--- title: "What does an AI operating partner do in private equity?" description: "What an AI operating partner does in private equity: retained by the fund, deployed where the value is, full-time or fractional, year one KPIs, red flags." url: https://anywayventures.com/insights/ai-operating-partner-private-equity updated: 2026-09-29 --- [← ALL WRITING](https://anywayventures.com/insights) OPERATING PARTNER # What does an AI operating partner do in private equity? Plenty of funds now say they want one. Far fewer can describe the job. Here is what it contains, how a fund should deploy it, and how to tell in year one whether it's working. WRITTENSeptember 2026 READING · Ten minutes FOR · Fund partners, portfolio directors, operating partners RELATES TO[Fractional AI operating partner](https://anywayventures.com/ai-operating-partner) An AI operating partner is the person a fund holds accountable for turning AI from a paragraph in the investment memo into money in a P&L. For most mid-market funds, the version that works is retained by the fund and deployed by the deal team: pointed at the one target or portfolio company where AI can move the numbers most, inside that business until the change is live, then moved to the next priority. If the person in the seat can't point to something running in production, they aren't doing this job. The title is spreading faster than the definition. [Korn Ferry](https://www.kornferry.com/institute/the-ai-operating-partner-the-latest-pe-portfolio-value-creation-role) notes that large PE firms have created AI operating partner roles, filled either by full-time employees or by part-time advisers. In practice the same title covers a former CIO running an IT forum, a data scientist building dashboards and a consultant with a good slide library. Useful people, doing a different job. This piece sets out the job as a deployable resource, how a fund decides where to point it, and how to judge it in year one. > If the person in the seat can't point to something running in production, they aren't doing this job. ## Why funds are creating the role now The [Bain & Company and StepStone Group 2026 Private Equity GP Outlook](https://www.bain.com/insights/private-equitys-reality-check-gp-outlook-2026/) found GPs report their highest returns from generative AI in their own deal sourcing and due diligence. Inside portfolio companies, the benefits skew to cost savings, and nearly 40% of GPs don't expect material financial impact from AI in their portfolio companies in 2026. So the fund's own team uses AI well. The companies it owns, mostly, don't yet. [FTI Consulting's 2026 Private Equity AI Radar](https://www.fticonsulting.com/insights/reports/2026-private-equity-ai-radar) puts a number on the gap. Only 36% of portfolio companies deploy AI in production with real operational impact, and just 7% of respondents describe AI as fully integrated across the portfolio. The top constraint to scaling, named by 35%, is talent. That last figure is why the role exists. What's scarce is people who can walk into a £60m distribution business, find the three workflows worth automating, get them built and keep them running. Most portfolio companies can't hire that person and wouldn't keep them busy. A fund can hold one centrally and send them where they're needed. ## What an AI operating partner actually does as a deployable resource Larger funds run in-house operating teams as a kind of SWAT team, held at the centre and sent into portfolio companies as needed. Bain Capital's portfolio group is the example people usually cite. The retained AI operating partner is the same idea, sized for a lower-mid-market fund. It is deliberately not a light-touch review of the whole portfolio. No baseline of every company, no monthly session with every CEO, no plan for every company. Spread evenly, a few senior days a month buy each business an hour of conversation. Concentrated, they get something built. A standing counterpart to the deal team. Between deployments, the operating partner is on call for the weekly questions: is this AI claim in the information memorandum real, and which company should go next? A brief for every deployment. Before going in, the operating partner and the deal partner agree one page: the outcome, the baseline, and the exit criteria that say when the capacity moves on. No brief, no deployment. Hands-on deployments. Inside the business, with its management team. A one-week diagnostic, then the highest-value changes built and put into production over 4 to 8 weeks, then a handover: documented, owned by a named person, measured against the baseline. Plans without hands are how funds end up with a folder of roadmaps. Diligence on the pipeline. A deal in exclusivity often outranks everything else, because diligence is where the value case and the price get set. The operating partner sizes AI exposure and upside firmly enough to go into the model. It's lighter than full [AI due diligence](https://anywayventures.com/ai-due-diligence) , and a good one knows when to call for the heavier version. 100-day plans for new acquisitions. Three to five named workflows, each with an owner, a cost and a number in hours or pounds. I've set out how to build one in [the AI value-creation plan](https://anywayventures.com/insights/ai-value-creation-plan) . The person who diligenced the target goes back in and builds what the plan promised. Reporting per deployment and to the IC. Each deployment ends with a one-page report to the deal partner: baseline, what went live, hours returned, and what is still running six months on. Each quarter, a summary goes to the IC. [Bain's Global Private Equity Report 2025](https://www.bain.com/insights/field-notes-from-generative-ai-insurgency-global-private-equity-report-2025/) describes Vista Equity Partners asking portfolio companies to submit goals and quantified benefits for GenAI initiatives. That's the discipline worth copying: quantified commitments, reviewed on a rhythm. Exit evidence. A buyer's diligence team will discount anything described as "AI-enabled" that can't be shown working, so a business 12 to 18 months from exit is often the best place to deploy. Systems in production, usage