Every hour AI saves your delivery team is, eventually, an hour a client stops paying for. If your firm still prices its work by the hour, AI is not quietly helping your margins — it is quietly shrinking the number you bill against. This is not a future risk. For firms on time-and-material (T&M) contracts, it is already happening, one efficient sprint at a time.

The firms that escape this pattern are not working fewer hours to compensate. They are changing what they sell — from execution time to outcomes. Not “we staffed your project,” but “we solved your problem, faster, because of how we’ve built our team around AI.” That shift sounds like a pricing decision. It is actually an organisational one, and it is the single most consequential business decision AI adoption forces on a mid-market IT firm. 

Why This Is a People Problem, Not a Pricing Problem

Consider what happens inside a T&M-billed team once AI tools genuinely work. A developer who used to need ten hours to complete a task now needs three, using an AI coding assistant to handle the boilerplate. Under hourly billing, the firm now invoices for three hours instead of ten — a 70% revenue drop on that piece of work, even though the output quality is identical or better.

The developer notices this quickly. Being efficient with AI means reducing the firm’s income from their work. There is no policy memo that teaches this lesson — it is simply what the invoice shows at the end of the month. The rational response, for a perfectly good employee who wants to keep their team’s numbers looking healthy, is to slow down, avoid mentioning the time saved, or quietly stop using the tools that make the difference visible.

This is why AI adoption efforts that focus only on training and tool access so often stall in mid-sized IT firms. The training was not the problem. The billing model was actively working against the very behaviour the training was trying to encourage.

What the Research Says

Two separate bodies of research point to the same underlying pattern, from different directions.

Harvard Kennedy School and Mayo Clinic research on human-AI collaboration found that mixed human-AI teams reliably outperform either humans or AI working alone — but only when the collaboration is deliberately designed into how the work happens, not simply switched on as an available tool.

Handing a team an AI assistant and leaving the surrounding workflow, incentives, and role definitions untouched does not produce the performance gain the technology is capable of.

NASSCOM’s 2025 Tech SME Report reaches a compatible conclusion from the commercial side: 75% of Indian tech SME leaders cite cost and talent constraints as their biggest barrier to scaling AI, and more broadly, 94% of tech SMEs believe AI is essential to their business while only 36% have an actual strategic plan for how to use it. Buying the tool was never the hard part.

Redesigning the work — and the commercial model that sits underneath it — is where most firms are stuck. 

A Kerala Mid-Market Example 

Picture two IT services firms in Kochi, each with roughly 200 employees, each delivering comparable web and application development work for similar clients.

Firm A continues to bill by the hour. Its delivery teams adopt AI coding and testing tools enthusiastically at first — until team leads notice project hours (and therefore revenue) falling on every engagement where AI is used heavily. Within two quarters, AI tool usage on billable client work quietly drops, even though the same developers use the tools freely on internal projects, where there is no invoice attached to their output.

Firm B moves its highest-volume, most predictable service line — say, a recurring application maintenance and support retainer — onto a fixed monthly fee tied to defined outcomes: uptime, resolution time, and an agreed release cadence, rather than logged hours. AI-driven efficiency gains on that retainer now improve Firm B’s margin directly, because the fee does not shrink when the work takes less time. Delivery speed becomes something the team is rewarded for, not penalised for.

Same technology. Same starting point. Very different outcomes — because one firm’s commercial model was designed to convert AI efficiency into recognised value, and the other firm’s was not.

Where to Start, If Full Repricing Feels Too Big a Step

Outcome-based pricing across an entire client book is genuinely harder than hourly billing, which is exactly why most firms have not made the shift. It rarely needs to happen everywhere at once. The more workable path is to identify one recurring, well-understood piece of work — a support SLA, a maintenance retainer, a defined and repeatable feature category — and move that single line to outcome-based pricing first. The less predictable, more custom work can stay on T&M until there is enough delivery data from the outcome-priced work to reprice it with confidence.

This also tends to land better with clients than a wholesale pricing change. Clients resist outcome-based pricing when it feels like the same team, doing the same work, under a relabelled invoice. It is received very differently when the client can see that the underlying delivery model was actually redesigned around AI — faster turnaround, tighter scope definition, clearer commitments — not just repackaged.

The Real Question

For Kerala’s mid-market IT firms navigating the AI transition right now, the more useful question is rarely “which AI tools should we buy next.” It is this:

If your firm stopped selling hours tomorrow, what would you sell instead?

Firms that can answer that question specifically — not in general terms, but for one real, nameable piece of their own service line — are the ones positioned to convert AI adoption into margin, not just activity. Firms that cannot answer it yet are the ones most likely to see their best people quietly slow down the moment AI actually starts working.

Frequently Asked Questions

• Does AI-driven efficiency always reduce revenue under hourly billing?

Yes, mechanically — if a task takes fewer hours and the firm bills by the hour, the invoice for that task falls, regardless of how much value the client received. This is a structural effect of the billing model, not a flaw in the AI tool or the team using it.

• Is outcome-based pricing realistic for a mid-sized IT services firm?

It is realistic as a phased move rather than an overnight switch. Firms typically start with one recurring, predictable service line — a maintenance retainer or a defined feature category — before extending the model to less predictable engagements.

• Why do employees hide AI efficiency gains instead of reporting them? 

 Under a billing model where faster delivery reduces the hours (and therefore revenue) attributed to their work, slowing down or under-reporting time saved is a rational individual response — not a sign of resistance to AI itself. The incentive, not the attitude, is usually the root cause.

• Is this only a concern for firms serving international clients?

No. The mechanism applies to any engagement billed on a time-and-material basis, domestic or international. Kerala IT firms serving Kochi-, Kozhikode-, or Thiruvananthapuram-based clients on hourly contracts face the identical structural pressure as those serving US or European accounts.

If this pattern sounds familiar in your own firm, a FREE 45-minute Organisational Health Conversation is a low-commitment way to talk it through.