William Waldron on how the conversation with senior technology buyers has shifted, what AI-assisted delivery actually changes in practice, and where Forge fits.
I have been hearing the same thing in conversations with senior buyers this year. The question has changed. They have moved past whether AI is going to matter to their business. That one is settled. What they are wrestling with now is harder and more practical. How does delivery actually change when AI is sitting inside it, and what should they be asking suppliers to do differently?
This is not drawn from one engagement. It is the pattern across a number of them over the last while, in different sectors and at different sizes. A few things come up consistently.
They want to own more of what defines how the business operates.
The first is a clear preference for owning more of the software that defines how the business operates. Sometimes that lands at board level as a concern about vendor concentration. In smaller organisations it is usually simpler than that. The realisation that a vendor platform no longer fits how the business has actually evolved. Licence costs matter, and for some organisations they are what tips it. For most, what really pushes the conversation forward is the experience of operating inside someone else's product roadmap. People have got tired of asking for changes that never arrive.
The specification almost never holds.
The second is quieter but ends up mattering more. The specification a customer arrives with is almost never the specification they actually need. That is not anyone's fault. Real requirements only become visible once something is being built against them, and people start to see gaps in their own thinking that no document could ever have surfaced. The first version of any meaningful set of requirements starts to move within weeks of work starting on them. Agile was a real attempt at handling that movement. AI-assisted delivery picks up the same problem and addresses it differently. The cost of reshaping the build itself drops sharply, which means the team can absorb the movement without leaning as heavily on process to manage it. Delivery models that depend on a fixed scope, on the other hand, still struggle in the same way they always did, and the strain usually shows up as a scope-creep conversation neither side really wanted to have.
There is now a proof gap.
The third is a proof gap. Customers who have sat through enough proof of concept presentations have got better at spotting the difference between real software and a slide about it. The bar at the early stages of a conversation has moved. A working demo wins over a deck describing one. That has changed the early commercial conversation more than people realise.
Throughput matters as much as speed.
The fourth is about throughput as much as speed. When a customer has a backlog to work through rather than a single project to deliver, what matters is how many things move in parallel, not how fast any one of them ships. The real economic effect is what happens to a team that can hold more in flight without constantly having to increase headcount, and the more experienced buyers are starting to ask exactly that question. Traditional commercial models built around day rates and headcount are still working through how to answer it.
The bottleneck is rarely the build itself.
Across all of this, the build itself is rarely the part of the lifecycle that determines when something goes live. The time goes on everything around the code, such as access, approvals, security review, environment readiness, integration onboarding, data migration. Which is why speed of code generation is not really where the value is. What AI-assisted delivery actually changes is how much a smaller team can carry, and what kind of team that is.
Where Forge fits.
Forge is what we built at SevTech to handle the kind of work I have been describing. Rob walked through how Forge came together earlier this month. The Forge product page covers the engine in detail.
The short version is this. A smaller team can now carry the workload that used to need a much bigger one. AI now handles most of the engineering production work. The humans set the direction, make the judgement calls, and review what comes back. That frees them to concentrate on the parts of the engagement that genuinely need them. Architecture. Integration with real-world systems. Security. Governance. The audit trail and documentation come out of the build itself, rather than being added on at the end. That matters more than it used to, given the direction of travel in Europe on AI regulation.
For the customer, that lands as working software sooner. Requirements that can move without breaking the contract. Ownership of the code at the end of it. And a delivery setup that can carry more than one priority at a time.
For anyone working through any of this, get in touch.
