How private equity firms use AI: a working guide
Where AI does real work in private equity, split by deal team and portfolio company, what recent surveys show, and a sensible order to start in.
King & Company
In short
Private equity firms use AI in two separate places: on the deal team for screening, diligence reading, and memo drafting, and inside portfolio companies for finance, customer-facing work, and sponsor reporting. Recent surveys show wide experimentation and very little at enterprise scale, because the value sits in specific workflows that have to be rebuilt against each company's own documents with a named reviewer. The order that works is to train the people, pick one workflow, build it until its owner uses it weekly, and then repeat it.
The clearest way to understand how private equity firms use AI is to ask whose week the work comes out of. On the deal team it is the associate reading a data room and the VP drafting the investment committee memo, and inside a portfolio company it is the controller closing the books, the customer success manager preparing for renewals, and the analyst assembling the board package.
Those are two different jobs with different owners, different documents, and different risks, and most firms that are making progress treat them separately. This guide maps both, says what the surveys show about how far along the industry is, and suggests an order to start in.
Where AI is in private equity right now, according to the surveys
Activity is wide and depth is thin. FTI Consulting's 2026 Private Equity AI Radar, a survey of 200 fund and operating leaders, reports 36% of portfolio companies using AI across use cases and 7% at enterprise scale. Bain's survey of investors representing $3.2 trillion in assets, taken in September 2024, found that a majority of portfolio companies were in some phase of generative AI testing and that nearly 20% had operationalized use cases and were seeing concrete results.
The sponsor side looks similar. Accordion and Wakefield Research surveyed 150 operating partners in May 2026 about AI in portfolio company finance functions and found that 41% of firms are deploying AI across multiple companies without an operational playbook, and fewer than one in ten has a fully operational AI center of excellence.
There is one encouraging number beside those. In the same FTI survey, 95% of funds report AI initiatives meeting or exceeding their original business case. Read together, the figures say that the projects firms finish tend to pay off, and that few firms have finished many of them.
Two different jobs: AI for the deal team and AI inside the portfolio company
| Deal team | Portfolio company | |
|---|---|---|
| Whose week | Associate, VP, principal | Controller, customer success manager, FP&A analyst |
| Source material | Data room, CIM, management presentations, expert call notes | General ledger, CRM, contracts, support tickets, prior board packages |
| Output | Screening notes, diligence findings, a draft IC memo | A closed month, a renewal brief, a board package |
| Who reviews | The deal lead and the investment committee | The function head who already signs off on that output |
| Main risk | A confident summary of a document nobody re-read | A workflow built once and never adopted |
The sponsor controls the first column directly. The second column belongs to each company's management team, which is why portfolio work moves at the speed of the company and not the fund.
What deal teams use AI for: screening, diligence reading, and memo drafting
Deloitte's 2025 survey of 1,000 corporate and private equity leaders found that 86% of responding organizations had integrated generative AI into their M&A workflows. Among adopters, 40% apply it to strategy and market assessment, 35% to target identification and screening, and 35% to due diligence.
In a working week, that breaks into three tasks.
- Screening. Reading a CIM or teaser against the fund's own criteria and producing a one-page view in the format the team already uses for its Monday pipeline meeting.
- Diligence reading. Working through customer contracts, employment agreements, and financial schedules in a data room, pulling the terms the team cares about into a table, and citing the page each one came from so a person can check it.
- Memo drafting. Assembling the first draft of the IC memo sections that are mostly synthesis (market, company history, diligence findings) from the team's own notes and prior memos, leaving the thesis and the judgment to the people who own them.
The same Deloitte survey found that 67% of respondents named data security as a leading concern, followed by data quality and availability at 65%. Much of the response to both sits in how the workflow is designed, with source citations on every extracted item and a written rule about where confidential deal documents may go. That rule is one for the firm's compliance lead and counsel to set. We cover that design in how to use AI in private equity due diligence.
What portfolio companies use AI for: finance, customer-facing teams, and reporting to the sponsor
Portfolio company work is closer to operations than to analysis, and it clusters in three places.
Finance. Variance commentary, reconciliation support, accrual schedules, and the narrative that goes with the monthly close. The controller's team already produces these from the same systems each month, which makes them good candidates for a rebuilt workflow.
Customer-facing teams. Account research before a renewal, call preparation, summaries of support history, and first drafts of proposals from prior ones. FTI's respondents named revenue acceleration as the top priority, cited by 41%, which points sponsors toward this group.
Reporting to the sponsor. The board package and the monthly KPI submission, assembled from finance, sales, and operations data by an analyst who spends days collecting inputs before writing a word.
In each case the workflow has the same shape. AI does the collecting and the first draft against the company's own data, and the person who already owned the output reviews it at the point where they reviewed it before.
