Firm-Level or Portfolio-Level AI: Where Should Private Equity Firms Start?

When AI lands on the agenda, most PE firms do one of two things: mandate it across every portfolio company at once, or wait for a flagship internal use case before touching the portfolio at all. Both are expensive bets, and the difference between a return and a write-off often comes down to which one they pick.

Organizations with fully integrated AI are nearly four times more likely to report AI-driven revenue growth than those still piloting, 58% versus 15%, according to Grant Thornton's 2026 AI Impact Survey.

However, most of the industry is investing in AI and still waiting on the return.

The real gap comes down to matching the sequence to where value, urgency, and ownership already exist. This post lays out a simple framework for deciding where to begin, and shows why the answer doesn't have to be firm-first or portfolio-first at all.

Firm-level AI and portfolio-level AI solve different problems

Before picking a starting point, it helps to separate what each layer actually does.

Firm-level AI is the investment team's daily work — deal screening, diligence, monitoring, reporting.

Portfolio-level AI runs inside the companies the firm owns, touching sales, finance, operations, and customer service. The breakdown below shows how little overlap there is.

Software specialist Hg has leaned into that separation deliberately. A centralized value-creation team of more than 100 AI experts develops shared playbooks, tools, and benchmarking that portfolio companies draw on, while each company's own AI work stays specific to its business.

The case for building AI capability inside the firm first

Building internal capability first has a clear logic. The firm develops enough hands-on understanding to judge AI opportunities without relying entirely on vendors or portfolio executives to explain what's realistic. Internal wins also create the credibility a firm needs when it later asks portfolio companies to spend their own time and budget on the same thing.

Internal capability means the firm knows enough to lead, evaluate, and set standards before the rest of the portfolio starts asking for guidance. None of that requires every investment professional to become an AI specialist.

AI Operator's case study with Source Capital shows what that looks like in practice. An AI-powered background screen built during a 12-week cohort caught a federal litigation case a standard search missed entirely, a finding that could have cost hundreds of thousands of dollars in diligence spend if it had surfaced later in the process. A second internal tool cut deal-evaluation time by more than half, over seven hours per deal.

The case for starting with a portfolio company first

Waiting for firm-wide readiness can cost real, available value. A portfolio company may already have an urgent, high-value problem, strong data, or a leadership team ready to move now.

Apollo's Shutterfly generated $5 million in new revenue in its first year from a single AI feature, along with a 22% productivity gain on a major software project, according to Bain's 2025 Global Private Equity Report. The same report cites Apollo's Cengage, where a portfolio company running eight separate AI projects saw costs drop 40% in select content production processes and 15 to 20% through automated lead generation.

Inside Source Capital's own portfolio, three individual leaders built tools worth $95,000 or more in combined annualized value, ranging from $15,000 to $50,000 per tool, in a matter of weeks, without waiting on any firm-wide rollout. That's often the faster path to a measurable result, and it's exactly why firm-first can't be treated as a fixed rule.

Why a portfolio-wide AI mandate usually backfires

Portfolio companies vary too much in maturity, data, and leadership for one mandate to work across all of them.

The confidence gap tells the same story from another angle. Just 7% of firms still piloting AI feel confident they could pass a governance audit within 90 days, compared to 74% of firms with AI fully integrated, according to Grant Thornton's 2026 AI Impact Survey. Shallow, broad rollouts create a governance problem on top of a performance one.

The better approach runs both layers in parallel, narrowly: build central capability while piloting in the one or two portfolio companies with the strongest case.

Central capability should develop early. Implementation should begin wherever the strongest business case exists.

Someone at the firm level still needs to assess opportunities, set standards, compare results, and stop fragmented adoption before it starts, whether or not the first pilot happens inside the firm.

Not sure where your firm stands right now? Take the two-minute AI Audit for an instant read on where your firm and portfolio have the strongest starting point.

A framework for choosing where to start

A short checklist works better than a long debate. Assess any potential starting point, inside the firm or inside a specific portfolio company, against six criteria:

• Value. How much upside is realistically on the table?
• Urgency. Is there a real reason to act now, not later?
• Readiness. Are the data and systems actually usable?
• Ownership. Is there an accountable senior owner?
• Measurability. Can ROI be shown, not just claimed?
• Speed to value. Can a first result land within a reasonable timeframe?

Ownership and measurability tend to decide the outcome more than the technology does. A pilot with no named owner rarely survives past the first leadership change, and a result nobody can measure is a result nobody can defend at the next partner meeting.

Start where these line up, not where the AI use case sounds most impressive on a slide.

What Source Capital's approach proves, and what it doesn't

Source Capital built capability inside the firm first, then expanded into portfolio-company leadership, then extended the same program to their peer network. It's a useful example of how a firm-first sequence can create momentum. But it reflects one firm's specific starting point, not a universal rule every PE firm should copy.

In Source Capital's peer network, 82% of participating PE professionals are now exploring AI training for their own portfolios, per AI Operator's ongoing engagement with the firm. That's meaningfully ahead of an industry where most portfolio companies are still starting from zero on AI strategy, which is exactly why Source Capital's sequence is worth learning from, not copying line for line.

Find the strongest starting point before you commit to a sequence

Your portfolio companies are already competing against leaner, AI-enabled businesses, whether or not your firm has picked a starting point yet. The firms pulling ahead aren't the ones with the boldest AI mandate. They're the ones matching their sequence to where value, readiness, and ownership are already strongest, while building central capability alongside it from day one.

Book a Strategy Call with AI Operator to map the strongest starting point for your firm and portfolio.

AI for Private Equity: Frequently Asked Questions

What's the difference between firm-level and portfolio-level AI?

Firm-level AI supports the investment team's own work: deal screening, due diligence, portfolio monitoring, and reporting. Portfolio-level AI improves operations inside the companies the firm owns, like sales, finance, and customer service. They solve different problems and usually need different owners.

Should private equity firms build AI capability at the firm level or the portfolio level first?

Start wherever value, urgency, ownership, and readiness are strongest, whether that's inside the firm or inside a specific portfolio company. Most firms get better results running both in parallel: central capability building alongside a narrow, high-conviction pilot.

Why doesn't a portfolio-wide AI mandate work?

Portfolio companies vary too much in data maturity, leadership readiness, and existing AI strategy for one rollout to fit all of them. A mandate applied evenly across a mixed portfolio tends to create shallow adoption everywhere instead of real results anywhere.

How fast should a PE firm expect to see results from an AI pilot?

Examples worth learning from move in weeks — Source Capital's portfolio leaders built tools generating measurable value in a matter of weeks. If a pilot can't produce a first, defensible result within a quarter, that's a sign the starting point was wrong, not that AI needs more time.

Who should own an AI pilot inside a portfolio company?

A named, accountable senior leader, not a committee. Pilots without a clear owner rarely survive a leadership change, and results nobody can measure are results nobody can defend at the next partner meeting.