An AI slowdown call buys software and services time, and the exits show what buyers will pay for.

Two software processes this week show what buyers will still pay for. Hellman & Friedman agreed to sell a stake of around 10% in Italian accounting, payroll and business software group TeamSystem to Francisco Partners, with a further 5% or so going to other investors including KKR, according to people familiar with the matter ([1]). The deal values TeamSystem at between €8 billion and €10 billion ([2]). Reuters noted that it follows a steep sell-off in software stocks this year, partly driven by concerns about AI, which has threatened to derail transactions across the industry ([3]).

The reason given for the price matters more than the price. One source said TeamSystem's products are closely integrated with government e-invoicing systems used by small and medium-sized companies, making them harder for AI-driven competitors to replicate ([4]). The €8 billion valuation H&F was reported to be seeking in July implied roughly 16.5 to 17 times 2025 adjusted EBITDA of €476 million ([5]). The structure matters too: a so-called private IPO lets H&F return capital at a time when sponsors have struggled to list their largest portfolio companies ([6]). In the same week Silver Lake said it would merge French software companies Cegid and Silae into a group valued at more than €10 billion ([7]).

Vista Equity Partners is exploring options for Finastra, including a sale, with Blackstone among prospective bidders ([8]). Vista is seeking to capitalise on renewed interest in financial software even as sector valuations struggle amid uncertainty over how AI could disrupt business models and profitability ([9]). Finastra expects EBITDA of $650 million this year, and one source said traditional multiples for specialised software would value it as high as $12 billion ([10]).

The common thread is defensibility that a buyer can see. Regulatory plumbing, embedded payments and lending rails, and deep workflow integration are being priced as protection against AI substitutes. Exit-bound software assets should document that case explicitly, with usage and retention data, before diligence starts.

Waystar, the healthcare payments software company taken public in 2024 by EQT, the Canada Pension Plan Investment Board and Bain Capital, is exploring options including a sale that could return it to private hands, according to seven sources ([1]). It has hired Evercore, and Barclays is also advising ([2]). Its shares rose more than 8% after the report, recovering part of a slide that had removed nearly a quarter of its market value this year, leaving it worth about $5.2 billion ([3]). EQT remains the largest shareholder with 13%, followed by CPPIB with 10% ([4]).

The case matters because of how Waystar chose to describe itself. It positioned itself as a healthcare software company that automates administrative work, rather than a healthcare services business that relies on people to do those tasks, in order to win the higher valuations typically given to technology companies ([5]). Investors initially accepted the story and the shares rose from $20 to a 2025 peak of $45, but they came under pressure as investors grew concerned that advances in AI could disrupt software companies, Morgan Stanley analysts said in July ([6]). Reuters said an auction could show whether investor appetite for software businesses is returning ([7]).

For sponsors, Waystar turns the software-or-services label into the central valuation question. The label that earned a technology multiple now invites the question of whether AI makes the product more valuable or makes it easier to replace. Healthcare revenue cycle, claims and administrative automation assets are common in mid-market portfolios, and this process is the comparable to watch.

The practical test for an operating partner is to show where AI is already embedded in the product and what it does to customer outcomes and gross margin. A buyer weighing a take-private in this sector will want proof that the software is doing the work that AI agents might otherwise take on, not just a narrative that says so.

The week brought two clear signals on how AI is changing labour-heavy professional services, and where outside capital is entering. Morgan & Morgan, which describes itself as the largest US personal injury law firm, committed to invest at least $1 billion in AI and technology over the next decade and will start marketing its own AI platform to other law firms ([1]). It has already spent $300 million building the platform, called MX2, which it uses to extract medical information, generate case documents and prepare for trials ([2]). Founder John Morgan argued that firms billing by the hour to draft and review agreements or read thousands of pages of documents are the practices AI will replace ([3]). Reuters reported in June that the firm, with annual revenue of $2.4 billion, had hired JPMorgan to explore a minority stake sale ([4]).

Separately, FairPlay Law launched on 16 September as an AI-native firm focused on employment disputes, affiliated with a management services organisation, FairPlay Global, that is backed by investor capital ([5]). MSOs can be backed by outside capital, unlike law firms, which are largely prohibited from investor ownership, and these AI firms often avoid hourly billing and list flat fees ([6]). A directory kept by legal technology company Lupl now lists at least 60 AI-native law firms ([7]).

For private equity this matters in two directions. On the buy side, the MSO structure is opening a route for capital into legal services, with AI as the operating thesis and flat-fee pricing as the commercial model. On the hold side, any portfolio company that sells professional time, from legal process outsourcing to compliance, consulting or back-office services, faces competitors built without the cost base of hourly staffing.

Diligence should now ask what share of revenue is priced by the hour, how quickly that share is falling, and whether the target can show margin holding as prices move to fixed fees.

Deloitte's first UK GenAI Workforce Survey, based on 25,000 workers, estimates that employees spend £958 million a year of their own money on GenAI tools for work ([1]). Nearly a third (31%) of users say they use GenAI without their employer's knowledge, and almost a quarter of UK workers now use it every day ([2]). Around half of users say they have had no formal training on safe and effective use, and 65% report a lack of convincing leadership on how it should be used ([3]). Workers claim to save 70 minutes a week on average, mostly spent doing more work for the same employer ([4]).

For portfolio CFOs the finding cuts both ways. The productivity gain is real but ungoverned, so it rarely shows up in the P&L or in a value-creation plan. The exposure is also real: client data and confidential material moving through personal accounts is a diligence finding waiting to happen, and a buyer will ask about it. The first step in any portfolio AI programme should be an audit of what staff already use, followed by approved tools, training and a way to measure time saved against cost.

Sponsors are also turning AI inwards. Carlyle joined the MIT Generative AI Impact Consortium on 15 September for a two-year research programme that will draw on its proprietary private markets data ([5]). The work will test how large language models can identify relevant information, generate investment insights and support portfolio construction and capital allocation ([6]). Key research areas include evaluating AI performance on investment-related tasks and extracting investment signals from complex datasets ([7]). Chief executive Harvey Schwartz said Carlyle sees significant potential to use AI to make better-informed investment decisions ([8]).

The gap between these two stories is the point. Sponsors are investing in AI for their own decisions, while many portfolio workforces are adopting it without guidance. Closing that gap is where measurable value sits.