Why most portfolios stall between pilot and production
The surveys name several causes, depending on who was asked. Accordion's operating partners, asked what keeps portfolio CFOs from scaling AI, ranked data infrastructure first at 71%, followed by talent at 63%, legacy technology at 58%, and change management at 52%. FTI found talent to be the primary constraint to scaling, cited by 35% of respondents. BCG's survey of 100 senior PE investors found 90% citing competing priorities as the top blocker to digital transformation and 76% citing unclear ROI.
Our reading of those numbers is that a license is the easy part. The value sits in specific workflows that have to be rebuilt against one company's documents and systems, with a named person reviewing the output. That is work done by people inside the company, during weeks that are already full, and a portfolio company that buys seats without assigning that work ends up with a pilot that a few enthusiasts use.
Measurement is the other gap. In EY's AI Pulse survey, 31% of PE respondents strongly agreed and another 31% somewhat agreed that their organizations struggle to link specific productivity gains to AI adoption. A gain is much easier to show when it belongs to one named workflow with a before and an after.
A sensible order: train the people, pick one workflow, build it, then repeat it
We think the unit of progress is one working workflow that its owner uses every week. Everything else in the sequence serves that.
- Train the people first. A team that has used AI on its own documents can tell you which of its tasks are worth rebuilding. Firms are already budgeting for it: in EY's AI Pulse survey, 72% of PE firms expected to spend on employee training as part of their responsible AI commitments, against 65% of private companies overall. The reasoning is in why AI training comes before AI tools at a portfolio company.
- Pick one workflow. Choose something frequent, document-heavy, and owned by one person who wants it fixed. Our guide to choosing the first AI workflow at a portfolio company covers the selection.
- Build it against real work. Use last month's close or last quarter's board package as the test, and keep going until the owner uses it without being reminded.
- Repeat it. A finished build at one company is the start of the playbook that many sponsors are scaling without. Bain describes three ways sponsors spread what works: Vista asks portfolio companies for generative AI goals in annual planning, Apollo runs a center of excellence, and Hg encourages its similar software companies to share solutions with each other. We describe the mechanics in how to repeat one AI build across a private equity portfolio.
Who should own the work, and what the company should own when it is done
The owner of each workflow is the function head whose team produces the output, with the operating partner as sponsor and not as builder. In our view, a workflow that is owned by the fund and handed to a company is at risk of being used only for as long as someone from the fund is asking about it.
The build itself should belong to the company: the instructions, the templates, the integrations, and the documentation, all sitting in the company's own AI workspace alongside its data. That matters at exit. Asked whether AI-enabled finance commands a premium, 44% of Accordion's operating partners said buyers are asking in diligence but the premium has not yet shown up in price, and BCG found that only 11% of investors explicitly link digital progress to exit narratives. When a buyer asks what transfers, a documented workflow the company owns and runs is a clear answer.
What to ask before you hire anyone to help
BCG found that only 45% of successful firms systematically ensure knowledge transfer from external partners to internal teams. These questions test for that before the work starts.
- Who does the building, and is it the same person who scoped the work?
- Will the work be done inside the company's own workspace and systems, on its own documents?
- Who at the company will be using the workflow on the day the engagement ends, and how were they trained?
- What exactly does the company own at the end, and is there any platform fee or dependency on the outside firm to keep it running?
- Where does the review step sit, and who signs off on the output?
This is the kind of work we do as embedded engineers inside portfolio companies, and if you are forming a view for your firm or your sponsor, you can get in touch to compare notes.
Common questions
How are private equity firms using AI in 2026?
Deal teams use it for market assessment, target screening, diligence reading, and first drafts of investment committee memos. Portfolio companies use it in finance, in customer-facing teams, and in assembling reporting for the sponsor. Most of this is still in pilot or in a handful of use cases per company.
What percentage of PE portfolio companies are using AI in production?
FTI Consulting's 2026 Private Equity AI Radar, a survey of 200 fund and operating leaders, reports 36% of portfolio companies using AI across use cases and 7% at enterprise scale. Bain's September 2024 investor survey found nearly 20% of portfolio companies had operationalized generative AI use cases with concrete results.
What is the biggest barrier to AI adoption in portfolio companies?
The surveys name two. Operating partners in Accordion's 2026 benchmark, which covers portfolio company finance functions, ranked data infrastructure first at 71%, ahead of talent at 63% and legacy systems at 58%. FTI's respondents named talent as the primary constraint to scaling. Our reading is that both come down to whether someone inside the company has the time and skill to rebuild a workflow against its real data.
Should a PE firm build an AI center of excellence or work company by company?
Start company by company and let the central function form around what works. Accordion found fewer than one in ten sponsors has a fully operational center of excellence, and a central team is most useful once it has one finished build to carry from company to company.
Does AI in a portfolio company increase exit value?
The evidence so far is that buyers ask and rarely pay for it yet. In Accordion's 2026 benchmark, which asked about AI in the finance function, 44% of operating partners said buyers ask about it in diligence but the premium has not shown up in price, and 9% had seen a clear premium in a completed transaction. A company that can show a working, documented workflow it owns has a better answer to that question than one that can show licenses.