Moving a Consultancy Off Billable Hours

Trinzo sells hours. That is the model, and for most of its history it was a good one: a client with a regulatory problem buys expert time, and the expertise is real, and the time is genuinely required. Then the time required starts falling, and the model starts eating itself.

Nobody sends a letter announcing this. There was no lost bid, no client asking why a deliverable took three weeks, no competitor undercutting us on rate. I want to be precise about that, because it is the only unusual thing in the story and everything else follows from it. We started before it hurt.

Four things carry this account. That the threat was identifiable in advance and we acted on it early. That the first adoption win has to be placed where it costs nobody their billable hours. That the commercial model has actually changed, not just the delivery cost. And that the expensive failure in the middle is what produced the thing that now works.

Section 1The threat, before it arrived

Trinzo is a life sciences consultancy and I am its Head of Innovation. I did the research myself, over months, and what I found was not one pressure but four arriving together on different timescales.

Rates.

Clients who can see a task compressing do not keep paying the old price for it. This is the slowest of the four and the most negotiable.

Hours per deliverable.

The direct one. A submission section, a gap assessment, a remediation plan: each of these took a known number of days, and the number was falling. In a time and materials business, efficiency is revenue reduction. That sentence is the whole problem and everybody in a consultancy already knows it.

Clients doing it themselves.

The work we were most comfortable selling, structured, procedural, evidence assembly, is exactly the work a client can now attempt in house. Not all of it, and not well, but enough of it to change what they are willing to outsource.

New entrants pricing differently.

Firms with no legacy day rate to protect, quoting for an outcome instead of a team. They do not have to beat us on quality. They have to make our pricing look like an artefact of how we are organised.

Any one of these is a margin problem. All four at once is a model problem.

High confidence in the four pressures; I have watched all of them operate. Lower confidence in their relative speed, which I got wrong: I expected rate pressure first and it has been the slowest to bite.

Section 2Skunkworks, then a business case

It did not begin as a strategy. It began as a small piece of self interest: I wanted my own operations team to be faster and better, so I trained them on the tools and we worked out what actually helped. Productivity first, upskilling alongside it. No budget line, no mandate, no announcement.

That was the right way round, and not because I planned it that way. A skunkworks that improves one team's work produces evidence. A strategy paper produces a discussion.

Once I understood the size of both the threat and the opportunity, I took a business case to the leadership team. I was on the leadership team, which sounds like an advantage and is a mixed one. I was not asking a boss for permission. I was asking peers for budget, against their own priorities, in a business where every euro spent on this was a euro not spent on something with a nearer return. Buy in had to be agreed communally. Nobody could impose it, including me.

If I were doing this again in another firm I would still do it in that order. Evidence from one team, then the case. The case on its own does not survive contact with a finance discussion.

Section 3Where to put the first win

There was a lot of resistance at the start, and most of it was reasonable. If your compensation, your standing and your sense of professional worth are all measured in billable hours, and someone arrives proposing to reduce the hours a task needs, scepticism is not obstruction. It is a correct reading of your own position.

What turned it was a sequence of early successes and a great deal of communication, sustained past the point where I was bored of saying the same thing. The first substantial win was rolling Copilot out to the sales team.

I would like to claim I chose that target for strategic reasons. Partly I did and partly it was where the appetite was. Either way it is the most useful thing in this account, so I will state it as a principle: put the first win where it does not reduce anybody's billable hours.

Sales was ideal. There is no billing model to threaten, the output is proposals and research and follow up rather than regulated deliverables, and the improvement is visible within days. People who had heard the argument in the abstract saw a colleague doing more with the same week. Delivery teams did not have to take my word for it and they did not have to go first.

Adoption in a professional services firm is not a training problem. It is a sequencing problem.

Section 4The year the floor moved

Then the market turned, and it turned hard, near the start of the programme. The core business lines suffered. Trinzo contracted sharply and we lost a substantial proportion of our people. To be exact about the causation, because it matters more than anything else on this page: that was the market, not the AI programme. The two happened at the same time and one did not cause the other.

We kept investing anyway.

That is the decision the whole thing rests on, and it was not obviously correct at the time. The safe move when revenue falls is to protect the core, defer the speculative work and revisit it when conditions improve. Plenty of firms did exactly that, and I would not call them wrong on the information they had.

The argument that won was that the contraction and the disruption were not separate events to be handled in sequence. A market that has turned is a market in which clients are more price sensitive, more willing to try doing it themselves, and more open to a supplier pricing differently. Waiting for conditions to improve meant arriving late to the thing that was changing the conditions.

