AI in Private Equity: Where Value Starts | ClearWork

AI in Private Equity: Where AI Value Comes From in Portfolio Companies

Avery Brooks
November 16, 2026

Updated: November 16, 2026

AI in private equity creates value when a portfolio company applies it to specific, high-volume work it understands in detail. Start with a role-level baseline of how the work runs today, build a ranked backlog of AI opportunities, measure each one before and after, then repeat the method across the portfolio rather than running one-off pilots.

What sponsors are asking portfolio companies for

Most sponsors now expect every portfolio company to show progress on AI. Accordion's survey of PE sponsors and portfolio CFOs (2025) found that 98% of sponsors have directed their CFOs to prioritize AI adoption. The same survey found that most CFOs had not acted, and 68% of them said the main reason was not knowing where to begin or who to turn to for help.

The request usually arrives in three forms. The board wants an AI section in the value creation plan with named initiatives and owners. The operating partner wants a short list of use cases that will move EBITDA inside the hold period. And at exit, buyers will want to see AI that is running in the business with measured results, not a list of experiments.

For a portfolio company CEO or COO, that adds up to a practical question: which parts of our operation would AI actually improve, by how much, and how will we prove it? Answering it starts with the work, not with the tools.

Why AI pilots stall

The results so far are thin. PwC's analysis of PE-backed CEOs in its 2026 Global CEO Survey found that only 14% say AI has contributed to both higher revenues and lower costs, and more than half report no upside at all.

Pilots tend to stall for the same few reasons:

  • They start from a tool, not a process. A team picks a product, then looks for somewhere to use it. The use case fits the demo better than the business.
  • Nobody knows how the work runs today. The process lives in people's heads, in exceptions and in spreadsheets on one desktop. You can't automate a process that nobody has written down.
  • The volume is too small to matter. A pilot saves a few hours a week in one team and never scales.
  • There is no baseline. Without a measured before, nobody can show the after. EY's AI Pulse survey (2025) found 62% of PE respondents say their organizations struggle to link productivity gains to AI adoption.

BCG's work on AI-first portfolio companies (2026) attributes about 70% of AI value to people, organizations and processes, with 20% from technology and 10% from algorithms. The tool is the smaller part of the problem.

Start from the work: a role-level baseline

A role-level baseline describes, for each role in a focus area, what the work is, how often it happens and where it goes wrong. It is the map an AI backlog is built on.

For each role, capture:

  • The tasks the role performs, step by step, including the decisions and business rules behind them.
  • Volumes and frequency: how many invoices, orders, tickets or quotes pass through each week.
  • Exceptions and workarounds: the cases that don't follow the standard path and how people handle them.
  • Systems and data each step touches, including the spreadsheets and inboxes that sit outside the core systems.
  • Handoffs between roles, where work waits or gets re-keyed.
  • Pain points in the words of the people who do the work.

The fastest way to get this is to hear from everyone who does the work, not a sample. ClearWork is an AI discovery platform for processes and requirements: its AI agent, Clarity, runs short async interviews (voice or chat, about 20 minutes) with the people in each role, people can add screen recordings of how a task is done, and ClearWork ingests existing documents. Process owners validate each round, every line traces back to its source, and the outputs include improvement and AI opportunities by role. If a focus area needs more detail, run another round on it.

If you have already run an AI readiness assessment for operations, the current-state work there is the starting point for this baseline.

Build and rank the AI backlog

With the baseline in hand, list every place AI could change the work. Most opportunities fall into a few patterns: reading and keying documents, answering repeat questions, drafting routine outputs, checking work against rules, and routing work to the right person. The baseline tells you where each pattern shows up and at what volume.

Then rank the list. Score each opportunity on value, effort and risk, using the evidence from the baseline rather than opinion. The table shows the columns to keep for every line.

OpportunityRoleWeekly volumeEffortRiskValue
Extract data from supplier invoices that arrive by emailAccounts payable clerkFrom the baselineLow to mediumLowHours saved, faster close
Draft first responses to repeat customer order questionsCustomer service repFrom the baselineMediumMediumResponse time, capacity
Check purchase orders against approval rules before releaseProcurement analystFrom the baselineMediumLowFewer errors and exceptions
Pre-fill quotes from past orders and the price bookInside salesFrom the baselineMediumMediumQuote turnaround
Summarize service tickets for the weekly operations reviewOperations managerFrom the baselineLowLowManager time

Score effort with the people who own the systems and data, and score risk with the process owner. For a fuller scoring method, the AI use case prioritization framework sets out impact and feasibility criteria you can reuse. Keep the value column in the units the board tracks, and enter your own estimates; the baseline gives you the volumes to support them.

