Start a New Value Creation Plan

Bain tells private equity to start with the value creation plan. The instruction is sound. The plan is the problem.

What Bain found

On 16 September, Bain & Company published "Getting Past the AI Value Paradox in Private Equity." The diagnosis holds up. In Bain's 2026 CEO survey, 82% of chief executives say their AI transformation is at best partially delivering what they intended. The sample is 100 CEOs, so read the figure as direction, not precision.

Bain names five causes: use-case swirl, the micro-productivity trap, tools-first bias, capability gaps and sponsorship gaps. Its prescription is to start with the value creation plan, isolate three to five opportunities inside it, and redesign those workflows before choosing a tool.

The worked example is a PE-owned company that manages large hospitality and leisure properties. Management had 30 candidate processes and narrowed them to four tied to revenue: finance automation, lead generation, proposal automation and client value reporting. The company is tracking towards a 40 to 50 percent EBITDA increase by 2030.

That is a disciplined story. It is also one unnamed company, with no baseline disclosed and a forecast in place of a result.

Where Kevin Mannion pushed back

Kevin Mannion, founder and CEO of Workstuff AI, read the same article and did not accept the conclusion. His argument is that filtering use cases for value cannot be the whole answer, or every portfolio company would be up 50 percent on EBITDA.

His diagnosis goes further. A typical value creation plan holds two kinds of lever: sell more of the existing product, and take cost out of existing processes. AI is strongest somewhere else. It brings new products and offerings to market, and it makes operating-model redesigns feasible that were once judged too hard to count.

So his answer is not to start with the value creation plan. It is to start a new one.

I agree, and this essay exists because he made the case first.

Why the plan caps the answer

A value creation plan is written before close, from a data room. It models the business as the seller presented it, and it prices what the deal team could see.

Two things follow. The levers are the ones a spreadsheet can price. And the baseline for anything harder is invisible on day one. That second point is my inference, not Bain's, and it comes from operating experience.

At a global pharma client, content was produced by more than 1,000 agencies across 12 major markets, with a long tail across more than 40 others. Content reuse stood at 6%. About $100 million of content creation spend could not be attributed to anything.

No diligence process would have priced a lever there. Nobody could see the thing the lever would move.

The work that followed put a single platform in place, redesigned the workflow, automated more of it and consolidated production into five global entities by therapeutic area. Content reuse rose from 6% to 70%. The client saved more than $11 million and avoided more than $15 million in further cost.

That result was not a use case. It was an operating model. A filter that asked which of 30 AI ideas maps to a line in the plan would never have found it, because the line did not exist.

The part that is organisational

I have run the other version of this problem. At eg+ in London, nine agency production units were merged into one business over twelve months, roughly 200 people and revenue in the region of €40 million. No client left during the consolidation, and revenue rose 25% in the first year. I owned the client side: relationships, service levels and the sales proposition.

It was a carve-out and merge of internal units, not a buy-and-build, and it did not go cleanly. We lost people to relocation, and the systems consolidation struggled at first.

The lesson transfers. The tools were never the hard part. The hard part was deciding who owned what once the units merged. AI redesign raises the same question at higher speed, and private equity operating partners already know it from every integration they have run.

Where Bain and Kevin overlap

The gap between them is narrower than Kevin's post suggests. Bain's second example, a parts distributor, cut the turnaround on order inquiries from over an hour to under five minutes and saw its conversion rate double. That reads as a new capability, not a cost line. My reading is that Bain's practice runs ahead of its headline instruction.

The disagreement is about what the plan admits at the start. Bain says to isolate the opportunities embedded in it. Kevin says the plan is where the ceiling sits. On that point I side with Kevin.

What a second plan needs

The entry plan stays. Lenders and investors underwrote it, and it remains the floor. The second plan is an upside case, rewritten once the work is visible. I would set that point at around the first 100 days. That timing is my judgement, not a finding.

It needs three things.

A baseline before a target. Spend the first weeks establishing where the work happens, what it costs and who owns it. A plan that prices a lever it cannot measure is a guess.

Room for operating-model levers. New offerings and redesigned operating models go in the plan with their own risk rating, alongside revenue and cost.

One owner with authority across functions. Bain calls the failure a sponsorship gap. The fix is a named executive who can move budget and headcount, not an innovation team running pilots.

The first plan prices the business you bought. The second prices the one AI lets you build.