Investors and portfolio companies

The gap you can get ahead of

The mood in private equity is oddly split. In Grant Thornton's Private Equity Pulse survey (2026), 72% of UK respondents called AI more hype than impact, while a majority of the same market keeps AI near the top of the value-creation agenda. Both positions are rational, because most firms have adoption without measurement: tools in use across the portfolio, spend on the P&L, and no defensible line from either to EBITDA. If that sounds like your portfolio, you are closer than you might think: the tools and the activity are already there, and measurement is the part that can be added quickly.

Timing is on your side too, if you move inside the hold. EY's 2026 research found 86% of general partners saying that structured preparation in the final twelve to twenty-four months improves valuations, and AI claims are now part of what buyers' advisers test. A story that survives that testing is a lever; a story that collapses under it becomes a price chip on the other side of the table. Test your own stories first, on your own clock, and you get to fix anything the testing finds before it ever meets a buyer. Giving you that early, unhurried position is the review's whole job.

The Portfolio AI Review

You choose three to five portfolio companies. Over three to four weeks we screen each one and answer the three questions you actually have about it: which of its AI claims are real, where its material risk sits, and where it is leaving value on the table. It is triage, priced and scoped as triage: it tells you where to look harder, and it says plainly what a screen cannot establish.

What we examine

The artefacts each of your companies actually has: system lists, model documentation, evaluation evidence, spend, governance papers, contracts by title. We sit down with each management team twice, once on the commercial reality and once on the technical, and run a challenge session wherever a claim is material to value.

What you get on each company

A one-page summary you can circulate: what its AI estate actually is, the three findings that matter with the evidence behind each, scores across six dimensions from value linkage to key-person dependency, and the top risks with suggested owners. Each company also keeps its model-risk register: every system listed with what it does, who owns it, what is owned versus rented, and what would detect degradation.

What you get on the portfolio

A heat-map across your companies and the six dimensions, the shared exposures you are carrying (common vendor dependencies, repeated governance gaps), the transferable wins where one company's solution fits another's gap, our suggested sequencing for where your attention buys the most value first, and a ninety-minute readout with your partners.

What it costs

A fixed fee, agreed before we start and scaled to the number of companies you choose, from a three-company minimum. Where the screen finds something that deserves a full examination, per-company deep dives are scoped and priced separately.

How the scoring stays honest

Every finding is graded by the strength of the evidence behind it: measured, verified against an artefact, demonstrated to us, or asserted to us. The grade travels with the finding into the report, and one structural rule does most of the work: no dimension can score above the mid-point on asserted evidence alone. Scores move on new evidence, not new narrative.

Where a claim is material to value, we test it with a structured challenge rather than a checklist: walk the headline capability from input data to business action; show us yesterday's real output and what happened next; tell us what users would notice if the model were switched off tomorrow; show us the thing that would reveal it quietly getting worse; and tell us exactly what the company owns, model, prompts, data or none of it. A frequent pattern in the market is rented capability described as proprietary. It is material to valuation, and the review's job is to find it while there is still time to put it right, on your terms rather than a buyer's.

What you are buying is a structured examination of documented findings against stated criteria at a date, referencing recognised frameworks for AI management and assurance. It is not an audit and does not certify anything; at screening depth we do not read code or test models against ground truth, and the report says so. Where those depths matter, that is what the deep dive is for.

How your companies experience it

However carefully a fund positions it, a commissioned review can feel like an audit from the company's side of the table, and defensive management teams produce poor findings. So this one is designed to be worth having, from their side as well as yours. Each interview starts with what that company wants from AI, not an interrogation. Each of your management teams receives a private feedback note with its own findings, quick wins under its own control, and credit for what is genuinely working, in a form its board can use, and none of them sees another's findings. Teams that engage candidly get something valuable back, which is also what keeps the evidence you receive honest.

The people in those rooms are Ballista's founders, personally. You can meet them here before we meet your companies.

Proving one workflow

Where the question is not "what is real across the portfolio" but "what is this initiative actually worth", the sharpest instrument is the Evidence Sprint: ten days inside one portfolio-company workflow, measured with and without AI against pre-agreed success criteria, ending in a go, change or stop decision with the evidence pack to defend it. It is the fixed-fee way to turn one of your companies' AI claims into the kind of measured result an exit narrative can lean on, and a low-risk way for you to evaluate how we work.

When a deal is live

When a deal is live, the same discipline applies under exclusivity timescales: what is actually built, whether the intelligence is owned or rented, what the data assets really are and what licences underpin them, whether the team can sustain what it has built, and whether the roadmap is deliverable. Findings arrive as evidence, assumptions, risks and unknowns, with the cost of each risk sized in engineering effort rather than adjectives, in language an investment committee can interrogate line by line. Diligence work is scoped per deal; if it is on your horizon, talk to us early.

In the hold period

Findings become value only when someone acts on them. Where a review or diligence exposes work worth doing, the same team can carry it: delivery and integration work inside portfolio companies, sequenced around what each company can absorb, and fractional technology leadership for the season between the deal and the permanent hire. Your companies engage us directly, with the work scoped to their own interest, which keeps the incentives clean and your management relationships intact.

Independence and boundaries

Assessment is kept separate from any later delivery interest: we do not soften findings to sell the remediation, and we surface potential conflicts before you find them. Conversations and materials are handled in confidence as a matter of course. We do not provide legal, accounting or regulated financial advice, and our reports say what they are: an independent technical and commercial assessment, addressed to you, offered to inform your judgement rather than replace it.