---
title: "AI disruption risk portfolio review: a triage framework"
description: "An AI disruption risk portfolio review for mid-market PE: seven scoring factors, a four-week process for 10 to 30 holdings, and what to tell LPs and boards."
url: https://anywayventures.com/insights/portfolio-ai-disruption-triage
updated: 2026-10-25
---

[← ALL WRITING](https://anywayventures.com/insights) DILIGENCE

# AI disruption risk across a portfolio: a triage framework

A fund with 20 holdings can't run full diligence on each one every time the AI question comes up. It can triage them in a few weeks, sort them into four classes and give LPs and boards an answer they can check.

WRITTENOctober 2026
READING · Twelve minutes
FOR · Operating partners, deal partners, portfolio CEOs and CFOs, investor relations teams
RELATES TO[AI disruption risk review](https://anywayventures.com/ai-disruption-risk)

An AI disruption risk portfolio review scores every holding against the same short set of factors (how much revenue sits on tasks AI now does, how the business prices, whether it sells labour, and what protects it), then sorts each company into one of four classes: defend, adapt, accelerate or exit early. For a mid-market fund with 10 to 30 companies it takes four to six weeks, not a full diligence on each. The businesses most at risk from AI are those selling hours or seats for work that ends in a document, a ticket or a line of code, with nothing proprietary sitting between the customer and a cheaper alternative.

I've written separately about [assessing AI disruption risk in a single target](https://anywayventures.com/insights/ai-disruption-risk-assessment) , with an eight-factor grid and a revenue-at-risk calculation. That's the right tool when you have a data room and a price to set. This article is about the other problem: the fund already owns twenty businesses, an LP has asked "what's your AI exposure?", and the honest answer is that nobody has looked at them all on one basis.

> A portfolio review that puts every company in the middle has told you nothing.

## Why funds are running portfolio AI risk reviews now

The market stopped waiting. [EY's Q1 2026 Private Equity Pulse](https://www.ey.com/en_gl/newsroom/2026/04/pe-navigates-a-more-complex-geopolitical-and-macroeconomic-environment-ey-analysis) found around 60% of GPs had increased diligence on AI disruption risk, and nearly two-thirds were taking a more targeted approach to new investment. Technology fell from roughly 30% of global PE deployment by value in 2025 to 12% in the first quarter of 2026. Bain's [Private Equity Midyear Report 2026](https://www.bain.com/insights/private-equity-midyear-report-2026/) puts the fall in software valuations inside PE portfolios at roughly 8% in Q1 marks, after a near 30% drop in listed software in February.

The largest firms are already doing this across their books. Bloomberg reported in June 2026 that [Apollo is assessing every new software opportunity for AI disruption risk](https://news.bloomberglaw.com/artificial-intelligence/apollo-is-screening-all-software-investments-for-ai-threat-risk) , using a framework that splits software into 12 to 14 categories ranked by susceptibility. Rob Bittencourt, its head of thematic investing, said the firm had also overlaid that framework on its existing portfolio to decide where it might reposition.

Software is only half of it. Livingstone's [2026 IT Services Market Update](https://livingstonepartners.com/wp-content/uploads/2026/08/Livingstone-2026-IT-Services-Market-Update.pdf) describes BPO multiples resetting from around 12 to 13 times EBITDA to around 5 to 7 times as agentic AI erodes transaction-based revenue, and digital engineering falling from above 23 times in December 2024 to around 7.4 times. Its summary is blunt: the market is repricing labour-arbitrage and FTE-billing models, not the sector as a whole.

A mid-market fund can do the same thing with a consistent test and a few weeks of work.

## Which businesses are most at risk from AI?

Exposure concentrates in a recognisable set of business models. In rough order of how hard the market is already pricing it:

Business model
Why it's exposed
What usually protects it

BPO, contact centres, outsourced back office
Revenue is priced per FTE or per transaction for work AI agents increasingly do
Owning the automation, regulated processes, deep integration

IT services and digital engineering
Billed by the day for code, testing and support that AI tools now compress
Outcome pricing, managed services contracts, domain depth

Agencies (content, SEO, creative production, translation)
Output is text, images or copy a model drafts acceptably
Strategy and relationships, performance-based fees

Seat-priced SaaS for knowledge workers
Revenue scales with customers' headcount, which AI may shrink
Systems of record, proprietary data, workflow lock-in

Document-heavy professional services
Hours billed for drafting, review and summarising
Regulated sign-off, liability, long relationships

Physical, licensed, on-site services
Little of the core service can be done by software
Already resilient; the risk is in the back office and in sales channels

That table is where the review starts, not where it ends. A BPO business that runs its own automation platform for regulated insurance claims is a very different asset from one supplying 300 agents to a retailer's customer service queue. Livingstone notes that Capgemini paid 13.5 times EBITDA for WNS on exactly that logic: a platform that owns the automation rather than competing with it. The review exists to tell those two apart.

