06 Enterprise AI 2.0: Past Hype, Bring the Receipts

Ideas got AI in the door. Results are what will keep it there.

That’s exactly where the 2025 Wharton + GBK AI Adoption Report says we are right now. Gen AI has fast-tracked into the enterprise, it’s being used weekly and even daily, budgets are going up—and now leaders want accountable acceleration. Not more pilots. Not more proofs of concept. Proof of value.

Here’s the part I really like: the report basically confirms what we see in real client work—the constraint has shifted from tools to people. As AI becomes “everyday work,” the gaps aren’t in the model anymore. They’re in skills, uneven training, culture, and change management. Human capital is now “the decisive lever that converts usage into scalable ROI.” That’s their language, and I’m here for it. Here are some interesting highlights from the report.

✅1. AI is mainstream. Accountability is new.

82% using Gen AI regularly means “we’re experimenting” isn’t a good story anymore. If your team is using AI, leaders will (and should) ask: what did it improve, for whom, and how do we know? 72% are now measuring ROI, and most are seeing it—that’s a big maturity jump.

✅2. People and policies are now the friction.

This is the flavor we can’t ignore: employee resistance and lack of trust are still in the top 10 barriers—especially for laggards. So it’s not “AI doesn’t work”—it’s “we haven’t brought people along,” and sometimes, “we over-locked it down.”

At the same time, orgs have tightened usage policies: data security, usage restrictions, explicit approvals, councils and boards, human oversight—all up. That’s good governance, but it also means some teams are trying to adopt inside a maze of rules. The report literally calls out that IT and finance boosted rigor the most. So yes—people and policy are holding back adoption as much as tech.

✅3. Training isn’t keeping pace

Another callout I love: lack of training entered the top 10 barriers in 2025. That tells you leaders know AI is here to stay but haven’t resourced capability-building at the same speed as tool rollouts. You can’t skip the “teach people to think with AI” step and expect enterprise-grade outcomes.

This is where my interrogator brain comes back in—AI still needs someone to ask a good question. Someone to validate. Someone to connect it back to the business goal. Tools don’t replace judgment. They expose where judgment is missing.

💡 Clear Take

Gen AI is in. Now it has to earn its seat.

• We’re not waiting on models anymore. We’re waiting on people, policy, and culture to catch up.

• Measure it or lose it.

• Train for it or stall it.

• Govern it—but don’t strangle it.