Industrializing AI Across the Portfolio

For private equity, the next phase of AI adoption is unlikely to be defined by how many employees have access to ChatGPT, Claude or Copilot. The more consequential question is whether sponsors can convert thousands of individual experiments into a systematic portfolio capability—one with shared tools, reusable workflows, training, governance and measurable economics.

The distinction matters because AI usage and AI value creation are not the same thing. Employees can save time drafting emails or summarizing documents without materially changing a company's cost structure, commercial performance or growth trajectory. The emerging challenge for PE firms is therefore to move beyond democratizing AI access and begin industrializing deployment.

Our data suggests there is still a substantial gap between those two stages.

In our proprietary survey of 134 industry professionals, 32.8% said they had yet to see clear efficiency gains from AI, making it the single largest response. Among PE sponsors, that figure reached 50%, while two-thirds of consultants reported no clear gains. The remainder of respondents were relatively evenly divided between faster deal evaluation, savings on manual processes and lower back-office costs.

That distribution is telling. AI is producing benefits, but they remain fragmented. There is no dominant workflow where the industry has collectively cracked the code. Instead, organizations appear to be accumulating pockets of productivity without necessarily translating those gains into an enterprise-wide operating model.

From AI Users to AI Infrastructure

Clearhaven Partners founder Michelle Noon offers a useful model for what the next stage could look like. In discussing how her firm is responding to AI, Noon has described building an internal Claude skills marketplace and thinking about functions such as sales from a blank sheet rather than simply retrofitting existing processes with AI.

The idea of a skills marketplace is particularly relevant for PE because it transforms AI knowledge from something owned by individuals into something owned by the institution.

Consider a portfolio company that develops an effective AI workflow for prospect research, qualification, meeting preparation, call transcription and CRM updates. Under the traditional model, that capability largely stays inside that business. Another portfolio company starts the same journey from scratch.

An industrialized model works differently. The sponsor captures the reusable pieces—the workflow architecture, approved models, prompts, integrations, controls and KPIs—and makes them available across the portfolio. Company B starts from version two. Company C starts from version three.

Every successful implementation reduces the cost and time required for the next one.

That is where portfolio ownership potentially becomes an advantage.

McKinsey describes a similar emerging model. After studying 471 PE-backed businesses across 31 industries, it argues that many sponsors will benefit from a hybrid architecture: common infrastructure, capabilities and experts can be centralized at the fund level, while company-specific teams handle applications requiring proprietary data, product integration or deeper business-model transformation.

The objective is not to force every company onto an identical AI stack. It is to standardize what can be standardized and localize what creates competitive differentiation.

AI Fits Directly Into the Value Creation Playbook

This shift also matters because AI touches several of the levers investors already consider most important.

In another proprietary survey we conducted, operational efficiency and cost reduction and pricing and commercial strategy each captured 27% of responses, while 23% selected digital transformation and technology enablement. Talent and leadership accounted for the remaining 22%.

Industrialized AI sits at the intersection of virtually all four.

AI can automate back-office workflows, but it can also increase sales capacity, improve pricing decisions, change customer service economics and augment management teams. The opportunity therefore extends beyond simply cutting headcount. The more important evolution is from task productivity to operating-model redesign.

McKinsey's research divides portfolio companies into four AI maturity levels: opportunistic adoption, operating-model enhancement, product transformation and AI-enabled business building. The differences in observed valuation are substantial.

Companies at Level 1—where AI consists mainly of isolated experiments and productivity tools—had a median revenue multiple of approximately 13x. Level 2 businesses, which embed AI into operating workflows, reached 14x. But the larger step occurred when AI moved closer to the customer: Level 3 companies embedding AI into products averaged 20x, while Level 4 companies building AI-enabled businesses or new revenue streams reached 31x.

These figures should not be interpreted as evidence that AI itself causes higher valuations; companies with stronger technology, data and growth profiles may also be better positioned to reach higher maturity levels. But the pattern is nevertheless important. The observed valuation gap between Levels 1 and 2 is relatively modest. The larger differentiation appears once AI begins changing what the company sells, rather than only how efficiently employees perform existing tasks. McKinsey itself highlights this distinction.

For sponsors, that suggests productivity is the starting point—not the destination.

Industrialization Requires More Than Tools

Getting there requires a portfolio architecture built around several common capabilities.

First is a shared technology layer: approved models, enterprise licenses, security standards and connectors to common systems such as CRM, ERP and knowledge repositories.

Second is role-specific training. Generic prompt-engineering workshops may encourage experimentation, but they rarely redesign a business. Training should instead show salespeople, finance teams, developers or customer-service representatives how AI changes a specific workflow.

Third is governance. Questions around sensitive data, model access, autonomous agents, human review and customer-facing outputs should not have to be solved independently by every portfolio company.

And fourth is deployment discipline. AI projects need business owners, implementation timelines and measurable objectives. Otherwise, the portfolio accumulates pilots rather than capabilities.

Measure Economics, Not Prompts

The final element is measurement.

McKinsey's data shows another important relationship. At the median, Level 2 companies generated approximately $118,000 of annual revenue per employee versus $99,000 for Level 1 companies—a 19% difference. The gap appears across the broader distribution as well, including $70,000 versus $43,000 at the 25th percentile and $203,000 versus $183,000 at the 75th percentile.

That is closer to how sponsors should ultimately judge AI.

The KPI cannot simply be licenses activated, prompts submitted or even hours saved. It should become revenue per employee, cost per transaction, sales capacity, customer retention, working-capital improvement, margin expansion and ultimately EBITDA growth.

That transition—from tracking adoption to measuring economics—is what separates experimentation from industrialization.

AI's portfolio opportunity is therefore not principally about giving every employee a better assistant. It is about creating a repeatable system for identifying what works, capturing the knowledge, deploying it elsewhere and improving it with every implementation.

The winners may not be the firms with the most AI pilots. They may be the firms that learn how to make each portfolio company's AI progress compound across the rest of the fund.