## A portfolio AI risk triage framework: seven factors

The single-company grid has eight factors and needs evidence from a data room. For a portfolio you want something lighter that can be scored from a standard data request and one management interview, then deepened only where it matters. I use seven factors, each scored 0 to 3, where 3 is most exposed. A fuller model with weights and an Excel version is in the [AI disruption risk scorecard](https://anywayventures.com/resources/ai-disruption-risk-scorecard) .

#
Factor
0 (protected)
3 (exposed)
Scored from

1
Revenue tied to tasks AI automates
Under 10% of revenue
Over 50% of revenue
Revenue by service line, sample deliverables

2
Pricing basis
Outcome, asset, visit or contract pricing
Priced per seat, per hour or per FTE
Sales ledger split by billing basis

3
Labour-arbitrage model
Margin comes from the product or asset
Margin comes from the gap between labour cost and bill rate
Gross margin by line, headcount mix

4
Proprietary data and workflow
Owns structured data and rights to use it
Works on clients' data with generic tools
Data inventory, contract terms

5
Regulation and accountability
Licensed or named sign-off required
Anyone can do it, no liability attaches
Licences, PI insurance, accreditation

6
Switching costs
Integrated, multi-year, painful to unpick
Project work, re-tendered, easy to split
Retention cohorts, contract terms

7
AI-native competition
No credible AI-native entrant in the category
Funded AI-native rivals winning customers now
Competitor scan, lost deal reasons

That gives an exposure score from 0 to 21. Factors 1 to 3 describe the threat. Factors 4 to 6 describe the moat. Factor 7 tells you whether the threat is theoretical or already in the pipeline. I weight factor 7 more heavily in conversation than the arithmetic does: a company scoring 3 there is losing deals this quarter, whatever the total says.

Two more inputs sit outside the score, because they don't measure exposure. They measure what the fund can do about it.

The first is capacity to respond: does management have someone who has shipped AI into operations, cash or covenant headroom to fund a change in pricing or delivery, and a board that will back it? Score it high, medium or low. The second is time to exit. A business with four years left in the hold has time to change its model. A business going to market in 12 months does not.

## How to classify: defend, adapt, accelerate or exit early

The classification comes from putting exposure next to capacity and time. Each company gets one class, a short reason and three actions.

Class
Typical profile
What the fund does

Defend
Exposure 0 to 7; moat factors strong
Protect the moat, take the cost upside in the back office, don't overpay for add-ons on an "AI-proof" story

Adapt
Exposure 8 to 14, or higher with strong capacity and time
Fund a specific change: reprice exposed lines, put AI into delivery, move seats to usage or outcome pricing

Accelerate
Any exposure level, but the company owns data, distribution or the automation itself
Invest more: product, buy-and-build, bring forward capex, consider whether this becomes the platform

Exit early
Exposure 15 or above with low capacity or under two years left
Bring forward a sale while buyers still separate this asset from its sector, or restructure the plan and reset the mark

Accelerate is the class people forget. The same shift that hurts a labour-arbitrage business helps a company sitting on ten years of structured claims data, or one with a distribution channel an AI-native entrant would have to spend years building. Those businesses deserve more capital, not a risk flag. I'd expect two or three in a typical 20-company portfolio, and I'd expect them to be under-funded, because nobody has written down why AI helps them.

Exit early needs care. In some categories the market has already priced the risk, and selling a BPO business into a 5 times market simply crystallises the loss. The case for an early exit is strongest where the asset still trades on old assumptions, often because it sits in a niche buyers don't yet associate with AI, and where the honest plan says the company can't change fast enough. Where that isn't true, the right answer may be to move it to adapt with a smaller, cheaper plan and reset expectations.

