Operational alpha is back in PE, and most value creation plans are already out of date

Every serious voice in private equity is converging on the same diagnosis. 

Apollo, in January, argued the asset class needs to return to its roots: buy well, improve operations, exit with discipline, because cheap debt and multiple expansion are gone. Bain’s 2026 Global PE Report goes harder, with the line that has been quoted everywhere since February: ‘12 is the new 5.’ The deals that used to clear their return hurdle on roughly 5% annual EBITDA growth now need 10 to 12. Bain’s advice to GPs is blunt: build durable operating capability rather than marketing it, invest in talent and AI, and move from diagnosing an asset’s full potential to executing from the first day of ownership. Blackstone’s leadership calls AI ‘the megatrend of megatrends’ and has built a 300-strong analytics community across its portfolio. KKR has AI embedded across 225-plus portfolio companies through its Capstone team. 

The diagnosis is unanimous. Financial engineering is no longer enough. Operational value creation is back, and AI is meant to be in the plan. 

So why are most of the value creation plans still answering half the question? 

We’ve spent the last year in back-to-back conversations with PE firms and portfolio companies. The same four questions keep appearing, with either a clear solution emerging or managers getting caught flat-footed. The four questions below reflect what we’ve learned from seeing organisations tackle these challenges effectively, and where others have struggled. 

The variable the industry keeps underpricing

Bain and StepStone’s GP Outlook, published this February, contains one of the strangest data points of the year. GPs report their best returns from generative AI in deal sourcing and due diligence. Inside their own portfolios, the benefits skew to cost savings, and nearly 40% of GPs expect no material financial impact from AI in 2026.. 

Read that twice. While Blackstone is putting data scientists inside its companies and KKR is doing the same at scale, four in ten GPs are treating AI in their existing book as a cost story or a non-event. 

That is a mistake, and there is now a cautionary tale big enough that no GP should miss it. 

What Chegg should teach every GP

If you haven’t come across it, Chegg is a US-listed education technology company. Its core product was online homework help: students paid a monthly subscription for access to a huge library of textbook solutions and step-by-step answers to academic questions. It was exactly the kind of business private equity has spent two decades loving. A subscription model with predictable revenue and a ten-year content library that looked like a real moat. In 2021 it was worth around $14.5bn. 

Then ChatGPT arrived and Google bolted AI overviews onto search. The product became substitutable overnight, and the channel that delivered its customers went dark at the same time. By late 2025 Chegg’s revenue had roughly halved year on year, it had cut close to 45% of its workforce, and a company once worth $14.5bn was worth nearer $100m. A ten-year moat, gone in barely two years. 

Here is the part to sit with. Chegg saw it coming and tried to respond. In 2023, it launched an AI study tool built in partnership with OpenAI itself. It failed to hold subscribers, because students who could already use ChatGPT directly had no reason to pay for a wrapper around the same technology. Bolting a generic AI feature onto a business under threat did not save it. Remember that the next time someone walks into a board meeting and presents an ‘AI feature’ as a moat. 

Chegg is extreme, but it is not exotic. The categories private equity has prized for their defensibility are now often the most exposed: vertical SaaS that codified a workflow, BPO and outsourced services, content and publishing, legal, marketing and customer-support operations. The sticky, recurring-revenue businesses LPs underwrote on. 

This is not just my read. MIT’s ‘State of AI in Business 2025’, the study behind the widely quoted finding that 95% of corporate AI pilots produce no measurable P&L impact, also found that the use cases delivering the strongest returns were AI replacing business process outsourcing, cutting agency costs and automating back-office work. Put plainly: the clearest evidence we have of AI actually working in the enterprise is AI eating exactly the kind of labour-heavy service businesses that sit in a lot of portfolios. 

If your value creation plan assumed a four-year window of operational improvement on a business AI compresses in eighteen months, you do not have a plan. You have a problem. 

So, the new playbook has a second leg. Operational value creation in this cycle has to be AI-literate by default. The firms doing it well are not running a thematic AI bet on the side. They are treating AI as operational hygiene across every company they own. 

The four questions I keep being asked

These are the conversations I’m having every week. Operating partners, deal teams, portfolio company CEOs and the occasional fund principal who has done the maths and realised the plan on the shelf is not going to deliver. 

1. Which of our investments could AI take to zero?

Always the first question, and the one most firms answer too slowly. It is not a portfolio review. It is triage. And triage means separating noise from terminal risk. Almost every company is under threat from AI in the loose sense that its competitors now have new tools. That is not the question worth board time. The real one is which of these businesses AI could take to zero, the way it nearly did to Chegg. You are looking for three things: 

• Where does AI compress the unit economics? Usually labour-heavy operations, content production and decision-support. 

• Where does AI erode the moat? Network effects tend to survive. Codified expertise does not. 

• Where does AI open a new entrant route into your company’s market? Watch the seed-stage deck flow in your sectors. 

The MIT number cuts both ways here. Yes, 95% of pilots fail, but that is an argument for ruthless prioritisation, not for sitting still. The exposure map tells you where to spend. Without it you fund a hundred pilots and get five outcomes. Even the most AI-forward buyers are not immune. Uber burned through its entire 2026 AI budget in roughly four months after gamifying adoption internally and its own COO has admitted he cannot yet draw a clear line from that spend to features customers actually value. Money spent on AI is not the same as value created by it. 

