Eventus DIGITALReadiness check
← Insights

Evidencing Value Creation

AI can cut a contract's margin by 50%. Or increase it by 52%. What's the secret?

On one contract, sold on inputs, AI cuts the margin by half. Sold on the impact it delivers for the customer, the same contract lifts gross margin by 52%. Same engineers. Same customer. Same technology.

George Davies · 18 October 2026 · 9 min read

You lead or back a B2B service-centric business: people plus software, and now AI. Managed services, cloud services, outsourcing, customer services. You are deploying AI. Your team does more with fewer people. Your customer knows it, and expects to see the saving in its own costs.

So how are you going to drive business value from AI, when your business is set up to deliver hours, seats, headcount and tickets? Every figure here comes from one contract: an IT helpdesk, followed from Value in Exchange to Value in Use, from both sides.

1. Value in Exchange: the squeeze from both sides

Most service-centric businesses sell Value in Exchange. They are paid for what they hand over, not for what it achieves: engineers and tickets in IT, seats in contact centres, headcount in outsourced operations, day rates in consulting. They plan capacity, measure performance and reward their people on the same units. AI now attacks that model from two sides at once.

From the inside, your own AI. Tickets resolve faster. Junior engineers fix what used to need a senior. Gartner expects agentic AI to resolve 80% of common customer service issues without a person by 2029. Salesforce has already taken its own support headcount from about 9,000 to about 5,000. Under a contract priced on inputs, every unit your AI removes is a unit the customer stops paying for at the next renewal — and the customer expects it.

From the outside, low-cost AI providers quote per ticket, per seat or per hour at a fraction of your cost. ISG reports that unit prices in managed services contracts now fall about 20% by the end of year two, double the historical 10%. HFS Research, with Capgemini, found 77% of buyers assume they are overcharged when they have no price benchmark.

The example: the customer's own team handles about 7,800 working-day cases a year. The supplier handles nights, weekends and overflow — 4,200 cases, 35% of the total. Cases the supplier cannot resolve go to the customer's six senior engineers.

−25% revenue · margin nearly halved

What AI does to the contract under input pricing. Add ISG's 20% unit-price fall and revenue drops to £198,000 against £200,000 of cost: the contract stops making money.

The service did not get worse. It got better, and a business driven by inputs is paid less every time it gets better. Paid for engineers and measured on response times, the supplier has no reason to resolve more cases, so the customer's escalations stay where they were.

Investors see it. Everest Group reports private equity cooling on IT services firms built on staffing and volume. N2M Capital found break-fix and project-heavy managed services businesses selling for 3 to 5 times EBITDA, against 9 to 13 times for AI-enabled platforms.

2. Value in Use: the same AI, with a contract built to recognise contribution

Now sell the same contract as Value in Use: not the engineers the customer receives, but what the service does for the customer's business, measured in the customer's own numbers.

  1. 1

    A base fee

    For the service itself.

  2. 2

    A share of the efficiency gain

    The saving the service creates in the customer's cost.

  3. 3

    A performance fee

    For the improvement it delivers — such as better renewals.

What changes from one business to the next is what is measured, not how the contract is built. Paid for capacity, the supplier uses AI to cut its own cost; paid for outcomes, it uses AI to cut the customer's. The low-cost AI that threatened your price becomes part of your cost base, and you keep the saving.

2.9×

Year-one gross margin under Value in Use versus Value in Exchange. Same engineers, same customer, same AI.

£573,058 (44%) vs £163,500 (20%)

Three-year gross margin: 3.5 times. The customer's base grows 44% while engineers rise 20%, and margin moves from 38.5% to 49%.

3. The customer's view: why it works

The customer is a software company with £60m of recurring revenue, where each point of customer renewal is worth £600,000 a year. Under Value in Exchange, its only saving is the cut in the supplier's fee. Its senior engineers still handle 1,050 escalations a year at a cost of £210,000. Under Value in Use, that cost falls by £84,000, because the supplier is paid to reduce escalations.

