Buyers no longer give credit for exposure to AI. They want to see whether this particular company can prove AI is moving its margin.
AI raises the stakes on the operating leverage these businesses have long promised. The ones that can show it in their own numbers will have an exit story. The rest will have a list of tools.
Tech-enabled services businesses have long promised operating leverage: technology that lets the same people serve more clients at lower cost. AI raises the stakes on that promise. Buyers no longer give credit for exposure to AI as a sector theme. McKinsey’s 2026 private markets report finds that underwriting has shifted to the individual asset: whether this particular company can use AI to protect its margins or accelerate growth.
For a sponsor or a CEO, that turns a technology question into a measurement question.
In the third quarter of 2026, the ISG Index recorded a record $52.5 billion in large technology and business services contracts, driven by cloud infrastructure for AI. Traditional managed services grew only about 2 percent, IT outsourcing was flat, and ISG pointed to AI-related pricing pressure on providers.
Bain’s research on IT services firms puts numbers on the risk. Firms that carry on as usual could lose up to 30 percent of revenue, and discounting to hold onto work could cost several points of margin. Firms that rebuild their processes with AI at the core, by contrast, could add 200 to 300 basis points of margin. PwC reaches a similar conclusion for industrial and business services: the strongest opportunities lie where AI improves productivity and strengthens margins, not simply where it cuts headcount.
Between a tool and a better margin sit four links.
The research shows why each link matters. A study of generative AI adoption by Bick, Blandin, and Deming found average time savings of about 5 percent among workers who used it, but only about 1 percent across all workers, because many did not use it at all. And a widely cited study of customer-support agents by Brynjolfsson, Li, and Raymond found a 15 percent productivity gain from one AI tool that guided agents through calls, with the largest gains among less experienced staff. That is strong evidence for that kind of tool in that setting, not a general rule about AI.
Tools that suggest next steps and check for errors, much as a strong manager would. They tend to pay off first where management is thin or turnover is high.
Automating routine tasks so skilled people can spend their time on harder work.
AI agents running a whole process end to end, with people handling exceptions. Larger gains, but it needs documented processes and clean data.
Many PE-backed services businesses are in transition after an acquisition, and they generally do better starting with guidance and selective offloading while they document processes and build the data foundation that redesign requires.
The gains also depend on the whole process, not a single step. Making one step five times faster does little if that step is a small share of the work. As one bottleneck clears, the next one, often people, management, or governance, becomes the constraint.
Once AI lowers the cost of delivering a service, pricing becomes the deciding factor. If a business bills by the hour or the head and clients expect the savings, the efficiency gain flows to clients rather than to margin. Businesses that move toward pricing by unit, outcome, or subscription before clients push will keep more of the gain.
Results will be mixed. Many AI investments will level off well short of the business impact they promised, because that impact does not happen automatically; it has to be planned for and managed. The firms that manage AI deliberately, choosing where to deploy it, tracking each link to margin, and redesigning work and pricing around it, will be able to lower prices while holding or even raising margins. Competitors that cannot match that combination will struggle to keep up.
Measurement is only the start. The analytics that pay off are the ones tied to actions the business can take. For a tech-enabled services business, those include:
Identify the teams, sites, and roles that lag in adoption and why, and target training or workflow changes where they will close the gap.
Track where freed hours go, and redeploy them deliberately: to backlog, to new clients, or to lower overtime and contractor spend, with staffing plans adjusted to match.
Monitor error and rework rates after each rollout, and route exceptions to the people best placed to resolve them.
Flag the contracts where cost to serve has fallen most, and prepare unit- or outcome-based pricing before renewal.
Usage logs sit in the tool, time and labor data in workforce systems, delivery and quality data in operational systems, and revenue and cost data in finance, each with its own definitions. The first step is a governed data foundation: a semantic layer and a shared model of clients, contracts, work, and labor that connects them.
The second step is economics. That means separating the effect of AI from changes in volume, mix, and staffing; estimating how much freed capacity actually turned into margin; and modeling how a change in pricing would affect margin by contract. A dashboard is only the last step, the place where that work becomes visible to the people who act on it.
The same foundation also changes who can do the analysis. With clear data governance in place, defining who owns each data domain, who stewards its definitions and quality, and who maintains the systems as custodian, operations and finance leaders no longer have to queue every question behind IT. A well-built semantic layer, AI tools that work against it, and the right training let the business build the analytics it needs, while IT concentrates on keeping systems reliable and available.
We have done this kind of work before, including analytics and hands-on work with operators that lifted labor productivity 15 percent across about 100 facilities.
The services businesses that sell well in this market will not be the ones with the most AI tools. They will be the ones that can show, in their own data, that the tools changed what it costs to serve a client, and that pricing kept the gain.
A short conversation is enough to see which of these measures your data can support today, and where the foundation needs work.
An independent read on whether a portfolio company’s AI story will hold up with buyers.
Compare NotesA governed view of adoption, capacity, and cost to serve, and the economics to price for it.
Map the Leverage