Proserv & PE Insights · Monday, September 14, 2026

Hi, this is Koko from Koko Knows.

Let's get right into what's new, because there's a real signal this time, not just noise. EY has disclosed a one hundred million dollar bonus pool tied explicitly to human skills. Not billings, not utilization, human skills. That's a firm putting real money behind the idea that judgment, relationship-building, and client trust are becoming the scarce, billable asset precisely because AI is eating the routine execution work. If you're running a professional services practice and you haven't thought about how you comp for the things AI can't do, EY just forced the question onto your agenda. Expect competitors to either match this or have to explain publicly why they're not.

Second big one: Cognizant stood up an AI innovation center for Citizens Bank, went from build to production in under six months, hit zero attrition along the way, and it's targeting one hundred million dollars in annual run-rate savings. I want to underline the zero attrition part, because that's usually where these transformation stories fall apart. Clients are terrified of AI-first delivery models that either don't work or cause a talent exodus that tanks service quality mid-transition. Cognizant appears to have cracked a delivery model that gets the cost savings without the disruption. If you're pitching AI-enabled delivery to clients or evaluating vendors for your own portfolio companies, that's the proof point to ask about by name.

And then the number that should be sobering anyone with a PE value-creation plan on their desk right now: Bain's research says only about twenty percent of portfolio companies have generative AI actually in production with measurable results. Twenty percent. That means four out of five companies where somebody built an investment thesis assuming AI-driven EBITDA improvement are running ahead of operating reality. If your fund has been modeling AI upside into hold-period plans, this is the moment to go back and pressure-test which of those assumptions are real deployments versus pilots that never left the lab.

So here's how I'd tie these together for you. EY and Cognizant are showing you both ends of the same trade. Firms are monetizing the human side of the transition by paying premiums for differentiated judgment, and simultaneously banking hard savings from automated delivery. That's a sophisticated two-sided play. But Bain's twenty percent figure is the reality check underneath it all: most organizations, especially portfolio companies, are nowhere near that level of sophistication yet. The gap between what's possible and what's actually deployed is wide, and it's exactly where diligence teams and operating partners need to be spending their time right now, not on the AI narrative in the pitch deck, but on what's actually running in production.

There's a governance thread running underneath all of this too, and it's worth naming directly. We're seeing data point after data point suggesting that a huge share of AI spend, some surveys put it as high as eighty percent of token spend, can't be traced to any actual business outcome. Pair that with reports that roughly ninety percent of agent pilots never make it to production, and you've got a pattern: boards and investment committees are approving AI budgets without demanding the same outcome accountability they'd demand of any other capital allocation. If you sit on a board or you're running diligence on a target, put ROI attribution on the same agenda as the spend approval itself. Don't let adoption metrics substitute for unit economics. And if you're advising clients on this, know that procurement and finance buyers are about to start asking for exactly this kind of traceability in every renewal conversation. The vendors who've built cost-to-outcome measurement into the product, not bolted on afterward, are going to win those conversations by default.

One more thread worth flagging on the talent side. Accenture is building a multi-year cybersecurity apprenticeship pipeline with PeopleShores, sourcing entry-level talent years ahead of projected 2031 demand. That's a quiet but important bet: rather than fight everyone else for scarce senior security hires, build your own pipeline now and own the cost and retention curve. Worth considering for any practice or portfolio company staring down a talent gap that's only going to get tighter.

So three things to carry into your next partner meeting or portfolio review. First, human-skills compensation is now a live competitive signal, not a nice-to-have. Second, delivery models that cut cost without attrition are the new client-pitch differentiator, and Cognizant just set the bar. And third, before anyone signs off on another AI-driven value-creation thesis, ask for the production evidence, not the roadmap.

That's the rundown for today. I'm Koko, and this has been Koko Knows. Talk soon.