WEBVTT

NOTE
{
    "characteristics": "public.accessibility.transcribes-spoken-dialog",
    "language": "en-US",
    "language-name": "English (US)"
}

00:00:07.120 --> 00:00:12.560
Let's talk about decision-making, especially as it relates to AI. AI has the potential to transform

00:00:12.560 --> 00:00:17.999
human decision-making. In our 2026 Global Human Capital Trends survey, as many as 60% of

00:00:18.000 --> 00:00:22.959
executives said they regularly now use AI to support their decisions. Gartner projects, by just

00:00:22.959 --> 00:00:29.240
next year, 2027, half of all business decisions will be augmented or automated in some way using

00:00:29.280 --> 00:00:35.560
AI. We're still getting the balance of AI autonomy and human agency right. At the heart of this issue,

00:00:35.600 --> 00:00:40.120
what do you think we're really grappling with when it comes to AI and human decision-making? I

00:00:40.120 --> 00:00:46.440
think it's about legibility to ourselves and each other. When you as a human make a choice, even a

00:00:46.440 --> 00:00:51.520
bad one, you’ve got some sense of “why,” right? There’s a story you can tell yourself: “I picked

00:00:51.520 --> 00:00:57.600
this because it felt right, or I trusted my instinct, my guts.” That's how you make sense of who

00:00:57.640 --> 00:01:03.889
you are. With AI in the mix, that starts breaking down very quickly. Not just because AI is a black

00:01:03.889 --> 00:01:09.610
box, though it can be, but because the decision doesn't always pass through you anymore. The

00:01:09.610 --> 00:01:15.169
intermediate steps don't go through you. You can't reconstruct the reasoning. You can't really

00:01:15.169 --> 00:01:21.529
challenge it. You're just accepting an output. And here's what bugs me the most. We might lose the

00:01:21.529 --> 00:01:28.529
ability to account for ourselves, to look someone in the eye and say, “Here’s why I did that” in a

00:01:28.529 --> 00:01:34.849
way that connects to our core values. Responsibility needs that, so does trust. Plus, and

00:01:34.849 --> 00:01:41.010
this sounds small, but it's really not, if we stop exercising our judgment, what happens to it? Does

00:01:41.010 --> 00:01:47.609
it atrophy? Do we forget how to navigate the messy, ambiguous stuff that actually matters? What do you

00:01:47.609 --> 00:01:51.769
think are the tipping points that will force leaders to address this human-in-the-loop

00:01:51.769 --> 00:01:58.609
question head on? The tipping points are probably complexity, speed, volume, and really the first

00:01:58.609 --> 00:02:04.710
major disaster with no clear owner. Autonomous car kills someone. Hiring algorithm discriminates at

00:02:04.710 --> 00:02:11.630
scale. And when everyone asks who's responsible? The answer is a shrug. That will force change

00:02:11.630 --> 00:02:18.550
fast. That will be one of the major tipping points. And maybe it is about human setting

00:02:18.550 --> 00:02:24.910
boundaries up front and AI operating within them. Maybe it is more retrospective audits of AI

00:02:24.910 --> 00:02:30.789
outcomes, instead of real-time approval. Maybe it is about humans handling exceptions only, the

00:02:30.789 --> 00:02:37.069
weird edge cases. So really moving, you know, leaders moving from the role of not just decision-maker

00:02:37.070 --> 00:02:43.470
but system steward, someone who understands how the AI behaves and when it drifts and when to

00:02:43.509 --> 00:02:45.470
jump in and take control.