Loading in progress, please wait

Retour vers le haut

Three points of margin: why the most profitable innovation in Professional Services is a habit, not an agent

Home > Blog article list > 2026 > October > Three points of margin: why the most profitable innovation in Professional Services is a habit, not an agent

Blog | Analysis and insights on post-sales, customer success and AI

Three points of margin: why the most profitable innovation in Professional Services is a habit, not an agent

Three points of margin: why the most profitable innovation in Professional Services is a habit, not an agent

Services firms are being shown AI business cases every budget season. The surest return available to them this year comes from a Friday timesheet and a monthly re-estimate. What the data says, the rituals that produce it, and where AI fits.

In 2025, billable utilisation across professional services fell to 66.4%. That is the lowest level in the nineteen years SPI Research has run its benchmark, built this year on 509 organisations, and it came with an average project overrun of 10.7%. The best performers in the same study ran at 75.0% utilisation, against 64.9% for everyone else.

Now take a figure from a different source. Across the customer base of Stafiz, a professional services automation platform, the teams that update their remaining work and replan their projects on a regular rhythm show a project margin at least three points higher, on average, than the teams that do not.

Three points of margin from a ritual. Ten points of utilisation between the disciplined and the rest. I have rarely been shown an AI business case with returns that certain.


The number was simply old


Aymeric de Ferrières of Stafiz showed me the cleanest example I have seen of how margin disappears. A project budgeted at 93 days. 80 days already consumed. Everyone involved believed it was 65% complete. When the remaining work was properly re-estimated, the answer was 35%.

Nobody had lied about that project. The number was simply old.

I call that gap truth latency: the delay between what is true on a project and what management knows about it. We tell ourselves that margin is lost in the negotiation, in the discount conceded or the rate card under pressure. Some of it is. Most of it is lost later, quietly, in that delay.


What one point is worth


Aymeric put a price on it. Take a firm of 100 people with an average daily rate of around 1,000 euros. One extra point of utilisation is worth about 220,000 euros of additional revenue a year, with the same people, and almost all of it reaches the bottom line.

Apply that to the ten-point gap in the SPI benchmark and the firm of 100 is looking at more than two million euros a year. Stafiz's own customers, who by definition plan their load in a tool, average 77% utilisation.


The rituals, as I run them


None of this is sophisticated. That is the point.

  • Every Friday, everyone declares their time, and project managers approve the timesheets the same day.
  • Every month, a project and account review. Account by account: what was delivered, the margin, the risks, how we work around them, the action plan.
  • A staffing meeting, where allocations are optimised against a detailed skills matrix.
  • A revenue recognition review with finance. Project by project: time declared, milestones validated, invoicing, revenue.

The last one matters more than it looks, because signed is not earned. On a fixed price, you still have to deliver the milestones. On time and materials, you still have to invoice time that had value for the customer. Services teams deliver without invoicing on time far more often than they admit, which is why I follow the conversion of backlog into billing as closely as the margin itself.

Three refinements I learned the hard way.

Time and materials needs this as much as fixed price. A customer who discovers the remaining work on the invoice is just as unhappy as one whose fixed-price project is late. There is always a scope, written or not.

Calibrate the person, not only the project. Some project managers are optimists. Others raise the alarm early and often. An estimate is only credible once you know which one produced it.

Do not worship utilisation. I compute it by country, because holidays differ and cross-border staffing distorts any global average. And I set a ceiling as well as a floor. A target that is too aggressive degrades the quality of delivery, and lost quality costs far more margin than the point of utilisation ever earned.


Where AI fits


In the teams I have led, AI went to delivery first: shared knowledge bases, preparation of workshops, generation of design documents. It did not go to the steering of projects, because the project directors were experienced and the return was clearer elsewhere.

That does not make AI irrelevant to steering. In the same SPI benchmark, the firms that apply AI widely in finance and operations, with measurable benefits, report an EBITDA of 24.2%, against 8.8% for those that do not use it there.

The two findings are not in conflict. An agent can warn you that a project is drifting only if someone updated the remaining work on Friday. The tools now pre-fill timesheets from calendars and chase the people who are late. They cannot invent what the project manager knows. The discipline is the data.

I once had a customer freeze a project because of an internal audit that had nothing to do with our delivery. The restart was pushed back month after month. No model had that information. The project manager did. A forecast is built from the bottom up, from what the people closest to the customer know this week. AI can turn that into scenarios and staffing options faster than any of us. It cannot go and find it out.


Five moves for the executive committee


If I were structuring this for a services organisation this quarter, five moves.

  • Fund the ritual before the agent. If the Friday timesheet and the monthly re-estimate are not in place, an AI investment in steering will produce confident forecasts from stale numbers.
  • Ask for remaining days, never for percent complete. Percent complete is an opinion. Remaining days can be compared with the budget, and argued with.
  • Review backlog conversion every month. What was signed, what was delivered, what was invoiced, what was recognised. Four numbers that should move together and rarely do.
  • Set utilisation by country and by profile, with a ceiling. A manager and a junior do not have the same target, and neither should be at 100%.
  • Point AI at delivery first, and at steering only once the data is fresh. The return on design documents and shared knowledge is immediate. The return on a forecasting agent depends entirely on the ritual underneath it.


The habit, redefined


Stop looking for margin in your next AI use case until you have collected the three points that are already on the table.

The real question is how old the numbers are in your Monday review. If the answer is more than a week, the most valuable innovation in your organisation is a habit, and it costs nothing.

Update the number. Then automate it.

This article is adapted from edition #[N] of my LinkedIn newsletter, Professional Services & Tech. Subscribe on LinkedIn to receive future editions.

Sources:

SPI Research, 2026 Professional Services Maturity Benchmark (19th edition);
Stafiz, figures from its customer base shared during the webinar of 8 October 2026 with Aymeric de Ferrières.



Mathilde HENRY
Executive Leader in Enterprise Software & AI Transformation. 20 years working at the intersection of software, consulting and business transformation.

Blog – Analysis and Reflections by Mathilde Henry

A conversation amongst peers on post-sales, AI and valuation.

For CEOs, division heads, CFOs, board members and private equity operating partners who wish to engage with an executive-level analysis of post-sales SaaS, AI monetisation and transformation at C-suite level.

Use of data

We and our partners use different technologies, such as cookies, to personalize content and ads, provide social media features, and analyze our traffic. Use the buttons to agree or decline.