AI puts $4.7 trillion of profits in play, and markets are already repricing software.
A $4.7 trillion profit shift gives portfolios a triage list.
Bain published a sector-by-sector estimate, covering 92 sectors, of how far AI will move corporate profits. It puts $4.7 trillion of global business profits at stake between now and 2035, and says 71% of sectors face structural transformation, against 41% for the internet ([1]). For a sponsor, the split matters more than the headline. Productivity gains, the line most value creation plans are built on, are worth about $1.1 trillion, just 24% of the total shift ([2]). The rest comes from new AI-enabled products and from $1.3 trillion of existing profit moving between competitors ([3]).
That changes how the AI line in a value creation plan should be sized. A plan that books only cost-out captures the smallest of the three pools and does nothing to defend against the largest. Bain groups sectors into four clusters. In parts of the cluster it calls Rewired, worth $1.5 trillion, it sees no predetermined winners, naming enterprise software, advertising, consulting, corporate law, freight and logistics and several others ([4]). In the smallest cluster, worth $0.3 trillion, customer support, IT services and online tutoring face replacement of their entire delivery model, and the profit pool migrates to whoever owns the AI layer ([5]).
Bain also puts a clock on it: the gap between fast and slow adopters opens within years, not decades ([6]). For most current holdings, that gap opens before exit. Lead author Dunigan O'Keeffe said that the company two years behind cannot buy its way back ([7]).
The practical step for operating partners this quarter is a portfolio triage. Place each holding in a cluster, then ask whether its value creation plan targets the right pool. Services businesses exposed to delivery-model replacement need a clear answer on who will own the AI layer in their market before the next refinancing or sale process starts.
Software comps now reprice with every model release.
On 8 September the S&P 500 software and services index fell 1.4%, its second straight daily fall. Salesforce and Intuit fell about 4%, while ServiceNow lost 5% ([1]). Reuters tied the move to OpenAI's new GPT-6 Astra model and concerns that AI could displace specialised software services. One portfolio manager told Reuters that Astra had reignited the software disruption fears ([2]). In the same session Intel jumped 9% after an Amazon deal for custom AI chips ([3]). The market rewarded the infrastructure layer and marked down applications, the same split Bain describes.
For sponsors, this is more than a two-day move in large-cap names. Public software comps feed quarterly marks, lender appetite and the multiples buyers will anchor on in a sale process. Each frontier model release now acts as a repricing event for sponsor-owned software, whatever the asset's own trading.
Bain's view of enterprise software gives a test buyers will use. A market leader with deep customer integrations, high switching costs and defensible data moats starts with a structural advantage, but only if it treats AI as a board-level priority and ships features customers will pay for ([4]). The key phrase is ships features customers will pay for. A roadmap is not evidence.
Value creation leads should build the exit evidence pack now rather than defend against the same questions in diligence. That means revenue attributable to AI features, renewal and expansion rates on accounts using them, pricing that holds as seats fall, and usage data that shows the product is embedded in customer workflows. Assets that can show this will hold their multiple better when the next model release hits the comps.
Software lenders are paying more and watching their loan books.
Blue Owl Technology Finance Corp, a technology-focused business development company, raised $150m of senior unsecured notes. The notes carry a 7.6% coupon and mature in September 2032, the vehicle's highest coupon on a bond issuance since September 2023 ([1]). The report notes that investors have become more cautious about technology-focused private credit, citing lending standards and the potential for AI to disrupt the software businesses that form a significant part of private credit portfolios ([2]).
The credit picture is mixed rather than broken. Management says its borrowers have so far shown limited evidence of material AI-related disruption, yet the shares remain around 21% lower for the year and non-accruals rose to 0.6% of the portfolio at cost in the second quarter, from 0.3% ([3]). Liquidity is also tight elsewhere in the asset class. Cliffwater's $31bn Corporate Lending Fund again capped quarterly redemptions at 5% after investors asked to withdraw around 16% of their holdings ([4]).
For portfolio CFOs, the signal is about the cost and terms of debt. When a lender's own funding costs rise and its investors question software exposure, that caution reaches borrowers at the next refinancing, repricing or add-on facility. Software platforms that refinance in 2027 should expect lenders to ask the same AI questions equity buyers ask.
A short AI exposure memo is worth preparing before those conversations begin. It should show the share of revenue tied to seats that AI could reduce, renewal cohorts since 2024, attach rates on AI products and any evidence that customers are consolidating vendors. Lenders that can see this evidence early are better placed to price the risk accurately, and a well-evidenced borrower is likely to get better terms than one that leaves the lender to assume the worst.
OpenAI builds an analyst for deal work, and a rival for some portfolios.
OpenAI launched ChatGPT for Financial Services on 10 September. Its early work with Morgan Stanley and Evercore has steered where it started: investment banking and equity research ([1]). The product includes datasets from providers like Daloopa, PitchBook, LSEG News and Crunchbase covering earnings transcripts, financial statements, company fundamentals and private companies ([2]). It is available to eligible financial institutions ([3]).
Private equity is not named as a launch audience, but the work in scope is the same first-pass work that absorbs sponsor associates in origination and early diligence: pulling company and market data, building comparable sets, and tracing every figure back to a source. Private company data from PitchBook built into the tool is what buy-side screening runs on.
Two consequences follow for sponsors. First, the pace of sale processes is likely to rise. If sell-side banks produce materials and answers faster, buyers need an equally fast way to test a target's claims, including its AI claims, or they will either drop out or overpay. Sponsors that have not yet standardised how they test AI capability in diligence should do so before the next competitive auction.
Second, the same product category is a disruption risk inside portfolios. Businesses that sell research, data compilation or analyst hours to financial institutions now compete with a general tool backed by a frontier lab and premium data. These assets sit close to the delivery-model risk Bain describes for services. Owners should check whether their value creation plan assumes pricing that a client with this tool would still accept.
Cheaper analysis also raises the bar for evidence. A polished comp set produced in minutes is not proof of value. Investment committees will need to ask where each figure came from.
UK policy backs AI adoption as the mid-market keeps buying people-heavy services.
Speaking at the UK Private Capital Summit on 10 September, Economic Secretary to the Treasury Lucy Rigby told the industry that AI is one of the most powerful drivers of future productivity, and restated the government's ambition for the UK to be the fastest adopter of AI in the G7 ([1]). She pointed to the AI Adoption Plan for financial services published in July, which she said focuses on supporting innovation while maintaining high standards of consumer protection, operational resilience and trust ([2]).
For UK sponsors and their portfolio companies, the signal is that policy is pushing adoption rather than adding new constraints this year. The regulatory question for most mid-market assets is less about permission and more about governance: whether a portfolio company can show who approved an AI use, what data it touches and how its outputs are checked. That evidence also helps at exit, when a buyer's diligence team will ask for it.
The same week, ECI Partners backed Shaw Gibbs, a UK accountancy, tax and advisory group, as Apiary Capital exited. Since Apiary invested in 2022, Shaw Gibbs has completed 14 acquisitions and grown its workforce from around 100 employees to more than 800 ([3]). The deal shows UK mid-market sponsors are still paying for people-heavy professional services platforms.
Bain lists consulting and corporate law among the sectors where AI leaves no predetermined winners ([4]). Accountancy shares their labour model. For the next owner of a roll-up like this, the exit story will rest on whether AI raises fee income per head and margin across the integrated firms, or whether clients use the same tools to push fees down. The value creation plan for these assets should show which of the two it is planning for, and how it will measure it.