AI Across the Portfolio: Should Operating Partners Build a Team, Hire a Consultant or Run a Program?

1. The Short Answer: One Owner, Outside Help, Vendors per Function

Sponsors that get real value from AI across a portfolio tend to end up with the same shape. One person at the firm owns it. Outside help runs the first wave of companies. Software vendors handle specific functions, one at a time.

What changes from firm to firm is the mix. A sponsor with 40 software companies can justify its own AI team. A mid-market fund with eight industrial and services businesses usually cannot, and a small portfolio may only need a part-time owner and a standard playbook.

So start with the jobs you need covered and the number of companies, and let those pick the provider. This guide covers the six resources operating partners use, what each costs, how fast each moves and where each falls short.

2. Why Portfolio AI Stalls: Activity Without Value

Almost every portfolio company is doing something with AI. Far fewer can show it in EBITDA.

Bain calls this the AI Value Paradox. In a September 2026 article, its private equity partners wrote that most portfolio companies show little correlation between AI spend and value, and named the usual causes: too many use cases chasing too few resources, small productivity gains that never reach enterprise value, tools picked before workflows are redesigned, gaps in data and talent, and no visible owner at the top.

Other research points the same way. Accenture's analysis of hundreds of AI use cases at nearly 40 client portfolio companies found that nearly 90 percent never moved beyond the pilot stage (Accenture, November 2025). In Accordion's 2026 survey of 150 AI, data and technology operating partners, 41 percent said their firms were deploying AI across multiple companies without an operational playbook (Accordion).

The shortage is focus and ownership more than tools. That is why the choice of help matters. The right resource brings a way to pick a few use cases per company, get them into production and keep score across the book.

3. What a Portfolio AI Program Has to Do

Before comparing providers, list the jobs. A working program covers six:

• Prioritize. Find the few opportunities in each company's value-creation plan where AI moves a number that matters. Bain's advice is to isolate three to five.
• Set the rules. An acceptable-use and data-handling policy at each company before tools spread, so staff stop pasting customer data into personal accounts.
• Train. Hands-on training for the function that pays first, on the company's own work.
• Implement. Buy or build the tools for each priority workflow and get them into production.
• Measure. One scorecard across the portfolio, so the investment committee compares like with like.
• Sustain. Keep policies, tools and training current as models and vendors change.

No single provider type does all six well. Strategy firms are strongest at the first job and vendors at the fourth, while training is often the job nobody owns. Hold each option below against this list.

4. AI Advisory Resources for Operating Partners at a Glance

Six resources, compared on the columns that decide most choices. Costs are typical market ranges, and real quotes move with scope.

Most firms combine two or three rows. The sections below take each in turn.

5. Building an Operating-Partner AI Team

The largest sponsors have built their own. Bain's 2025 Global Private Equity Report profiled how three sponsors organize for AI, and two of them show the range. Vista built what Bain called an internal army of professionals to help its 85-plus portfolio companies apply AI, and required each company to submit goals and quantified benefits from its generative AI work as part of annual planning. Apollo set up a center of excellence staffed with two partners and an advisory board of external AI experts, which in turn built an ecosystem of AI specialists, technology partners and service providers.

Notice that the second model is a hybrid. Even a center of excellence at one of the largest sponsors leans on outside specialists. Building in-house settles who owns the program, and the doing can still be bought.

The case for building is continuity: an owner who knows the companies, stays for the whole hold and carries what one company learns into the next. The case against is cost and hiring. FTI Consulting's 2026 Private Equity AI Radar, a survey of 200 fund and operating leaders, found talent was the primary constraint on scaling AI, cited by 35 percent. A senior AI leader plus a small team is typically a seven-figure annual commitment before it ships anything.

Best fit: large or sector-concentrated portfolios, long holds, and theses where AI decides the outcome. The guide to who should own AI covers when a firm needs a head of AI at all.

6. Strategy Firms: Bain, McKinsey and BCG

The three big strategy firms each sell AI work through a dedicated unit: Bain through its AI, Insights, and Solutions practice, McKinsey through QuantumBlack, AI by McKinsey, and BCG through BCG X, its tech build and design unit.

Their strength is the big bet. When a company's value-creation plan depends on AI changing its pricing, product or sales motion, and the equity story at exit rests on it, a strategy team can size the opportunity and build the case that a board and the next buyer will read. They also bring patterns from many sponsors and sectors.

Best fit: the largest companies in the portfolio, and the few AI decisions big enough to move the exit multiple.

Limits: price and staying power. Engagements commonly run six to seven figures, which fits a large company and rarely fits a smaller one's budget. And the team rolls off when the engagement ends, so someone inside has to own what it leaves behind.