A portfolio review that puts every company in the middle has told you nothing. If 18 of 20 holdings come back as adapt, the scoring is being softened somewhere, usually by deal teams scoring their own companies.

## How to run a portfolio AI risk review across 10 to 30 holdings

The process below fits four to six weeks for a portfolio of 10 to 30 companies, with a light ask of each management team.

Week 0, one standard data request. Revenue by service line and by billing basis, headcount by function, gross margin by line, the top 20 customers with tenure, the product or service list, the five competitors management worries about most, and the last 12 months of AI tool spend. It should take a CFO half a day. Send it to every company in the same format.

Weeks 1 and 2, desk score and interviews. Score each company from the data, then hold a 60 to 90 minute interview with the CEO and whoever runs delivery or product. Ask what they've lost to cheaper or AI-led competitors in the last year, which customers have asked for price cuts citing AI, and what is actually running in production. The answers to the first two questions move scores more than anything in the data pack.

Weeks 2 and 3, deep tests on the companies that matter. Pick the five to eight companies scoring highest on exposure, plus any candidate for accelerate. Run the replicability test from the [single-company assessment](https://anywayventures.com/insights/ai-disruption-risk-assessment) on a sample of their deliverables, scan their category for AI-native entrants, and check renewal and pricing trends in the last four quarters.

Week 4, calibration. One session with the operating team and the deal partners, every company on one page, scored side by side. This is where the scores become consistent. Expect arguments. A company everyone agrees is resilient often turns out to have 40% of revenue on hourly billing for drafting work.

Weeks 4 to 6, classification, actions and write-up. Each company gets a class, three actions with owners, and a revenue-at-risk range. The portfolio gets a single view: NAV and EBITDA by class, and the handful of decisions the fund needs to make.

The point of the triage is to decide where to spend the deeper work. Most companies need nothing more than the one-page output and a check at the next board. Two or three will need the full single-company assessment, and possibly a rewritten [AI value creation plan](https://anywayventures.com/insights/ai-value-creation-plan) .

### What the one-page output for each company contains

Keep it to a page. The exposure score with a line of evidence against each factor, the capacity and time inputs, the class and the reason in two sentences, a revenue-at-risk range over the remaining hold, and three actions with an owner and a date. If an investment committee member can't read it in three minutes and say whether they agree, it's too long.

## What to tell LPs about AI disruption risk

LPs are already asking, and they expect the answer to separate managers. The [Coller Capital Global Private Capital Barometer, Summer 2026](https://allnews.ch/sites/default/files/ckfinder/userfiles/Coller%20Capital%20Global%20Private%20Capital%20Barometer%20Summer%202026.pdf) found 67% of LPs expect AI adoption to widen the gap between the best and worst performing funds. An LP who believes that wants to know which side of the gap you're on.

What works, in my experience, is method before results. Describe the test, show that it was applied the same way to every company, then show the portfolio by class as a share of NAV. Name the actions on the exit-early and adapt companies without naming the companies if confidentiality requires it. Say what changed in the marks, and why.

What doesn't work is "we have limited exposure to AI disruption". Every portfolio has some, if only in back offices and in how customers buy. An LP who hears that line from a fund with three IT services businesses will assume nobody looked. I would rather report that four companies sit in adapt, with funded plans and dates, than claim a clean portfolio and be contradicted by the next valuation round.

Be careful with the accelerate story too. LPs have heard a great deal about AI upside, and the [StepStone and Bain 2026 GP outlook](https://www.stocktitan.net/news/STEP/bain-company-and-step-stone-group-release-2026-private-equity-gp-c4gwni4xizvz.html) found nearly 40% of GPs not expecting a material financial impact from AI in their portfolios in 2026. Claim upside only where the plan has a number and an owner against it.

## What to tell portfolio boards

Each board should get its own page and a conversation. Keep the portfolio league table inside the fund. The CEO of a company classed as exit early needs to hear it from the deal partner first, with the reasoning, before anything changes in the board pack.

Then turn the class into standing agenda items. For adapt companies that means the pricing migration and the AI delivery work reported quarterly, with numbers: share of revenue on fixed or outcome pricing, cost per unit of delivery, win rates against named competitors. For defend companies, a shorter check on the moat factors and the back-office savings. The questions in [what boards should ask about AI](https://anywayventures.com/insights/board-questions-about-ai) map directly onto this.

