Intent Translation is our point of view on why building software so rarely produces what the business actually intended. The hard problem in modern delivery is not writing code — teams can build. It is preserving intent as it passes through people, teams, systems, and now agents. Every handoff is a translation, and every translation loses a little of the original meaning. By the time strategy reaches production, a surprising amount of the intent has quietly leaked away.
The hard problem is not building software. It is preserving intent as it moves through people, teams, systems and now agents.
Where intent leaks
A strategic intent passes through a chain of handoffs, and each one is a place where meaning can be lost rather than carried:
- Strategy → Capability — what the strategy actually depends on being able to do.
- Capability → Product — the bets and journeys that realise those capabilities.
- Product → Architecture — the structure that makes those bets buildable and changeable.
- Architecture → Engineering — the implementation that honours the structure.
- Engineering → Operations — running it safely, supportably, accountably.
- Operations → Outcomes — and whether any of it moved the thing the strategy cared about.
No single role owns the whole chain, so the loss is invisible: everyone assumes translation is happening because the boxes on the diagram exist. This is the same seam we name in The Missing Architecture Layer Between Strategy and Delivery and trace to its root in Why Most Product Transformations Fail Before Engineering Starts; Delivery Architecture: The Translation Layer is the operating model we propose for owning it.
Why it matters more in the age of AI
Agents add handoffs, not fewer. When generation becomes abundant, throughput stops being the constraint and the scarce step becomes acceptance — the judgement that what was produced is what was intended. That is The Acceptance Gap: Intent Translation at the seam between what is generated and what is trusted enough to ship. The cheaper generation gets, the more the leakage of intent — not the speed of typing — decides whether AI delivers value.
How the rest hangs off it
Most of what we write turns out to be one facet of this single idea. Decision Architecture is intent becoming a decision. Governance-to-Value is intent becoming authority without becoming drag. Provenance is intent becoming accountability after the fact. They are not separate methods; they are the same concern — preserving intent — viewed from different points in the chain.
Measuring it
You measure translation, not motion. Intent fidelity — how much of the original intent survives to the outcome — is the metric family that matters, alongside the DORA flow and stability signals. Velocity and story points measure how fast the chain moves, not how much meaning it keeps.
A note on honesty: this is a position we hold, formed from years of architecture, delivery, product and transformation work and from watching the agentic shift unfold — not a packaged methodology with a proven track record in AI. We think it is what the new world will need, and we refine it in the open.