The two vendors you'll meet
Put out a brief for AI help across your portfolio and two kinds of firm turn up.
The first is the strategy consultancy. They will interview your portco management teams, benchmark them against a maturity framework and deliver an intelligent, well-formatted roadmap. The portco still needs someone to build it, because the consultancy's engagement usually ends before production code. Advice was the agreed deliverable.
We keep meeting the aftermath. One sponsor-backed global services business we spoke to had just received an eight-week AI review: intelligent recommendations, presented to the exec, with no view on data readiness, dependencies or priorities. The person hired to act on it called it a nice deck without an execution plan. A Big-6 accountancy backed by a major PE fund told us it had spent a year being pitch-decked out by consultants before anything got built.
The second is the AI engineering house. Its team is genuinely technical and will build whatever you specify, but the client must supply the commercial judgement. Engineers tend to optimise the system rather than the size of the P&L impact when nobody on their side has owned a P&L or sat through an IC. They are poorly placed to decide which of five possible builds matters most to a business eighteen months from exit.
Both firms are good at their jobs. The problem is that a portfolio company in a hold period needs both jobs done by the same brain: judge what's worth building, then build it.
Why the gap hurts PE specifically
A corporate can afford the two-vendor model because it has time and internal translators. A portco operates against a hold-period clock. Value created in year four of a five-year hold barely compounds before exit, so every quarter spent writing specifications, running procurement and onboarding an integrator removes a quarter of impact from the bridge.
The gap also destroys accountability. When the roadmap's promised savings fail to appear, the consultants point at the implementation and the engineers point at the strategy. The operating partner owns the shortfall. We have watched this from board seats, and the pattern is depressingly consistent.
What "both" looks like in practice
The alternative is a team that carries an initiative from commercial judgement through to production system. Concretely, that means:
Prioritisation in deal language. The opportunity scan ranks builds by expected P&L impact against the hold period, not by technical interest. Some of the best advice is "don't build that": a recommendation engineering houses are not incentivised to give.
Building, in the portco, on their systems. The work should run in the portco's actual CRM, ERP and inboxes rather than remain a sandbox pilot. Deliver it in weeks, alongside the team who will use it. We go into the portcos and build, rather than advising from the sidelines.
A number on everything. Every build ships with a measurement: hours saved, conversion uplift, or cost per unit before and after. Boards can then judge the result on evidence rather than a vendor's description of the activity.
This is what we set Squirrel AI up to do. Between us, the founding team spent almost a decade in investment roles at firms including Apax, 3i and Deutsche Bank before crossing to the operating side, most recently leading AI at Healf, the FT's #1 fastest-growing company in Europe, where the projects we led added over £10m of annualised revenue. We were investors ourselves before this, so we cover the commercial side as well as the technical.
How to test a vendor for the gap
If you are evaluating AI partners for your portfolio, three questions expose which side of the gap a firm sits on:
- "Show me something you built that is running in production today." Strategy firms cannot answer this with a demo. Look for a system with live users.
- "Walk me through a build you advised against, and why." This tests an engineering team's commercial judgement and willingness to reduce its own scope.
- "What number did your last three projects put on the P&L?" Anyone who answers in adjectives is selling activity.
Few firms can answer all three well. Funds that find one can move from commercial judgement to production without losing quarters between separate advisers and builders.