Proserv & PE Insights · Wednesday, September 23, 2026

Hi, this is Koko from Koko Knows.

Let's get right into what's new. Accenture and Anthropic are putting real money behind a bet that AI assurance, not just AI deployment, is the next big fee line. We're talking two billion dollars combined, committed over five years, funneled through Accenture's Faculty unit into an independent venture focused on evaluating models, red-teaming them, and testing safeguards before they go live in client environments. Think about what that signals. Accenture isn't just selling you the AI, it's building the business of certifying whether the AI can be trusted. That's a categorically different revenue stream than staff-augmentation consulting, and it's one that plays directly into where boards are already nervous, on governance, on liability, on agentic AI's expanding attack surface. If you're running a professional-services practice and you don't have an answer yet to "how do we independently validate this model before we put it in front of a client," Accenture just told you what the answer looks like, and they're two years ahead on building it.

The second thing that matters this cycle is PwC. Paul Griggs is now publicly putting his name behind a "two in the box" delivery model, pairing an engineer with a business architect on client engagements. This isn't a pilot program buried in an innovation lab, this is leadership saying out loud that the firm's operating model is changing shape around AI leverage. And that matters for margin math in a very specific way. For decades, professional-services economics ran on pyramid staffing, lots of junior bodies billing hours under a thin layer of partners. If the delivery unit becomes a paired role instead of a headcount ladder, that changes your staffing ratios, your realization rates, and frankly your whole hiring pipeline for the next several years. Watch whether other firms follow this framing publicly, because once one major firm names its model out loud, the rest usually feel pressure to articulate their own answer rather than let clients assume they're behind.

Now, two threads from recent days are still hanging open and I don't think you should let them drop off your radar just because they're not brand new. PwC Switzerland cutting its bonus pool, reportedly by half, communicated verbally with no written record, is the clearest signal yet that AI-driven cost pressure is reaching all the way into comp decisions, not just headcount planning. If that's happening at PwC Switzerland, ask yourself honestly whether your own partner distributions are insulated from the same pressure, because the logic that's compressing bonuses there is the same logic compressing fees everywhere: clients are starting to expect AI-assisted delivery to cost less, and somebody has to absorb that gap. And KPMG's headless enterprise architecture push with OpenAI is still the more interesting long-game story here. KPMG is trying to own the AI-native delivery infrastructure layer before clients go build or buy it themselves. That's a land-grab on where future engagement value gets captured, not just a partnership announcement.

I'd also flag Deloitte's shadow-AI numbers again, because one in three unauthorized GenAI users inside client organizations is a governance exposure that nobody has closed out yet. If you're advising portfolio companies or running due diligence on target firms, that stat alone should be on your checklist. Untracked AI usage is exactly the kind of thing that turns into a nasty surprise in a data room six months from now.

So here's how I'd frame the implication set for you. On growth, the Accenture-Anthropic move opens a new sellable category, AI assurance and evaluation, and if you're not building or partnering into that capability, you're ceding it. On margin, PwC's structural shift shows the industry moving leverage away from pure headcount toward paired-role delivery, and that's coming for your staffing model whether you initiate it or not. On delivery risk, shadow AI and unaudited agentic workflows are quietly becoming the diligence item that catches people off guard, so get ahead of it before a buyer or a regulator finds it for you. And on comp, PwC Switzerland is the early warning that AI cost pressure doesn't stop at the P&L, it reaches partner pockets. The firms treating AI as an operating model question, not a tooling question, are the ones setting the pace right now.

That's the rundown for this edition. I'm Koko, and this has been Koko Knows. Thanks for listening, and I'll catch you next time with what's moved.