13.200
You can see over the last 10 years how these types of brands have exploded because they're being transparent about what you are consuming and they're saying what's real, what's fake.
00:21:13.519 --> 00:21:15.200
Think of the same thing, right?
00:21:15.440 --> 00:21:20.000
With what you are putting out there to the world, with what your company is putting out there to the world.
00:21:20.160 --> 00:21:31.920
Trust me, the trust, the trust crisis has barely started, and it is a much more significant issue than much than most people realize.
00:21:32.079 --> 00:21:41.119
And it's something that if you are like, uh, well, we'll we'll worry about this when we know it's a problem, when we hear rumblings of it, it's gonna be too late at that point, right?
00:21:41.200 --> 00:21:43.440
Because your competitors are already gonna be doing this.
00:21:43.519 --> 00:21:47.039
So I think the fifth one here uh is huge.
00:21:47.200 --> 00:21:52.319
You have to become overly transparent now as a competitive advantage.
00:21:52.480 --> 00:22:04.880
And again, it's not just uh how you communicate these things externally or to you know external stakeholders, to uh current clients, potential customers, etc.
00:22:05.200 --> 00:22:09.279
It's also how you communicate these things in your own organization, right?
00:22:09.839 --> 00:22:12.400
How do you deploy AI agents?
00:22:13.440 --> 00:22:16.000
Do all of your staff know?
00:22:16.160 --> 00:22:17.279
Do all of your leaders know?
00:22:17.440 --> 00:22:22.000
Do frontline workers know how these decisions are ultimately being made, right?
00:22:22.079 --> 00:22:24.720
When they go click that, you know, agent run?
00:22:25.279 --> 00:22:26.000
Probably not.
00:22:26.559 --> 00:22:34.319
Transparency starts at home and then it is going to expand uh right to your neighborhood or your potential clients.
00:22:35.279 --> 00:22:38.640
So as we wrap here, here's what I want to leave you with.
00:22:39.759 --> 00:22:42.000
The same thing with governance, right?
00:22:42.079 --> 00:22:46.160
The governance are the rules in which you carry this out, but responsible AI, right?
00:22:46.240 --> 00:22:48.799
When you think about ethics, what's right, what's wrong.
00:22:48.960 --> 00:22:55.279
And then you think about how we are actually uh, you know, the foundational rules that help set this.
00:22:56.240 --> 00:23:03.440
Remember that responsible AI turns your company's AI values into real decisions.
00:23:03.599 --> 00:23:07.279
It is the foundation on how you look at use cases, data, and oversight.
00:23:07.359 --> 00:23:11.599
And it also sets the pace for how you will ultimately govern AI.
00:23:11.839 --> 00:23:16.079
But responsible AI helps you technically go faster.
00:23:16.240 --> 00:23:20.960
In the same way people think governance slows your AI efforts down.
00:23:21.200 --> 00:23:26.720
Well, responsible AI doesn't slow your efforts down either because it builds trust.
00:23:26.880 --> 00:23:32.240
And trust uh builds the roads, it builds the highway for speed, right?
00:23:32.400 --> 00:23:39.839
Without those lanes uh of responsible AI, you're just gonna have your cars going all over the place, right?
00:23:40.000 --> 00:23:42.079
So trust is that infrastructure.
00:23:42.160 --> 00:23:48.160
The ethics provides the values, responsible AIs provide the decisions, and governance enforces both.
00:23:48.240 --> 00:23:57.519
And the companies that are earning trust now, while distrust is on the rise, will own the competitive advantage in 2027.
00:23:57.839 --> 00:23:59.279
All right, that's a wrap, y'all.
00:23:59.359 --> 00:24:03.519
Thanks for sticking with me as I can barely uh you know talk.
00:24:03.599 --> 00:24:05.200
It feels like my nose is plugged.
00:24:05.279 --> 00:24:08.720
Hopefully, at one point here, I'm not gonna be sick or have crazy allergies.
00:24:08.799 --> 00:24:09.839
So you can all actually hear me.
00:24:09.920 --> 00:24:10.880
But thank you for sticking around.
00:24:10.960 --> 00:24:16.160
And I hope this was helpful as we tackled the responsible AI playbook.
00:24:16.240 --> 00:24:20.480
Uh, and if this was helpful, make sure you go to starthearseries.com.
00:24:20.559 --> 00:24:25.039
That is gonna give you free access to our exclusive inner circle community.
00:24:25.119 --> 00:24:26.880
You're not gonna find it anywhere else.
00:24:26.960 --> 00:24:33.599
Uh, and if you haven't already, please do me a favor go subscribe to the podcast on Spotify or Apple.
00:24:33.759 --> 00:24:34.880
Leave us a rating if you could.
00:24:34.960 --> 00:24:35.839
I'd appreciate that.
00:24:36.000 --> 00:24:36.960
Thank you for tuning in.
00:24:37.039 --> 00:24:39.599
Hope to see you back tomorrow and every day for more.
00:24:39.680 --> 00:24:40.720
Every day, a lot.
00:24:40.799 --> 00:24:41.839
Thanks, y'all.
