Guide
How to roll out AI across portfolio companies
To roll AI out across a portfolio, treat it like any other operating improvement: prove it at two portfolio companies, package what worked into a playbook, then deploy that playbook company by company while tracking adoption in one view. The work happens with each company's non-technical teams (sales, service, finance and operations) inside the software they already use. Buying seats is not a rollout.
Why seats alone don't change much
Most portcos already have AI in the building. What they lack is setup, training and measurement. In Liberty Mutual's 2026 Independent Agency Growth Study, 65% of agency staff said they had used AI for work in the past year, but only 14% said their agency had actually put an AI tool in place. In Bullhorn's survey of staffing firms, only 10% had implemented agentic AI across their full workflow.
Individual use without company setup leaves the gains scattered: one person saves time, the team's process stays the same, and nobody can show the board what changed.
The four steps
| Step | What happens | What you have at the end |
|---|---|---|
| 1. Pilot at two portcos | Sit with every non-technical team, set AI up for their real work, train each person, and measure usage and hours saved | Working workflows and a baseline |
| 2. Portfolio playbook | Package what worked by function (sales, service, finance, ops) with setup steps for the shared software | A playbook the next portco can run |
| 3. Rollout | Deploy the playbook to the mid-tier portcos, then the top tier, adjusting for each company | Portcos live on the same workflows |
| 4. Portfolio view | Track which portcos are live, who is using AI and what it saves | One view for the operating partner |
Step 1: Pilot at two portcos
Pick the two companies that need it most. Good pilot candidates have:
- A leader who wants it. A pilot without a sponsor stalls at the first busy week.
- Repetitive desk work. Quotes, proposals, follow-ups, service replies and call notes are where AI saves time first.
- The same core software as the rest of the portfolio. If the pilot runs on the system most portcos use, what you build carries over.
- Something you can count. Proposals sent, response times, renewals reviewed: a number you can compare before and after.
Run the pilot team by team. For each team, watch how the work is done today, set AI up inside the tools they use, train each person on that setup, and record usage and time saved.
Step 2: Package the playbook
The playbook is what makes the next portco faster than the first. It should hold:
- Workflows by function, for example drafting quotes, proposals and follow-ups for sales, drafting replies for service, and getting call notes into the CRM.
- Setup steps for the shared software, so the next company's accounts and templates take hours, not weeks.
- Training material built on those workflows, not generic AI tips.
- An AI policy each company can adopt as-is. Our AI policy template for portfolio companies is a starting point.
- Metric definitions, so every portco reports the same numbers.
Step 3: Roll out by tier
Deploy to the mid-tier portcos next, then the top tier. Each rollout repeats the pilot's team-by-team setup and training, adjusted for the company, but starts from finished workflows instead of a blank page.
Step 4: Give the operating partner one view
Track the same few numbers at every portco: active users, workflows live, hours saved and one output measure per team. Our guide to measuring AI adoption across a portfolio covers which numbers to use and where they come from.
Why roll-ups are the easiest place to start
When a PE firm buys up many companies in one industry, those companies tend to run the same systems: Applied Epic or AMS360 at insurance agencies, Deltek at architecture and engineering firms, Bullhorn at staffing firms. A workflow built for one carries to the next with little rework, so the next portco doesn't start from zero.
Common mistakes
- Calling seat licenses a rollout. Licenses are the cheapest part. Setup and training are what change the work.
- Generic training. A class on prompting rarely survives the next busy week. Training on each team's own workflows does. See AI implementation vs. AI training.
- No baseline. Without a before measure, nobody can show what the pilot saved.
- No policy. Staff keep using personal accounts, and client data goes wherever they put it. See how to keep client data safe.
- Piloting where it's least needed. The strongest portco has the least to gain. Start where the gap is widest.
Questions
Should every portco in a roll-up use the same AI tools?
Mostly, yes. When companies in a roll-up run the same core software, one set of workflows, training and policy can carry from company to company, and the operating partner can compare adoption in one view. Leave room for a company to add tools specific to its work.
Do portfolio companies need technical staff to roll out AI?
No. The work is with non-technical teams in sales, service, finance and operations. What they need is someone to set AI up inside their existing software and train each person on that setup.
Which portcos should go first?
The ones that need it most: a willing leader, teams buried in repetitive desk work, and the same core software as the rest of the portfolio, so what works there carries over.