How to Build an AI Training & Support Program Your Portfolio Companies Actually Use

Key Takeaways

• Train on real workflows. Sessions built on each role's own tasks and documents produce habits. A feature tour produces a few weeks of curiosity.
• Name champions in every portfolio company. A respected user in each function, with protected time and a direct line to the support team, answers the questions that otherwise end in an idle seat.
• Plan support for the months after launch. Office hours, a shared prompt and Skill library, and refresh sessions as the tools ship features keep usage up once the launch energy is gone.
• Make managers set the expectation. In Section's research, manager expectations showed a stronger link to AI proficiency than training did.
• Measure task completion inside real workflows. Logins show who opened the tool. Workflow runs and completed tasks show whether the program works.

A portfolio company buys AI seats for every office employee. The operating partner books a launch webinar, IT sends the login instructions, and the usage dashboard looks healthy for a few weeks. By the next quarterly review, a small group uses the tool every day, most seats sit idle, and the people still logging in use it for the same quick lookups they used to run through a search engine.

The gap is in the training. In BCG's 2026 AI at Work survey, 72% of respondents said the skills expected of them have changed, yet only 36% felt they had received adequate upskilling. The same survey found that 66% of frontline employees who save time with AI get limited or no guidance on what to do with it.

For a PE firm, that gap repeats at every company in the portfolio, each with different systems, roles, and starting points. A single launch event cannot close it. A training and support program can: one that teaches people on their own work, gives them someone to ask, keeps up as the tools change, and measures whether the work actually gets done.

This guide walks through how to build an AI training and support program your portfolio companies will use, in four steps, with tables you can lift into your own plan and a checklist to grade the program you have today.

What Is an AI Training and Support Program?

An AI training and support program is the ongoing system a company uses to teach people to use AI tools in their own work, help them when they get stuck, and check that the tools are used for real tasks. The first workshop is part of it. Most of the work happens in the months after.

A launch webinar is one event. A program has owners, a calendar, and a metric.

For private equity, this is a portfolio-level problem. In FTI Consulting's 2026 Private Equity AI Radar, talent was the primary constraint to scaling AI adoption, cited by 35% of respondents. Licenses are easy to buy. People who know how to use them on Monday morning are harder to come by.

What are the core components of an AI training and support program?

Every program that holds up after launch has the same seven parts. Most failed rollouts skipped at least three of them.

Tip: To place each portfolio company before you design its training, use the four benchmark questions in our post on the 4 AI proficiency levels.

Why Do Portfolio Company AI Rollouts Get Early Adoption, Then Lose It?

Launch week has everything going for it: a new tool, leadership attention, and a calendar invite. Then the same five things wear usage down.

• The training covered the tool. A feature tour shows what the tool can do. It does not show an AP clerk how to clear the exceptions queue faster. In Section's AI Proficiency Report, 14% of workers said their most valuable AI use case was replacing Google search. That is what generic training produces.
• Nobody is there to ask. The first time a prompt returns a wrong answer or a connector fails, the user goes back to the old way of working, and usually stays there.
• Managers stay neutral. If nobody in the review chain asks whether a draft started with AI, the habit never becomes part of the job.
• The tools keep changing. AI vendors ship features every few weeks. Training content from launch day ages fast, and nobody shows the team what changed.
• The dashboard counts logins. Leadership sees active seats, assumes the program worked, and moves on.

Each of the four steps below fixes one or more of these causes.

Step 1: How Do You Anchor AI Training to Real Portfolio Company Workflows?

Real adoption comes from workflow fit and habit formation. A single training event can produce neither. Start every curriculum from the work each role already does.

Pick three to five workflows per role

Sit with each function and list the tasks that happen every week and have a clear finished output. Pick three to five per role. These become the curriculum, the practice material, and later the workflows you measure.

Train in the company's own environment

Run sessions inside the company's own AI workspace, with its connectors, shared projects, and real (non-sensitive) documents. When the practice environment is the work environment, there is nothing to translate afterward.

Give people enough hours

One afternoon is not enough. In BCG's 2025 AI at Work survey, regular usage was sharply higher for employees who received at least five hours of training. Plan for at least five hours per person in the first month, split into short sessions that each cover one workflow.

Teach people to check the output

Every curriculum needs a short block on where AI fails, how to verify an answer, and which data never goes into a prompt. In a PE-backed company, a wrong figure in a lender update costs more than the time the tool saved.

Step 2: How Do You Identify and Support AI Champions in Each Portfolio Company?

Champions are the people a colleague asks first. Every portfolio company needs at least one in each function the program covers.

Pick champions by the work they do

The best champion does the target workflow every day and is already the person peers ask for help. That is often a senior analyst, a controller, or a support lead. It is rarely the IT team, who know the tool but not the work.

Give champions time and a direct line

Champion work done after hours stops within a quarter. Treat it as part of the job.