2. How do we build a defensible moat, fast?

Once you have found the exposure you need an answer, and the honest version is that not every moat can be rebuilt and not every business is worth defending. Where the asset is worth saving, the logic is now fairly settled. AI raises the value of whatever it cannot replicate and destroys the value of anything that was a wrapper around expensive thinking. So: 

• Proprietary data becomes the moat. If you hold data your competitors cannot get, lean into it hard. 

• Distribution and trust outlast features. AI commoditises the build. It rarely commoditises the relationship. 

• Vertical depth beats horizontal breadth. Generic AI-enabled entrants win in flat markets and struggle where regulation, workflow and domain knowledge compound. 

A point the MIT data makes well: how you deploy matters as much as what you deploy. Buying from specialist vendors and building partnerships succeeded around two-thirds of the time in their dataset. Internal builds succeeded a third as often. The instinct to build your own model in-house is usually the expensive route to the same place. 

The mistake I see most often is still the Chegg mistake, scaled down. A generic AI feature, bolted on, presented as differentiation. It is table stakes, and your competitor is shipping the same thing on the same model. 

3. How do we use AI before and after the deal?

Pre-deal and post-deal are different problems, and running them together is where firms lose time. 

Before the deal, AI is changing the speed and depth of diligence, and this is exactly where the Bain and StepStone survey says GPs are seeing their strongest returns. Market-signal scraping, IP and data-asset valuation, customer-review mining at scale, target identification beyond the usual databases. Teams that have rebuilt diligence with AI in the loop are surfacing things their rivals will not see for another two months. 

There is also a pricing problem, and it is getting harder, not easier. How do you fairly value a target with real AI-driven growth potential when there is no settled set of comparables to price it against? It is the SpaceX question at portfolio-company scale: huge optionality, a very wide range of outcomes, and a real risk of paying up for a story. The mirror image matters just as much. Be ruthless about the difference between a business with a defensible AI advantage and an ‘AI-enabled’ business that is really just using an off-the-shelf copilot to write emails. The label is cheap. The moat is not. 

After the deal, AI is an operational lever, not a strategy, and this is where the industry is underperforming. The first 100 days should produce a prioritised view of where AI takes cost out, where it lifts revenue and where it protects the moat. Real numbers, real owners, real dates. Not a programme but a plan. 

4. How do we educate our own teams without creating a distraction?

The most underrated question on the list. Apollo does not touch it. Bain mentions it once. 

The risk is not that GPs and operating partners do not know about AI. It is that they know just enough to chase the wrong things. I’ve watched senior teams spend a partner meeting debating model selection while their company’s largest competitor quietly retrained its sales engine. 

The fix is unglamorous: structured education for the people doing the work, not theatre for the people running the firm. Enough working knowledge to ask sharp questions about where AI bites, what it costs and what good looks like. Not so much that everyone turns into an amateur machine-learning engineer. 

What the firms getting this right are doing

They have stopped treating AI as a thematic investment question and started treating it as an operational question on every company they own. They build an AI-exposure diagnostic into the 100-day plan. They refresh value creation plans on shorter cycles, because the half-life of a plan has shortened. They hold operating partners accountable for AI exposure the way they hold them accountable for working capital. And they educate their teams to create focus, not noise. 

They are doing it now, because the gap between the firms that work this out in 2026 and the ones that work it out in 2028 will show up in DPI for a decade. 

None of this is actually new

For all the noise, AI does not change the job. Private equity has always created value the same way: take a business, improve it through people, process and technology, and build something worth more than you paid. That is the discipline. AI is the newest, and by some distance the most powerful, lever in the technology column, but it sits inside the model the industry has always run. 

That distinction is the whole game. It is the difference between treating AI as a strategy and treating it as execution. AI may be the most revolutionary technology any of us work with in our careers, but revolutionary technology creates nothing on its own. The value is not in the idea, the thesis or the board deck. In this industry it never has been. It is in execution: the unglamorous work of rewiring how a company operates, function by function, until the numbers move. 

So, the firms that excel in the next decade will not be the ones with the best AI story. They will be the ones who treat AI the way good operators have always treated every lever worth having. Something you implement, measure and hold people accountable for. Buy well, improve relentlessly, return capital. 

Where this leaves us

The industry has the diagnosis right. Apollo, Bain, Blackstone, KKR. Back to fundamentals. Operational alpha. 12 is the new 5. 

What has changed is the ground beneath the assets. The business you are buying, the operations you are improving and the moat you are defending all sit on top of the largest productivity shift in a generation, and the firms writing about it in their annual reports are not always the ones acting on it inside their portfolios. Chegg is what the downside looks like when a good business meets that shift without a plan. 

The fix is the one this industry already knows. People, process, technology, executed hard. AI just raises the stakes of getting it right. 

If you are a GP or a portfolio company CEO and you are not certain your value creation plan accounts for this, that is the place to start, and it is worth doing before the next exit window opens. 

If you are facing the same challenges, we would love to talk.