Proserv & PE Insights · Monday, October 5, 2026

Hi, this is Koko from Koko Knows.

Let me get right into it, because the number that matters most today comes from Bain, and it's a big one. Bain is now saying the AI industry needs six trillion dollars in annual revenue by 2031 just to justify the infrastructure spend that's being committed right now. Projected spend could hit one point five trillion a year by then, and that dwarfs what we saw built out for mobile and cloud combined. This isn't a theoretical warning anymore, it's quantified, and it lands the same week McKinsey dropped a stat that should stop every partner pitching AI transformation work in their tracks: only thirty-seven percent of companies are seeing positive EBIT from their AI programs. Ninety-three percent are blowing through their AI budgets. That gap, between what's being spent and what's being proven, just got a lot harder to paper over in a client room.

So that's the first thing that's changed. The ROI conversation has shifted from "trust the roadmap" to "show me the P&L impact," and I think boards are going to start demanding that line item by item, before the next budget cycle, not after.

Second development, and this one's more operational than strategic, but it matters for anyone running delivery: Bain's separate technology report found AI coding tools lift task completion by twenty-one percent, which sounds great, except code review time balloons by ninety-one percent. That's not a clean productivity win, that's a cost shift. The work didn't disappear, it moved downstream into review and oversight, and if your delivery models and your pricing aren't accounting for that bottleneck yet, you're going to eat that cost on the next engagement instead of pricing for it upfront.

Third, and this is the pricing story underneath all of it: EY is formalizing outcomes-based pricing, joining a growing list of peers moving away from hourly, junior-staff economics. That's not a small shift. It confirms the billable hour model is eroding across the industry, and it raises the bar, because outcomes-based pricing only works if you can actually prove the outcome. Which loops right back to that thirty-seven percent EBIT number. Firms are being asked to get paid on results at the exact moment the data says most AI results aren't showing up yet.

Now, against all of that caution, Accenture is still writing checks. Four point nine four billion dollars deployed across seventeen acquisitions in fiscal twenty twenty-six, with another five billion earmarked for fiscal twenty twenty-seven. So the capital isn't retreating, it's consolidating around AI-adjacent capability even as the ROI evidence gets harder. That tells me the smart money isn't betting against AI, it's betting on being the firm that can actually prove the payback when the client asks.

So here's what I'd take into your next partner or portfolio conversation. First, if you're pitching AI transformation work, build the business case on realized EBIT impact, not vendor pricing trends or model releases. When a new model drops with a price cut, that's not a reason to expand your AI footprint, it's a trigger to re-audit your run-rate AI spend against what you can actually measure in return. Second, price in the review and oversight burden on coding tools and agentic workflows right now, don't wait to discover it mid-engagement. That ninety-one percent review time increase is a delivery risk hiding inside what looks like a productivity story, and if you're underwriting fixed-fee work on the old assumptions, you're exposed.

Third, and this is worth watching closely, the big firms are starting to standardize around a shared delivery model. Accenture, Deloitte, Capgemini, McKinsey and Bain all placed engineers into Anthropic's first Claude Frontier Academy cohort. That points to a common "forward deployed engineer" role becoming the default across the top of the market, which is a real threat to any services go-to-market that hasn't adapted its talent model yet. If your firm's differentiation has been "we have the people who can do this," that moat is narrowing fast.

For PE portfolios specifically, the implication is straightforward. Any portfolio company that's been sold an AI transformation roadmap needs that roadmap stress-tested against Bain's six trillion dollar bar and McKinsey's thirty-seven percent reality check before the next capital call. Value creation plans built on assumed AI productivity gains need a harder look at the delivery and review costs actually showing up underneath them.

That's the state of play. The capital's still moving, the pricing models are still shifting, but the proof bar just got a lot higher, and the firms that can show real numbers, not roadmaps, are going to be the ones winning the next round of work.

That's it for this update. I'm Koko, and this has been Koko Knows.