Get managers to set the expectation

Champions help people who want to use AI. Managers decide whether people want to. In Section's AI Proficiency Report, employees whose managers expect AI use were 2.6x more proficient than those whose managers discourage it, while trained employees were 1.5x more proficient than untrained ones. Only 7% of individual contributors said their managers expect daily AI use.

Brief each company's leadership team on which workflows are in scope and ask them to request AI-first drafts in the reviews they already run.

Connect champions across the portfolio

This is where a PE firm has an edge over a single company. A monthly call that brings champions from every portfolio company together lets a controller at one company share a variance commentary workflow with the finance team at another. The operating partner gets one view of what works, and good workflows travel across the portfolio.

Step 3: How Do You Build Ongoing, Embedded AI Support After Launch?

Portfolio companies with the highest sustained usage have embedded support that runs long after the launch webinar. Support is what turns the first month of training into a lasting habit.

Run weekly office hours for the first three months

A standing weekly slot where anyone can bring a stuck prompt, a failed connector, or an idea for a new use. Keep it weekly for the first three months, then move to every two weeks once the questions slow down.

Keep a shared prompt and Skill library

When someone builds a prompt or workflow that works, it should not stay on one laptop. A shared library of prompts and Agent Skills, owned by the champions and reviewed by the program lead, turns one person's habit into a team asset.

Refresh training as the tools ship features

Hold a short monthly session that shows new features on workflows people already run. Showing the feature on familiar work gets it used. A release note rarely does.

Write down the escalation path

Users go to their champion first, then to the support team, then to the admin or IT owner. Questions about sensitive data go straight to compliance. Write it down and share it, so nobody gives up because they did not know who to ask.

Tip: PressW runs role-based Claude training, champion programs, and weekly office hours inside each company's own environment. See how on our AI Training & Support capability page.

Step 4: How Do You Measure AI Adoption by Outcomes Instead of Logins?

Measuring adoption by logins is the wrong approach. A person who opened the tool once in a month counts the same as a person who runs three workflows a day. Measure task completion inside real workflows instead.

Take a baseline before launch

For each target workflow, record how long it takes today and how often it runs. Without a baseline, the business outcome rung of the ladder is guesswork.

Use the same definitions across the portfolio

If one company counts a "completed task" as any AI-drafted document and another counts only drafts the manager approved, the operating partner cannot compare them. Set the definitions once at the firm level and have every company report against them.

Turn every review into actions

A monthly usage review should end with decisions: reassign idle seats, schedule a refresh session for the workflow with low completion, publish the workflow one team built to the shared library. A report without actions is just a report.

Tip: To see adoption, spend, and workflow use by team in one dashboard, see our Token & Usage Analytics capability.

How Do Strong PE Firms Avoid Common AI Training Mistakes?

Firms whose portfolio companies keep using AI after launch plan the program before they buy the seats. The ones that stall make the same few mistakes.

Common pitfalls

• Treating the launch webinar as the program. One event creates awareness. Habits need weeks of practice and someone to ask.
• One curriculum for every role. A controller and a sales rep do different work. Generic training gets generic use.
• Champions without time. Naming champions and adding nothing to their calendar produces a title and no extra help.
• Training on demo data. If people never practice on their own documents, the first real task is where they get stuck.
• Reporting logins to the board. Seat activity says nothing about value, and the board will eventually ask about value.
• No plan for new features. The tools change monthly. A curriculum frozen at launch day falls behind within a quarter.
• The same program at every company. A portfolio company still learning the basics needs a different plan than one already building shared workflows.

Checklist for Evaluating Your AI Training and Support Program

Use this to grade the program at each portfolio company, whether your team runs it or a partner does.

• Three to five target workflows are defined for each role in scope
• Training runs in the company's own environment, on its own documents
• Each person gets at least five hours of hands-on training in the first month
• The curriculum includes checking output and handling sensitive data
• At least one champion is named per function, with protected time
• Managers have been briefed and ask for AI-first drafts in their reviews
• Office hours, a shared prompt and Skill library, and a written escalation path are in place
• Refresh sessions are on the calendar for new features
• A baseline was taken for each target workflow before launch
• Workflow runs and task completion are reviewed monthly, with the same definitions across the portfolio

Build an AI Training & Support Program with PressW

AI training sticks in portfolio companies when it starts from real workflows, has champions with time to help, keeps going after launch, and is measured by completed work. Get those four right and the seats you already bought start to pay for themselves.

PressW is an Official Anthropic Partner. Our AI Training & Support service runs hands-on Claude onboarding workshops on your team's own tasks, role-based tracks for executives, analysts, operators, and engineers, Claude Code training for engineering teams, and champion and leadership programs. We hold weekly office hours, add sessions as Claude ships new features, and track adoption by team. Training sits inside our Managed AI Services, alongside Claude deployment and usage analytics, so the team that trains your people also sees how the tools get used. See how we work with private equity firms.

If you are planning an AI rollout across your portfolio, or one is already losing steam, we will look at where each company stands and what its program needs. Book a Discovery Call.

FAQ