logs, before and after measures and run costs go into the equity story. Assertions get marked down. Patterns that transfer. Portfolio companies share more than they think: similar back offices, service queues and quote-to-cash processes. An invoice matching workflow built in one deployment is mostly configuration the second time, and the knowledge stays with the fund. The fund pays the retainer. Build work inside a portfolio company is usually a fixed-price engagement invoiced to that company, agreed with the fund against its LPA and fee offset terms. ## How a fund decides where to deploy it A retainer is only as valuable as the choice of where it goes. Before each deployment, the deal partner and the operating partner work through six questions. Size of the prize. Hours and margin at stake in work AI can realistically change today, in pounds rather than percentages. Readiness. Whether the data and systems can support it now, or what has to be fixed first and what that costs. Management appetite. A CEO and a named owner who want it. Deploying into a team that doesn't is the fastest way to waste the retainer. Timing in the hold. Early for the 100-day plan, mid-hold when a company is behind plan, 12 to 18 months before exit when the AI story needs evidence. What transfers. Work that can be reused in the next company costs less the second time. The pipeline. A deal in exclusivity usually jumps the queue. Picture a fund with ten companies. This quarter one deal is in exclusivity, one company is behind plan with a willing CEO, and one is heading for a sale next year. The sensible answer is two or three deployments in sequence and patience from the rest. That feels uncomfortable to funds used to treating every company equally. It's also how value gets created. ## How it differs from the roles it gets confused with Four alternatives come up in almost every conversation with funds. None covers the whole job. Alternative What it does well What it usually doesn't do Technology operating partner Infrastructure, ERP, cyber, IT cost, carve-outs Redesign workflows around AI or build AI systems in production CTO or CIO network Peer learning, shared vendor terms, benchmarking Own outcomes or deliver anything inside a specific company Big Four AI strategy engagement Structured assessment, board-ready materials, breadth Stay until it's running, build, or answer for results a year later Vendor partnerships Discounted tools, implementation help for their own product Give neutral advice on whether you need the product at all The technology operating partner is the closest cousin. A good one keeps systems secure, integrated and cheap to run. The AI job sits closer to operations than to IT: it starts with how work gets done and which hours can be returned. Some people do both well. Most are much better at one. ## Full-time or fractional Korn Ferry's split between full-time employee and part-time adviser is the right first decision. My working view: More than about 20 active portfolio companies, a heavy software weighting, or a thesis that leans on AI in every deal: a full-time AI operating partner, who will soon want a small team underneath. 6 to 15 companies and a small operating bench: a fractional AI operating partner retained by the fund, with a set number of senior days each month, deployed by the deal team into one priority at a time, and build work scoped at a fixed price inside the company that benefits. Fewer than about 5 companies, or a single large asset: put the capability inside the company, through an interim leader or an embedded delivery team. I've covered the economics of the full-time hire [in a separate piece](https://anywayventures.com/insights/why-mid-market-funds-cant-hire-an-ai-operating-partner) . The point here is what fractional should mean: retained and deployed, not spread. A part-time adviser who divides the same days so every CEO gets a slice is a different, cheaper and much less useful thing. ## What a good one has done before Three things, and I'd want evidence for each. They have shipped AI systems that run in production in real businesses, and can show you what's still running six months on. Demos don't count. Making a workflow survive messy data, a sceptical finance team and UK GDPR takes different skills. They have sat in an operator's chair: run a P&L and had to choose between the AI project and hitting the quarter. That earns a CEO's trust inside a short deployment. And they can talk to a CFO about EBITDA, with the arithmetic: hours, loaded cost, run cost, payback. The interview questions I'd ask a candidate: Show me a system you built that is still running unattended. What does it do, what does it cost to run, and who maintains it? Tell me about an AI project you stopped. How did you decide, and how long did it take? You have one week inside one of our portfolio companies. What does your diagnostic cover, and what data would you ask for? Size the EBITDA effect of automating invoice matching at a £40m business. Do the arithmetic out loud. Do you receive commission, referral fees or equity from any software vendor? Which ones? Looking at our portfolio and pipeline, where would you deploy first, and what would your one-page report to the deal partner say at the end? Question 4 usually tells you most of what you need to know. ## How to measure an AI operating partner in year one Measure deployments and outcomes, not activity. The year one scorecard I'd set: A prioritisation agreed with the deal team in the first month, with the first deployment named. A signed brief for every deployment: outcome, baseline and exit criteria. Workflows live in production, and how many are still running unattended six months after handover. Hours returned per month in each business deployed into, measured against that deployment's baseline. Annualised EBITDA effect net of run costs, agreed with each company's CFO rather than claimed by the operating partner. Deployments stopped or redirected early, and how quickly. A year with no kills means nobody was looking. Here is what one line of that scorecard looks like, as an illustration. Take a £40m services business with 120 staff, chosen for a deployment because the prize was