I am aware of how convenient that argument is coming from the person who wanted the budget. It happened to be right, so far. It could have been both convenient and wrong.

Section 5The expensive mistake

We committed to building an audit tool as a software product we could sell to clients. It was expensive, it was early, and I agreed to it before I fully believed in it.

Two things were wrong with it. The tool was never reliable enough for the use it was sold on, which is a fixable problem given enough time and money. The second one was not fixable: Trinzo is a consultancy, and we did not have the core competency to sell, support, price, version or maintain a software product. Those are different businesses. Being good at the domain a product serves is not the same as being able to run a product.

The learning that actually changed how I decide things is narrower than "validate before you build". It is this. When a proposal requires the business to become good at something it has never done, the risk is not in the thing being built. It is in the capability that has to exist around it, and that capability is almost never in the business case.

I would not remove this from the story if I could. The audit tool is what produced the answer in the next section.

Section 6What we actually sell now

The failure clarified the question. We are not going to sell software. We are going to sell the work we have always known how to sell, done differently underneath.

Two things changed commercially, and only the second of them is genuinely a model change.

The first is that the same work now takes fewer hours, which improves margin on any fixed fee and worsens revenue on any day rate. On its own that is not a transformation. It is the disruption happening to us slightly more comfortably than it happens to somebody else.

The second is that we can now price on project outcomes rather than days, and we have a product range built on that basis: remediation projects, lean improvement projects, quality and regulatory projects. These are productised services rather than software. The scope, the deliverables and the price are defined in advance, and the AI capability sits underneath making them deliverable at that price. A client buying one of these is not buying a team for a period. They are buying a defined result.

That is the part that matters. Everything else is efficiency, and efficiency inside a time and materials model is a slow way of getting smaller.

Section 7What moved, and how far I trust each figure

Separating what I can evidence from what I believe, because on a page written by the person responsible the distinction is the only thing worth reading.

Margin on the new work.

Roughly 40 per cent higher than comparable projects delivered before the change. This is the one hard number in the account and it is a like for like comparison of project margin, not a company figure. High confidence.

Leads and pipeline.

Marketing and sales leads are up substantially across the whole business, not only on the new offerings, and direct sales of the new work have landed and are growing. High confidence that the increase is real. Moderate confidence in attributing it to this programme rather than to a market where everybody is suddenly asking about AI. The honest position is that we did something visible in a period when the topic became unavoidable, and I cannot cleanly separate the two.

Brand and price protection.

Better quality of lead, and an easier time holding price in a market where price is under pressure. Moderate confidence. This is the kind of effect I can feel in a negotiation and cannot put a number on, and I would treat anybody who does put a number on it with suspicion.

Back office efficiency.

The back office is smaller than it was and supports more billable people than it did. I originally wanted to bank that as a result of automation, and I do not think I can. Trinzo contracted for market reasons in the same period, so the ratio improved for at least two reasons at once and I cannot separate them with the evidence I have. Low confidence as an AI result. Real as a fact about the business.

The bets.

That this line could eventually generate more revenue than the core business, that it supports every other line rather than competing with them, that it puts Trinzo in a leading position in a segment that is still forming, and that it lifts the valuation of the business through brand position. None of these are results. They are the reasons I argue for continued investment, and they are exactly the claims I would expect a reader to discount, coming from me.

Section 8The least flattering reading

Here is the version of this story that is unkind to me, stated properly rather than gestured at.

A senior person identified a threat that conveniently required his own function to be funded and expanded. He secured budget from peers during a downturn, partly on forecasts that cannot yet be tested. One expensive initiative failed outright. The clearest financial result is a margin comparison on a small number of projects that he selected and reports himself. The lead growth coincides with a period in which interest in the subject rose everywhere, and the operational savings coincide with a contraction that would have improved the ratio regardless. Two years in, the case still rests substantially on where he says this is going.

I do not think that reading is correct, but I cannot demonstrate that it is wrong yet, and a reader is entitled to hold it. The thing that would settle it is three to five more years of results, which is precisely what I do not have.

Section 9Where it is now

Nearly two years in, on something I expect to mature over another three to five. The return on what has been invested is already well ahead of what we projected, and the growth is continuing. That is the strongest sentence I am prepared to write.

What is not done is most of it. The outcome based product range is real and it is small relative to the core. The delivery teams have moved much further than they had a year ago and are not where they need to be. And the underlying question, whether a firm built on selling time can end up selling results without losing the expertise that made the time worth buying, is not answered by two good years.

I have written elsewhere about what organisations lose when they let capability drain out of their people. It would be an unusual failure to argue that in public and then run a programme that does it. That risk is live, I watch for it, and I would not claim to have solved it.