We are a software agency writing about whether software agencies have a future. That gives us a conflict. The useful way to handle that conflict is to argue the hard side properly, rather than work backward to a reassuring answer.
So we should say the position plainly. The traditional agency model is in more trouble than most agencies admit. The trouble is structural rather than cyclical. The parts that survive are different from the parts most firms are trying to protect.
Agencies sold capacity first, then skill, speed and accountability
Take away the positioning, the pitch decks and the language around partnership. A software agency has usually sold four things, in roughly this order of revenue.
- Capacity. You needed six engineers for nine months and did not want to hire six engineers. This was the largest line for most firms. It was never a skill. It was an arbitrage on employment friction.
- Skills you did not have. Mobile, or data engineering, or whatever was scarce that year. This had real value, and it was temporary by nature, because scarce skills stop being scarce.
- Speed. A team that has done this before moves faster than one learning. That value is real, and it is being compressed hardest right now.
- Accountability. Somebody outside your organisation can be held to a delivery. This was the smallest line historically. We think it is the one that survives.
The capacity business is the weakest part now
The capacity business is in the worst position. If a client’s own engineers become meaningfully more productive, the client needs fewer people. The marginal need that used to be outsourced disappears first.
Body-shop and staff-augmentation firms are feeling this now. Many describe it as a soft market. We do not think that explains it. The change comes from the structure of the work, not from a cycle that simply turns back.
The scarce-skills business is eroding in a different way. The knowledge premium on unfamiliar technology has collapsed. A competent engineer with a model can become adequate in an unfamiliar stack in days rather than months.
What remains valuable is narrower. It is judgement about that stack under load and at scale. That is a much smaller offering than simply knowing the technology.
The speed business is being competed away by the client. When your client’s team can build a working prototype in an afternoon, your speed is measured against their new baseline, not their old one.
Accountability has strengthened. The volume of software being produced went up. The average confidence in it went down. Somebody who will put their name on a delivery is worth more than they were.
The important number is how much revenue comes from accountability
The number that decides an agency’s future is simple to state. What proportion of your revenue comes from accountability rather than from capacity.
A firm where clients buy outcomes, with a scope, a price and consequences for failure, has a business. A firm where clients buy people by the month has an arbitrage that is closing.
Most firms sit somewhere in between and do not know the ratio. They have never had to distinguish between the two. Calculating it is useful, and the calculation is uncomfortable at most agencies we have discussed it with.
Keeping the old billing model makes the later change harder
The tempting response is obvious. Keep billing as before, deliver faster, and enjoy the margin. It works, briefly, and it has two problems.
- It is a bet that clients will not notice.
- It makes the eventual transition harder.
The same tools that made you faster made clients better at estimating. The information asymmetry that made the old arrangement comfortable is closing from their side, not yours.
The second problem is slower and more damaging. A firm that has spent three years quietly harvesting a margin has no fixed-price estimating discipline, no scoping practice and no track record of carrying delivery risk.
When the market forces the change, those capabilities take years to build. The firm has to build them while its revenue is falling.
We repriced, and it cost us
We repriced, and it cost us.
We moved to fixed scope, fixed price, with the conventional comparison stated in every proposal. Revenue per engagement fell materially. Volume and margin rose, because the offer is straightforwardly better for a buyer who can now check the arithmetic.
Two things were harder than expected.
- Estimating properly. It is a genuine discipline, and we were not as good at it as we assumed. We discovered that by being wrong and absorbing it.
- Declining work. It is culturally difficult, because every instinct in a services business says take the revenue.
We also stopped selling capacity. We do not offer staff augmentation. Principle was not the reason. It is the part of the business with no future, and keeping it would have kept us optimising for the wrong things.
The parts that survive are about risk and judgement
Four things survive, and none is about writing code faster.
- Carrying delivery risk. A client who wants a fixed outcome at a fixed price needs someone willing to be wrong at their own expense. That is a balance-sheet capability as much as a technical one. A freelancer or an internal team does not supply it.
- Judgement under consequence. This means knowing which architecture survives five years, which shortcut is fine and which is not. This is scar tissue more than stored knowledge, and it does not transfer through a model.
- Domain-shaped engineering. This means knowing what a HIPAA audit asks, how a payment reconciles, why a school laptop is the performance budget. It is sector knowledge fused with engineering, which is why we built out industry pages rather than service pages alone.
