The short answer.

An investor plans AI across portfolio companies by giving each company its own one-page plan, written in the same format, and by sharing the method rather than the workflows. Each portfolio company has its own teams, its own work and its own confidential information, so each company decides for itself which work AI prepares first and who owns it. Across the portfolio you share the plan's format, the way of choosing first workflows, general workflow templates and what each company learned, and you read every company's position on the same six stages of AI at work.

One company's workflows and information never move to another company.

Each portfolio company writes its own plan.

A portfolio-wide AI plan works as a set of company plans, one per company, each written by that company's own leaders. A software company and a food manufacturer share an investor and almost nothing else: different teams, different records, different customers and different work that repeats every week. A plan written at the fund lists tools; a plan written inside each company lists work.

Each company's plan follows the same headings, the ones in How organizations plan their use of AI:

  • Where the company is going.One sentence on what its teams will be able to do with AI by the end of the year.
  • Its teams, then the roles in them.Teams first, named for the work. Roles next.
  • The first workflow for each team.One piece of repeating work per team, what it starts from, what it produces and who decides.
  • An owner for each workflow.One name per workflow, inside the team that uses it.
  • The order and the review.What is being developed now, what is next, and who reviews the page each month.

The same headings in every company let you read five plans quickly and compare them line by line.

What travels across the portfolio, and what stays put.

The method travels across a portfolio; the workflows and the information stay in the company that owns them. A workflow is one company's best way of doing one piece of work, written from its own records, its own standards and its own people's know-how. That belongs to the company, and it often holds confidential customer and pricing information.

Four things are safe and useful to share between portfolio companies:

  1. The format of the plan.The same one page, the same headings, the same monthly review.
  2. The way of choosing first workflows.Score each candidate on how often the work happens, whether it starts from records someone can export, whether one person already reviews it, and what a miss would cost.
  3. General workflow templates.The free templates in the Workflows Directory are samples and a starting point. Each company adapts its own copy to its own work.
  4. What each company learned.Which kinds of work made good first workflows, what slowed a rollout, how the owner's routine settled in. Lessons travel as a short note from the company that learned them.

Three things stay where they are: a company's own workflows, its records and its customers' information. If two companies both want a month-end close checklist, each starts from the same general template and writes its own.

How to see where each company stands.

Read each portfolio company against the six stages of AI at work: not knowing it is possible, switched on, experimenting, planned, built and managed. You can check the stage in one conversation with the company's leader. Is there a one-page plan? Are the first workflows tested on real work and run by the people who own the work? Does every team run current versions with an owner and a monthly review? The stages of AI at work sets out each stage and what moves a company to the next one.

Most portfolio companies sit at switched on or experimenting, and experimenting was the right way to start. The move that matters is from experimenting to planned, and a company makes that move on its own page, with its own people.

Who owns the plan in a portfolio company.

Inside each portfolio company, one person owns the plan, by name: usually the head of operations, the chief of staff or the chief executive. That person keeps the page current, settles the order of work between teams and runs the monthly review. Who should own AI in a company your size? covers the choice in detail.

The operating partner owns the method: the format, the portfolio view and the habit of passing lessons between companies. The operating partner does not own any company's plan, and does not choose its workflows.

A one-page portfolio view to copy.

Copy these five headings into a document, one row per portfolio company. Fill each row from the company's own plan and a short conversation with its plan owner. Review the page once a month.

  1. Company.Its name and the kind of business it is.
  2. Stage.One of the six stages of AI at work, checked against its plan.
  3. Owner of the plan.One name inside the company.
  4. First workflows.The workflow being developed now in each team that has one, and who decides on its output.
  5. What it can share.Lessons and general templates the company is willing to pass on. Never its own workflows or information.

An example across three companies.

Illustrative example.

Take an operating partner whose portfolio holds a 45-person commercial HVAC services company, a 120-person specialty food manufacturer and an 80-person software company. Each company writes its own one-page plan with the shared headings. The HVAC company is experimenting; its plan owner is the operations manager, and its first workflow prepares the next day's service visit briefs for the dispatch team, with the dispatcher deciding. The manufacturer is switched on; its chief of staff owns the plan, and the first workflow drafts supplier quality follow-ups for the quality lead to approve. The software company is planned; its head of operations owns the plan, and support and finance each have a first workflow in testing. All three start from general templates in the Workflows Directory and write their own versions. The software company writes a short note on how its support team tested a workflow on last month's cases, and the other two use the note. No workflow and no customer record moves between the companies.

What is different when each company has under 200 people.

A portfolio company of under 200 people plans AI quickly and on one page. Most published advice assumes a large enterprise with a platform team and a steering committee, which a company of 45 or 120 people does not need.

  • The leader can name every team.The plan fits on one page and the plan owner fills it in with each team lead in one sitting.
  • The company already pays for an assistant.Workflows run in the company's own ChatGPT, Claude, Copilot or Gemini, on its own account, run by its own people. No new platform comes first.
  • The plan owner is an operator.Not an IT department and not the fund. The person who runs operations runs the plan.
  • Every team can have its first workflow soon.The people who decide sit in the same room, so the plan covers the whole company rather than one pilot.

This is where 10X the ROI of your existing AI comes from in a portfolio: the seats each company already pays for, pointed at named work, by companies small enough to decide quickly. How to manage AI across teams without everyone doing their own thing covers the same discipline inside one company.

Where to start.

Start with one portfolio company that has a willing plan owner, write its one-page plan, and use the format for the next company. Private equity and portfolios shows the work portfolio companies put AI to, and AI workflows for a private equity portfolio company shows the work AI prepares for a portfolio company's owners, from the monthly pack to the board pre-read. Each company's plan lives in FullyPowered Planner, where its leaders set out their teams, roles and the work between them, choose what to build first, and add colleagues to their own teams. When a company wants the workflows developed with it, a 3-Day Workflow Sprint or a 2-Week Workflow Sprint takes one piece of work to a tested workflow its people run.