A 100-day AI plan for a services portfolio company

The research describes the opportunity with unusual precision. McKinsey's late-2025 survey found 88% of organisations using AI somewhere and only 7% scaled into results across the business. Stanford's AI Index puts corporate AI investment at $582 billion in 2025, up 130% in a year. For an operator the gap is a frustration. For a fund that controls its companies it is a thesis: the portfolio that closes the gap earns operating improvement the market has not priced into entry multiples.

Why services portfolios first

The structural advantage sits with lower-middle-market services platforms: home services, dental, veterinary, insurance, accounting. Those businesses leak revenue in documented, fixable places: missed calls, slow lead response, unworked follow-up, manual back offices. Each leak is high volume, structured and countable, which is where the measured gains in the research sit. Harvard Business School's 776-person field experiment found an individual working with AI matched a two-person team without it; in a portco that is capacity without headcount.

Days 1 to 30: baseline every portco the same way

One page per company: calls and the share answered by a person, time to first response on leads, booked appointments, back-office hours by task, pipeline and win rate. Same definitions across the platform, so the IC can compare. Name one owner per company. Do not start a vendor bake-off; start a measurement.

Days 31 to 60: one engine per company, against its largest leak

Front office where calls are missed, conversion where leads go cold, back-office automation where hours are burned. One engine, one owner, one number, reported weekly against the baseline. The playbook exists for each vertical, so the deployment is a rollout, not a study.

Days 61 to 100: roll the playbook across the platform

What moved the number in the first company is deployed in the rest, with each company's rules. Data stays in the platform's systems. Results are reported to the IC as deltas against the day-one baselines, traceable to the operating lines they touched. What did not move the number is stopped, in writing.

What not to do: enterprise-style pilots without owners, proofs of concept that never meet a baseline, and AI budgets allocated before the leaks are measured. The 7% did not have better technology. They had an owner, a number and a rhythm.