The view from the board seat
Before Squirrel, I spent almost a decade on the investment side, including board and board-observer roles. I have watched the AI mandate arrive from the fund, and I have watched what happens to it next.
The pattern is familiar. The operating partner presents the portfolio AI initiative, the CEO agrees it matters, and a slide titled "AI Update" appears in the next board pack. Quarter by quarter, it moves down the agenda until it disappears.
Nobody formally cancels the work. It simply loses each prioritisation decision to a more immediate demand.
Why the CEO deprioritises it (rationally)
From the portco CEO's chair, the fund mandate competes with everything else the fund also wants: the pricing review, the bolt-on integration, the new CFO search, the budget. Against those, a generic AI initiative loses for three reasons:
1. It arrives without an owner. The mandate says "explore AI". It does not say who builds, who pays, or what done looks like. The CEO delegates it to whoever has spare capacity, which in a portco is nobody.
2. It arrives without a number. The pricing review promises margin. The bolt-on promises revenue. The AI initiative promises "opportunity". Without a number, the CEO cannot compare it with other calls on the budget.
3. It smells like risk. The CEO has read about AI projects that consumed a year and delivered a chatbot. Doing nothing is safer than being the case study.
None of this is resistance to technology. It is a rational response to a badly specified request.
We hear the same thing from inside portcos. One head of AI, three weeks into the role at a sponsor-backed services business, described the job to us as managing too many cooks: the sponsor pushing tools it uses elsewhere in the portfolio, a consultant review just landed, uncoordinated AI projects running in every team, and no central owner. At enterprise scale the failure looks like tool sprawl instead: one professional services firm we spoke to rolled out a copilot to thousands of staff and watched thousands of self-built agents get created and abandoned within about six weeks. A licence rollout produces activity; adoption needs an owner and a number.
What actually gets adopted
The AI work that survives in portcos usually has three practical characteristics.
It starts inside an existing pain. Ask where the finance team spends eight days a month producing reports the board reads for ten minutes, rather than asking where AI might fit. When the build addresses a pain the team already complains about, people have their own reason to use it.
It ships in weeks. The half-life of management enthusiasm is about six weeks. A build that goes live inside that window gives the team evidence. A programme with only a Q3 milestone plan gives them little reason to stay engaged.
Someone external does the building. Portcos cannot usually hire for this: people who can build production AI systems are expensive, hard to assess, and not needed full time once the systems are live. Giving the mandate to a stretched IT function means asking that team to learn a new discipline under board scrutiny, which is why delivery often stalls there.
How funds should issue the mandate
If you run value creation at a fund, three changes make the difference between a mandate that ships and one that stalls:
-
Send builders, not frameworks. Arrive at the portco with a team that will do the work instead of giving management a maturity assessment to complete. We were investors ourselves before this, so we cover the commercial side as well as the technical. The CEO receives delivery capacity without another internal workstream to staff.
-
Pick the first project for payback speed. The first build is a political asset. Choose something that touches revenue or an acute cost within one quarter. At Healf, the FT's #1 fastest-growing company in Europe, our fastest project paid back in twelve weeks, and that number did more for adoption than any presentation.
-
Report it like any other value creation lever. Put the AI line in the same table as pricing and procurement, with the same discipline: initiative, owner, cost, annualised impact, status. This gives the board the evidence needed to manage it as operating work.
The pattern to watch for
The tell that a mandate is dying is always the same: the board update changes from numbers to activity. "We have shortlisted vendors." "We are running a pilot." "We have formed a working group."
When you see activity language, intervene. Ask one question: what is running in production, and what has it saved or made? If the answer is nothing after two quarters, treat the mandate as failed delivery and change the approach.