7. The Big 4 and Accenture

The Big 4 and Accenture tend to come at portfolio AI from the systems side: finance platforms, data, ERP moves, integration after an acquisition and carve-outs. Deloitte, for example, markets AI services for private equity across diligence, the hold period and exit, and Accenture runs a dedicated private equity group.

Accenture's own research makes the case for foundations. The same analysis that found nearly 90 percent of portfolio-company use cases stuck in pilot found that companies with a sound governance structure and a clear data strategy were better placed to scale high-impact use cases and track return on investment. When data is the bottleneck, a firm that can fix the data earns its fee.

Best fit: large implementations, data foundations, and integration in buy-and-build platforms. The roll-up playbook covers where AI fits in integration.

Limits: these teams are built for scale, and pricing follows. A five-person finance function does not need a systems integrator. If the same firm audits the company or the fund, independence rules can also limit the non-audit work it may take on, so check before you scope.

8. Software Vendors, Function by Function

For a known problem in one function, the fastest route is often a proven product: revenue intelligence for a sales team, AI-assisted close and reporting for finance, AI in the contact center for customer service. Vendors run their own onboarding, and many include implementation help in the subscription or sell it as a services add-on.

The limit is scope. Each vendor sees one function at one company. Nobody in that picture owns the portfolio view, sets the rules across companies or decides which function goes first. Bain's list of failure modes includes exactly this: picking a copilot, or answering a vendor pitch, before redesigning the workflow.

Best fit: the moment you know which function pays first. The function guides compare the options for revenue intelligence, the finance back office and customer service.

9. Specialist Programs and Fractional AI Operating Partners

Between the large firms and doing it yourself sits a category of smaller specialist firms that sell portfolio AI as a repeatable program. Each company typically gets some mix of a diagnostic, a usage policy, training and a standard readout, usually at a fixed fee per company, with a few companies running at a time.

The appeal for a mid-market sponsor is one standard. Ten companies get the same diagnostic and the same scorecard, so the investment committee compares like with like, and the operating partner has one number per company to budget. The limits are bench depth and scope. These are small teams, so ask who does the work, and they are not built for an ERP rebuild or for putting AI inside a software company's product.

The fractional AI operating partner is the retained version: one senior person, part-time, who owns AI across the firm and its companies, runs office hours, reviews adoption and keeps the policies current. It suits sponsors that are not ready to hire a head of AI, and those that ran a first wave and need someone to keep it moving. The limit is capacity. One person can advise and coordinate, and building still needs hands.

Best fit: mid-market portfolios of roughly 3 to 15 companies, and sponsors who want one standard in place before they decide whether to build a team.

10. Firms and Roles That Support AI Implementation in Portfolio Companies

Whatever mix you buy, the same roles have to be filled at each company. Programs stall when one is empty.

• The operating partner sponsors the program, sets one standard across companies and ties each company's AI plan to its value-creation plan.
• A named owner at each company, usually the CFO or COO, owns the plan, the budget and the result. Bain's sponsorship gap is what happens without one.
• Function leads run the work in their own workflows and decide what goes to production.
• IT and security approve tools, set up access and enforce the data rules.
• Vendors' customer success teams handle onboarding and configuration for their own products.
• Outside help runs the diagnostic, the training and the first wave, from whichever resource above fits.

The portfolio AI maturity assessment helps score where each company starts, and training portfolio company teams covers the enablement job in detail.

11. How to Choose by Portfolio Shape

Portfolio shape tells you more than any brand name. Four common cases:

• Fewer than five companies, mid-market. A fractional owner or a specialist program for the first wave. A full team is hard to justify.
• Five to fifteen companies across sectors. One standard program across the book, a light internal owner, and vendors per function once priorities are set.
• A large or concentrated portfolio, such as software. An internal team, with strategy firms brought in for the biggest bets.
• One company with a company-defining bet or a broken data foundation. A strategy firm for the bet, a Big 4 or Accenture team for the foundation.

Then put these questions to any provider:

• Do you start from each company's value-creation plan or from a list of use cases?
• Will every company get the same readout, so we can compare them?
• Who trains the people, and on whose work?
• How do you set data and acceptable-use rules before tools spread?
• What stays behind when you leave, and who maintains it?
• Are you paid by any vendor you might recommend?

Red flags: a pitch that starts with a tool, pilots with no path to production, one policy pasted across every company regardless of sector, claims that a tool stores nothing, and fees you cannot tie to a result in a value-creation plan.

12. Where to Start

Pick two or three companies where the value-creation plan already leans on efficiency, reporting or sales productivity. Name an owner at each. Put the data rules in place before anyone buys a tool.

Run that first wave with whichever resource fits your portfolio shape, and measure each company on one number from its plan. After one wave you will know far more about whether to build a team, keep buying programs or retain a fractional owner.

The 100-day playbook covers how AI fits into a new company's first months, and build, buy or partner covers the capability decision at the firm level.