One caution. A classification is a judgement made with incomplete information, and management will sometimes have evidence that changes it. Give them a week to challenge the draft with data, and change the class if they're right. That is far better than a CEO who thinks the fund has written their business off on the basis of a desk review.

## How much does a portfolio AI disruption risk review cost?

Our [AI disruption risk review](https://anywayventures.com/ai-disruption-risk) is priced per company so a fund can scope it to the holdings that matter. Portfolio scans start from £4k per company, covering the data request, desk score, management interview, calibration and the one-page output. A full single-deal review, with the replicability test and revenue-at-risk modelling, runs £8k to £15k depending on the size and complexity of the business.

Most funds don't need the full review on every company. A sensible pattern is the scan across the whole portfolio, then the deeper review on the two to four companies where the class is contested or the decision is large.

## Mistakes that make a portfolio review useless

The most common is letting each deal team score its own holdings. People who championed an investment at IC are the last to score it as exposed. One small team, one standard.

The second is scoring the sector instead of the company. "IT services" isn't a score. An IT services firm with 70% of revenue on multi-year managed services contracts and its own monitoring platform is a different asset from one placing contractors by the day.

The third is stopping at the score. The value is in the class and the three actions. A heatmap without decisions is something to file.

The fourth is treating it as a one-off. Factor 7 can move in a quarter when a well-funded AI-native entrant starts winning in a category. Refresh every six months, and re-score any company starting exit preparation, where the questions buyers ask are covered in [will buyers pay for your AI?](https://anywayventures.com/insights/will-buyers-pay-for-your-ai) .

If you want an outside team to run the triage across your portfolio on one standard, that's what our [AI disruption risk review](https://anywayventures.com/ai-disruption-risk) does, and it starts with a 30-minute call to agree which holdings to include.
[FREE RESOURCE · PDF + XLSX AI disruption risk scorecard DOWNLOAD FREE →](https://anywayventures.com/resources/ai-disruption-risk-scorecard)

## Questions

### How often should a fund repeat a portfolio AI risk review?

Run the full triage once, then refresh it every six months, with a lighter quarterly check on the companies classed as adapt or exit early. The inputs that move fastest are competitor activity and customer behaviour at renewal, so the quarterly check should focus on those. A full re-score makes sense whenever a company starts exit preparation, makes a large acquisition or changes its pricing model.

### Who should own the portfolio AI risk review inside a fund?

The operating partner or head of portfolio, with the deal partner for each company signing off its classification. Investor relations should see the method before anything goes to LPs. Keep the scoring with one small team, internal or external, so the same standard applies to every company. When each deal team scores its own holdings, every company comes back as resilient.

### Should portfolio company management see their AI risk score?

Yes. The score is only useful if it changes what management does, and they will usually have evidence that corrects it. Share the draft with the CEO and CFO, give them a week to challenge it with data, and then fix it. The final classification belongs to the fund. The actions that follow belong to management, with named owners and dates.

### Does AI disruption risk apply to businesses that aren't software companies?

Yes, and some of the sharpest exposure sits outside software. Business process outsourcing, IT services, agencies, contact centres, translation, document-heavy professional services and outsourced back-office work all sell labour that AI can increasingly replace. Physical, licensed and on-site businesses are less exposed, though their back offices, sales channels and customers' buying habits can still change around them.

### How is a portfolio AI risk review different from AI due diligence on a new deal?

Deal diligence goes deep on one company in a short window, with access to the data room and a price to set. A portfolio review goes broad: the same light test across every holding, scored against one standard, to decide where to spend attention and capital. The portfolio triage tells you which two or three companies need the deeper single-company assessment.

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)
[DILIGENCE · 10 MIN AI disruption risk assessment: will AI hollow out the target? AI washing covers targets that overclaim AI. This covers the other problem: targets that AI threatens. An eight-factor grid, the evidence to request, and what the score should do to the deal.](https://anywayventures.com/insights/ai-disruption-risk-assessment) [BOARD · 9 MIN Board questions about AI: what to ask every quarter Boards tend to ignore AI or demand a strategy deck about it. Both are mistakes. A short list of questions, asked every quarter and answered with evidence, does far more.](https://anywayventures.com/insights/board-questions-about-ai) [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)