A good backlog has 10 to 30 lines per portfolio company. The top five become the AI section of the value creation plan.

Measure it: baseline before, results after

Every opportunity that moves into delivery needs a measured before. Pick two or three metrics per opportunity, record the baseline value, the date and where the number came from, then measure the same thing after the change has run for long enough to be fair.

Useful metrics are usually operational:

  • Cycle time for the process (for example, order to invoice).
  • Volume handled per person per week.
  • Error, rework or exception rate.
  • Time to respond to a customer or supplier.
  • Hours spent on the task, from time studies or system logs.

Agree the metrics with the CFO before work starts, so that results feed the KPI tracker the board already reviews. If a metric can't be measured today, that is a finding in itself: either add the measurement first or pick a different metric.

Measurement is also what makes AI credible at exit. A short record for each initiative, showing the baseline, the change, the result and the owner, is far stronger evidence than a list of tools in use.

Repeat across the portfolio

Sponsors get the most from AI when what works in one portfolio company is reused in the next. That depends on doing the work the same way each time.

  • Use one backlog format across portfolio companies, so opportunities can be compared and rolled up.
  • Keep a library of proven opportunities. An invoice-extraction use case that worked in one distribution business is a strong candidate in the next one, but check it against that company's own baseline first, because volumes and exceptions differ.
  • Roll up status every quarter: top opportunities per company, stage, owner and measured results.
  • Share what didn't work as well as what did, so the next company skips the same dead ends.

The operating partner's job here is to set the method and the format, not to pick every use case. Each portfolio company still owns its own backlog, because the people who do the work know where the real volume and pain sit. For more on how operators work with portfolio companies, see ClearWork's page for private equity operating teams.

Portfolio AI opportunity backlog template (download)

The free template puts the method in this article into one workbook. Download the portfolio AI opportunity backlog (Excel).

  • How to use: the steps, from baseline to roll-up, and how the tabs fit together.
  • Backlog: one row per opportunity, with the process, role, task, weekly volume, current pain, AI pattern, systems and data, owner and stage.
  • Scoring: value, effort, risk and data-readiness scores from 1 to 5, a total score and a rank, with notes on the evidence behind each score.
  • Measurement baseline: the metric, its baseline value, the date measured, the source, the target, the result after the change and the date of that result.
  • Portfolio roll-up: one row per portfolio company, with its top opportunities, stage, owner, measured result and next review date.
  • Illustrative example: a filled-in backlog, scoring and baseline for an invented regional distribution company.

Start with the Backlog tab during baseline work, score in a working session with process owners, and update the roll-up before each quarterly board meeting.

AI in private equity FAQs

Where should a portfolio company start with AI?

Start with one or two focus areas where volume is high and the work is well understood by the people who do it, such as accounts payable or order entry. Build a role-level baseline there first. The baseline shows where AI would remove real work and gives you the numbers to measure against.

How long does it take to build an AI backlog?

It depends on scope. The discovery that produces the baseline typically takes one to four weeks, depending on how many processes and people are involved. Scoring and ranking the backlog usually takes a few working sessions with process owners once the baseline exists.

Who should own AI in a portfolio company?

Each opportunity needs a business owner, usually the leader of the process it changes, not only the IT or data team. The CFO or COO typically owns the overall backlog and reports progress to the board. The operating partner sets the method and compares results across the portfolio.

How do you measure AI value in a portfolio company?

Record a baseline for each opportunity before work starts, using operational metrics such as cycle time, volume per person or error rate. Measure the same metrics after the change has run for a fair period, and report the difference in the units the board already tracks.

Will interviewing employees about their work unsettle them?

It can if it feels like an audit. Explain that the goal is to find work that is slow, repetitive or error-prone and make it easier, and share what you learn with the people involved. Process owners validate the findings, so people see their own work described accurately.

How is a portfolio AI backlog different from an AI strategy?

A strategy sets direction and principles. A backlog is a ranked list of specific changes to specific work, each with an owner, a value estimate and a baseline. Boards and buyers give more weight to a measured backlog than to a strategy document.

See how ClearWork builds the role-level baseline your AI backlog starts from. Book a demo.

Book a demo to see how ClearWork builds the role-level baseline an AI backlog starts from.

Where AI value comes from in PE portfolio companies: start from a role-level baseline, rank an AI backlog, measure before and after. Free Excel template.