If you use AI to do the work faster, your clients will notice. Some of them will ask you to take it off the bill. That is already happening at a much larger scale than a typical privately held firm, and the same logic travels down the chain: once a buyer can see that the work costs less to produce, the price becomes a live question.

Clients are already asking for the savings.

Infosys is a useful public case, not because your firm looks like Infosys, but because its clients made the request in public. In coverage of the company's October 2026 results, The Register reported that Infosys's AI revenue grew 29% from the prior quarter and now accounts for 10.5% of its incoming cash. Overall second-quarter revenue grew 1.5% year on year, and the company cut its full-year growth outlook from 1.5% to 3.0% down to 1.0% to 2.0%.

CFO Jayesh Sanghrajka told investors, "Growth was below our expectations. Volume was softer, and price increases were lower than anticipated amid high competitive intensity." The Register said one reason for those lower price increases was that customers expect Infosys to charge less when it uses AI, and that those cost savings are now "baked in" to its multi-year deals. CEO Salil Parekh pointed to AI as one reason the cost of services is falling, and said pricing pressure was coming both from competitors and "the discussion on AI productivity." On large programs, he said, clients are talking in a three- or five-year outlook.

Infosys is not simply giving the savings away in every deal. Sanghrajka said the company intends to hold margins at around 21%, and cited "a large deal that we walked out of, or we did not pursue because it did not meet our return expectations." That is the tension a smaller service firm will feel too: AI can help you sell more work, and it can also become the reason a client wants a lower price on the work you already do.

Nate B. Jones used that Infosys pattern as the hook for a Nov 15, 2026 briefing. The rest of his post is paywalled. This article uses only what his free preview states, and it takes the Infosys facts from the public coverage above, not from his paid text.

A bearded man in a grey blazer and a curly-haired woman in a white shirt go over papers with pens at a round table in a modern office cafe.

The people who buy AI are learning to spend less for more of it.

The same pressure shows up when businesses buy the models themselves. In the October 2026 Ramp AI Index, Ramp lead economist Ara Kharazian wrote that August and September had brought "two straight months of virtually no growth in token spend," even as token volumes rose. In the November index, published Nov 11, 2026, he reported that total AI spending among businesses on Ramp rose 1.65% in October, to a new peak. Token volume, he wrote, "continues to outpace token spend as AI gets cheaper and cheaper." The average price per million tokens fell 21% over the last month. Those figures come from Ramp's own customer base, not a census of every U.S. firm. They still describe a market in which usage can climb without the bill climbing at the same pace. If your AI costs are flat or falling while output rises, a client can reasonably ask whether their price should move too.

Once a buyer can test a cheaper option, the premium has to be explained.

Jones's free preview describes a change in how his enterprise conversations have gone. A few months ago, people wanted advice about which model to use. Now, he writes, they have "figured out a way to answer that themselves. They choose a relatively cheap model, test it against their requirements, and use it if it passes." For the work they are buying, "several models are interchangeable, and they can show you the evidence." He calls that "eval fluency," and says it becomes purchasing power once it reaches procurement, because the supplier then has to defend the premium.

You do not need his full briefing to apply the point. If you sell a service, your clients can try a cheaper tool, or another firm that uses one, and ask why your price did not move. If you buy AI, you can run that test before you renew a higher-priced model. In both directions, evidence that a cheaper option meets the requirement changes the conversation.

Decide which savings you keep, and which you pass on.

The mistake is to treat every AI efficiency as either a secret or an automatic discount. It will not stay a secret for long. An automatic discount can also leave you no better off.

Jones's preview includes a worked example, which is his illustration, not a figure from Infosys's accounts: AI lifts the margin from 40% to 50% after a price cut, "and the dollars kept per job end up where they started." The percentage looks healthier. The cash per job does not. That is what happens when you pass through the entire saving.

A more deliberate split looks like this:

  • Keep savings that buy speed, quality, or capacity. If AI lets you return work sooner, catch more errors, or serve more clients without hiring at the same pace, that is value you can hold, or turn into a better promise, rather than a lower unit price.
  • Pass on savings where the client can already do the comparison. If a task is now obviously cheaper to produce, and several firms can do it with similar tools, expect the price to move. Decide how much you will move before the client asks.
  • Walk away when the discount is larger than the gain. Infosys's public stance, as reported by The Register, was that some deals no longer met its return expectations. A privately held firm can make the same call on a smaller scale.

Write that split down for the services AI has actually changed. If you wait until a client emails you, you will be negotiating from their math.

Price the result, not the hours, on work AI has made faster.

Hourly billing is the easiest way to hand the client the whole saving. If a job that took ten hours now takes six, a rate card priced by the hour shrinks the invoice on its own. You did not choose to pass the savings through. The billing unit did it for you.

Pricing a defined outcome, a finished deliverable, a monthly service level, or a project with a clear result, lets you keep some of the time you no longer need, as long as the client still gets what they bought. You can still offer a lower price when you want to. You are no longer forced to. This is not a new product, and it is not a call to raise prices just because you use AI. It is a way to stop the clock from giving away a decision you have not made yet.

A practical next step is narrow. List the offerings where AI has changed the time or the cost. For each one, write whether you will keep the gain, share part of it, or cut the price. If that offering is still billed by the hour, and the hours are what AI reduced, rewrite the offer around the result before the next renewal or proposal.

Your customers can tell that the tools have changed the work. They are going to ask for the discount. The businesses that hold their earnings will already have decided what the answer is.

We can help you with that.

NTx AI Consulting works with owner-led and partner-led privately held businesses that are currently using AI tools in everyday workflows, or would like to be. The complimentary consultation is a working session, not a product demo. If a deeper engagement makes sense, Phase 1 is an AI Disruption Readiness Audit. You can learn more at https://ntx.ventures/ai-consulting/.

Sources

  • Simon Sharwood, "AI slows growth for Infosys and keeps it afloat at the same time," The Register, Oct 28, 2026. https://www.theregister.com/channel/2026/10/28/ai-slows-growth-for-infosys-and-keeps-it-afloat-at-the-same-time/5305879
  • Ara Kharazian, "October 2026 Ramp AI Index: Cracks in the AI thesis, part 3," Ramp, Oct 14, 2026. https://ramp.com/data/ai-index-oct-2026
  • Ara Kharazian, "November 2026 Ramp AI Index: New Fronts in the AI Trade," Ramp, Nov 11, 2026. https://ramp.com/data/ai-index-nov-2026
  • Nate B. Jones, "AI Evals for Buying AI: Keep More of the Savings" (preview), Nov 15, 2026. https://natesnewsletter.substack.com/p/ai-evals-pricing-power

Sources

  1. The Register reported. https://www.theregister.com/channel/2026/10/28/ai-slows-growth-for-infosys-and-keeps-it-afloat-at-the-same-time/5305879
  2. Nov 15, 2026 briefing. https://natesnewsletter.substack.com/p/ai-evals-pricing-power
  3. October 2026 Ramp AI Index. https://ramp.com/data/ai-index-oct-2026
  4. November index. https://ramp.com/data/ai-index-nov-2026