large and the CEO wanted it. The one-week diagnostic finds around 1,500 hours a month going into three workflows: proposal drafting, invoice matching and first-line customer queries. Six months after handover, systems in production are returning 600 of those hours. At a loaded cost of £35 an hour, that's £21,000 a month, or £252,000 a year. Take off £30,000 a year to run and maintain the systems and you have £222,000. That becomes EBITDA only if the business does something with the hours: slower hiring, absorbed growth, roles not back-filled. A good operating partner reports hours and money separately. If the £222,000 is real, at an illustrative 9x exit multiple it's worth roughly £2m of enterprise value from one deployment. Do that a few times a year and the retainer pays for itself several times over. ## A sample first 90 days The first month is terms and choosing. Agree the retainer with the fund: annual term, senior days each month, the six-month review, how build work is charged, and the report templates. Meet the deal partners, read the diligence and 100-day plans, run the six questions and pick the first deployment. Write its brief. Resist the urge to meet every CEO. The second month is the first deployment. A one-week diagnostic, then build. Check DPIAs under UK GDPR, and EU AI Act exposure if the company sells into the EU. Stop pilots that won't pay back. The third month is proof. The first workflow is live, measured and handed to a named owner. The first report goes to the deal partner, the first quarterly summary to the IC, and the next deployment is briefed. By day 90 the fund has seen something working, which is more than most portfolio-wide programmes show in a year. ## Red flags when hiring or retaining one Tool evangelism is the most common: every problem ends with the same product. Sometimes the answer is a spreadsheet and a better process. Vendor commissions are the most corrosive. Referral fees are legal, common and bend advice. Ask directly and make disclosure a term of the engagement. Decks without deliveries are the easiest to spot. After two quarters, count what's live. If it's mostly strategy documents and "phase two", the role has become the Big Four engagement it replaced. Spreading thin sounds diligent. A monthly check-in with every CEO turns the retainer into an hour per company, and nothing ships. Quieter ones: reports that measure activity, and no baseline, so every number is a claim. Review the role at six and twelve months against the scorecard. The simplest test I know is the house one: what is still running, unattended, six months later? A good AI operating partner will hand you a list. Everyone else will hand you a deck. If you're working out what this role should look like at your fund, here's how I work as a [fractional AI operating partner](https://anywayventures.com/ai-operating-partner) , or book a [free 30-minute call](https://anywayventures.com/contact) . ## Questions ### What does an AI operating partner do in private equity? An AI operating partner turns AI from a line in the investment memo into results in the P&L. In the model that suits most mid-market funds, the fund retains them and the deal team deploys them into whichever target or portfolio company has the strongest AI value case, such as a deal in exclusivity, a new acquisition's 100-day plan or a business nearing exit. Each deployment is hands on, measured against a baseline and handed over. ### How is an AI operating partner different from a technology operating partner? A technology operating partner usually covers infrastructure, ERP, cyber security, IT cost and carve-outs. An AI operating partner starts from how work gets done in each business, finds the hours and revenue AI can return, and gets systems built and running in production. Some people can do both, but most are much stronger at one, so funds should hire for the job they actually need. ### Should a fund hire a full-time or fractional AI operating partner? Funds with more than about 20 active portfolio companies or a heavy software weighting can usually support a full-time AI operating partner and a small team. Funds with 6 to 15 companies and a small operating bench are generally better served by a fractional operating partner retained by the fund, with a set number of senior days each month that the deal team deploys into one priority at a time rather than spreading them across the portfolio. ### How should a fund measure an AI operating partner in the first year? Measure deployments and outcomes, not activity. Useful year one measures are a prioritisation agreed with the deal team in the first month, a signed brief with a baseline and exit criteria for every deployment, workflows live in production, hours returned per month in each business, annualised EBITDA effect net of run costs, and what is still running unattended six months after handover. Brannan Coady Founder of Anyway Ventures. Exited founder, former CEO of YourParkingSpace, investor and board member; now building AI inside PE-backed and mid-market businesses. [More →](https://anywayventures.com/brannan-coady) ## Keep reading [ALL WRITING →](https://anywayventures.com/insights) [OPERATING PARTNER · 10 MIN How to build an AI value creation plan for a portfolio company A working method, step by step: where the hours go, what to rank first, how to size it in pounds, and how to report it so the exit story is evidenced rather than asserted.](https://anywayventures.com/insights/ai-value-creation-plan) [OPERATING PARTNER · 9 MIN Why mid-market funds can't hire an AI operating partner The people who can do the job are scarce, expensive, and mostly don't want a full-time seat at a £300m fund. There's a better structure, and it costs far less per portfolio company.](https://anywayventures.com/insights/why-mid-market-funds-cant-hire-an-ai-operating-partner) [TRAINING · 9 MIN How to train staff on AI so the work actually changes Most AI training is a lecture or a prompt session, and on Monday the work looks the same. Here is the programme I would run instead: small cohorts, real tasks, a working build per person and a six-week check.](https://anywayventures.com/insights/how-to-train-staff-on-ai)