- The uncomfortable conversation. This means telling a client the project is not worth doing, the vendor they like is wrong, or their own process is the constraint. Nobody internal can say this safely and no tool will say it at all.
Most firms will not replace the lost capacity revenue
The honest answer is that for most firms nothing replaces the capacity revenue. They will be smaller.
For firms that adapt, the replacement is some combination of three things.
- Higher-value engagements at lower volume, sold on outcome rather than headcount.
- Products, which is the route we took with Open Lance and BotUp, converting delivery capability into something that earns without a client.
- Specialisation deep enough that the alternative is genuinely not available, which usually means a sector rather than a technology.
All three are harder than selling capacity. That is why capacity was the largest line for so long.
Three types of agency are now moving in different directions
The following are composites, drawn from firms we know or have competed against. The abstract argument is easier to understand when it has shapes attached.
The staffing firm is exposed first
Sixty engineers, almost all placed on client sites at a day rate. No fixed-price work, no products, no sector specialisation.
Revenue held through last year and is now falling, because clients are renewing fewer seats rather than cancelling contracts. That makes the decline gradual and easy to attribute to a soft market.
This firm has no obvious route out. Everything it would need to build a different business takes years: estimating discipline, scoping practice, sector depth. It has to build them while revenue falls.
The generalist build shop can feel comfortable for a while
Twenty people, project work, mostly time and materials with a soft cap. Faster delivery has improved margin and nobody has repriced.
The next twelve months will be comfortable. The risk is that comfort delays the transition until buyers force it. At that point, the firm discovers it has never carried delivery risk and does not know how to price it.
The sector specialist is already closer to the surviving model
Twelve people, all in one regulated industry, selling fixed-scope work at prices that look high per unit and are cheap per outcome.
Demand has risen, because every client in that sector now has an AI question. Very few suppliers can answer it credibly while also passing a security review.
This firm is not competing on speed at all.
The instructive part is that the third firm did not do anything clever recently. It made a positioning choice years ago and the market moved toward it.
We repriced later than we should have
Since the purpose of this piece is to argue against our own comfort, two things we got wrong matter most.
- We repriced later than we should have.
- We underinvested in sector depth for too long.
There was a period of perhaps eight months where we knew the arithmetic had changed and kept billing as before. We told ourselves we were gathering data. We were enjoying the margin.
That period cost us nothing financially and cost us a lead we could have had.
We also underinvested in sector depth for too long, on the theory that engineering quality was portable across industries. It is portable, and it is insufficient.
The firms doing best right now are the ones who can talk about an audit, a settlement cycle or a school bell without needing it explained. Building that took us longer than repricing did and it matters more.
Buyers should separate capacity from accountability
If you engage agencies, the useful diagnostic is simple. Ask how they price, and what happens when they underestimate.
A firm that bills by the hour and treats overruns as additional invoices is selling capacity, whatever the deck says. That may be exactly what you want, and you should know that you are carrying the delivery risk.
A firm that prices fixed and absorbs its own errors is selling accountability. It costs more per unit of scope. The price includes something the first arrangement does not contain at all.
Neither is wrong. Confusing them is expensive, and most procurement processes compare them on rate as though they were the same purchase.
In-house teams face the same pressure
The same forces act inside client organisations. The conclusions are different but related.
If your team is faster, the question your finance function will eventually ask is why headcount has not fallen. The defensible answer is that capacity was never the point.
The work has shifted toward things that are harder to see: specification, review, integration and the judgement about what not to build. That answer is much easier to give if somebody has been measuring it.
The corollary is that the internal work most at risk is the same work that was most at risk externally. Routine implementation against a clear spec is being compressed everywhere, in-house included.
The roles that strengthen are the ones nearest the ambiguity. That means whoever turns a business problem into something buildable, and whoever decides whether what came back is acceptable.
Our argument depends on accountability staying valuable
Our position assumes accountability stays scarce and valuable. There is a plausible world where it does not. Verification could become cheap enough that clients can check outcomes themselves, and the willingness to carry risk could stop being worth a premium.
If that happens, this argument is wrong and we will have built the business around a moat that evaporated. We think it is unlikely within five years, because verification requires expertise and expertise is the scarce thing.
We would rather state the assumption than let it sit unexamined underneath everything else we have written.
The commercial structure is on engagement models, and what we commit to on every engagement is on how we work.