−16%

Customer support cost under Value in Use (£86,550) — more than repricing would give, with no renewal gain needed. Plus 420 fewer escalations and a desk that absorbs 44% growth on frozen headcount.

Renewals sit on top. If they improve by a point, the customer keeps £600,000 of revenue a year and pays the supplier 5% of it, £30,000. If they do not, it pays nothing for them.

That answers the questions raised on last week's edition. Paul Chapman asked how to make it easy to buy: the finance director sees a support cost 16% below today's, with every outcome fee paid only from a proven gain. Peter Skodny asked how to fund the gap: year-one revenue is £357,450 with every fee earned, and £327,450 before any renewal fee, against today's £330,000 and £247,500 under repricing — there is no gap to fund. Jana Zimova asked what happens if the outcome does not arrive: gross margin is £77,650 (26%), still above the £47,500 repricing gives.

So what is the secret?

It is not the AI. Everyone has the AI. It is not the price list either. Change the invoice alone and the first quarter ends in a dispute. The secret is proving the outcomes that matter: how you evidence the impact of your services on the outcomes that matter for the customer.

Outcome contracts without proof fail in predictable ways. The customer's season changes. The customer restructures. Someone questions the starting point. HFS found the most common reason buyers keep contracts labour-based is that the outcome cannot be defined, measured or tracked: 23% named it first. Yet the same buyers expect outcome or value-based pricing to grow from 12–15% of AI services spend to about 34% within two years. The demand is coming. The proof is what is missing.

  1. 1

    Define

    Agree the outcomes that matter to the customer — cost to serve, resolution, retention, revenue — and sign the starting point into the contract.

  2. 2

    Quantify

    Put a value on the expected improvement, in efficiency and performance, and agree a fair share before the work begins.

  3. 3

    Prove

    Show which interventions moved the result, so the supplier is credited only for the change it drove.

  4. 4

    Recognise

    Carry that evidence into monthly and quarterly reports, in the customer's own numbers, so the agreed share is paid without argument.

Get any of the four wrong and the contract becomes an argument. Get all four right and the fee is a calculation. This is also why a low-cost AI provider cannot follow you. It can undercut the price of a ticket. It cannot show what its service did to your customer's renewal rate. The baseline, the evidence and the customer knowledge sit with you.

The cost of waiting

Every contract that renews on inputs this year, while AI takes the effort out, locks the loss in for another term. How many of yours renew in the next 12 months? Your exact number depends on your contract terms, your ticket history and your customers' cost to serve. Only your data shows it.

+52% · +58% · 3 yrs

One IT managed services contract: a 52% gross margin improvement (30% to 46%), 58% revenue growth, and a three-year term, up from annual.

Sources

  1. Gartner, press release, 5 March 2025 (Daniel O'Sullivan): agentic AI to resolve 80% of common customer service issues without human intervention by 2029. Forecast.
  2. Salesforce support headcount from about 9,000 to about 5,000: Marc Benioff, The Logan Bartlett Show, reported by Fortune, 2 September 2025.
  3. ISG Index Insider, Stanton Jones and Mike Rose, 31 October 2025: unit prices in managed services contracts fall about 20% by end of year two, against a historical 10%.
  4. HFS Research with Capgemini, "Stop pricing AI like labor", Divya Iyer and Saurabh Gupta, 21 September 2026. 101 financial services firms. Financial services sample only.
  5. Everest Group, Abhishek Singh, 5 May 2026: private equity cooling on IT services as AI automates the base of the delivery pyramid. Qualitative.
  6. N2M Capital, MSP M&A Valuation Report 2026, June 2026. 120 transactions; break-fix and project-heavy 3–5x; AI-enabled platforms 9–13x. Indicative ranges.
  7. Illustrative Value in Use model of an IT helpdesk contract: six engineers billed at £55,000, direct cost £40,000 each, AI tool £20,000 a year; customer with £60m recurring revenue growing 20% a year, one renewal point worth £600,000; renewal fee 5% of proven gain; escalations valued at £200 each. Modelled, not reported results.