Why agencies that believe and sell Agile still ship waterfall

The agency believes in Agile. It says so on the site, its people have the certificates, and inside the delivery team it is telling the truth. Then the proposal arrives: 12 months, six figures, and a scope written out in full before anyone has seen the system. Nobody in that building is a hypocrite. The shape was decided somewhere they do not work.

Where the shape comes from

Start with the number that governs an agency’s week. It employs delivery people on salary, and they are paid whether or not there is work for them, which makes utilisation the figure everyone answers to. An unstaffed week is a loss that cannot be recovered later.

So the pipeline has to be kept full ahead of capacity rather than matched to it. Keeping it that way costs real money: business development, pitching, proposal writing, pre-sales consulting given away. Just over half of that spend buys nothing, because professional services firms win 48.1% of the bids they make. All of it lands before any revenue does.

That cost has to be recovered somewhere, and the only place it can go is the contract value. Spread across a 3-week engagement it is ruinous. Spread across 12 months and six figures it disappears. The long commitment is not ambition. It is arithmetic.

A long commitment also has to say what it is for. Nobody signs 12 months and six figures against “we’ll find out.” So the scope gets written out in full at the one moment in the relationship when the least is known about the problem.

Call that peak ignorance. Everything after the signature is negotiating away from a position taken in the dark.

What it does to the team you hired

Now watch what that does to the team you hired. They start work, they learn things, and everything they learn arrives as a change request. A change request is a commercial event. It gets priced, argued, and approved by someone who signed for the original scope and has a budget to defend.

The insight is real. The structure converts it into a problem. Your team learns, without anyone deciding to, that raising these things costs more than staying quiet.

Why the small version does not get sold either

The obvious fix is to buy less. A 2-week experiment instead of a 12-month build, and another one after that if the first goes well.

It rarely happens, and not because agencies refuse. Every piece of work has to be sold before it can be done: scoped, priced, written up, approved. In the agency I worked in, a proposal took days to write and several rounds to agree, and none of that got shorter when the work was small. Selling costs an agency 5% to 20% of revenue, and most of that is salaried time rather than advertising spend. On a 12-month build it is a rounding error. On 2 days of work it costs more than the work does.

So small experiments are uneconomic to transact even when both sides agree they are the right thing to do. Discovery gets bundled into a phase, and the phase into the contract, because bundling is the only way that overhead gets paid for.

What embedding actually buys

Change one thing and the arithmetic comes out differently. Put the same person inside the client’s team on a standing arrangement and the next experiment no longer needs a proposal. It needs a conversation.

The experiment was always cheap. What cost money was being allowed to run it.

That is what embedding buys: standing permission. The right to ask the next question without selling it first.

The part that sounds backwards

You came here wanting to commit less, and the argument has landed on a standing arrangement. Small in scope, not small in duration. Small units of work inside a durable relationship, not small relationships.

Neither side can buy a durable relationship on day one. It gets built, and a small piece of paid discovery is how it starts: cheap enough to risk on someone you have not worked with, and specific enough that both sides learn something real about the other by the end of it. You arrive at standing permission. You do not sign for it.

Read that as risk management, because it is competing with the usual kind. The other way to handle an unknown supplier is a large contract carrying penalties, liability caps and fixed-price terms, which manage the risk on paper and leave the exposure where it was. A first commitment small enough to walk away from manages it by limiting what is at stake.

From both sides of the table

I have sat on both sides of this. CTO inside an agency, watching the pipeline decide what we were able to sell, and alongside founders on the other side, reading the proposals that came back. The mechanism is invisible from either seat alone. From inside it looks like commercial reality. From outside it looks like the agency preferring big projects.

The clearest case I saw of it working was never sold. An existing relationship, enough trust already banked, and a piece of ad-hoc discovery nobody wrote a proposal for. It found quick wins worth doing straight away, and it grew into a run of projects of varying sizes.

It also killed things. A few days of experiments invalidated work that had been planned for months of delivery. Some of it we replaced with a better alternative. Some of it became a buying decision, because a product already did the job. An agency paid to build has no route to that second answer. Nothing was riding on this one, so the finding was allowed to be inconvenient.

The quiet corruption

There is a second effect and it is worse than the change requests. When the next tranche of work depends on this experiment looking successful, a negative result becomes commercially inconvenient. The discipline the whole method rests on is naming in advance what evidence would kill the idea, and that discipline does not survive contact with a renewal conversation.

Someone embedded on a standing arrangement can report that a thing did not work and call it a useful week. Someone pitching for the next phase cannot, comfortably.

The cost of the fix

The cost of that arrangement is concentration. One person inside the team is a dependency. For an agency that is a risk to be staffed around; for an embedded practitioner it is the operating model. Worth knowing before you choose it rather than after.

What you can actually do with this

If you are buying, the lever is the shape of what gets signed. Stages with exit criteria instead of one scope, each stage small enough that being wrong about it is survivable, and the next one priced only once the last has reported. A good agency will take that structure. One that cannot is telling you something about its cost base rather than about your project.

If you are inside the agency, you cannot fix this from the delivery side. The pressure is upstream of you, in utilisation and pipeline cost, and no amount of ceremony inside the team touches it. What you can change is what gets signed, and that is a conversation with the people who sell, not the people who build.

Nobody is lying to you

The account manager believes the methodology. The delivery lead believes it harder. Both of them are working inside a cost structure that decided the shape of your project before either of them met you.

That is worse than being lied to. There is no one to appeal to.