In short
An ongoing partnership does not mean unlimited software development for a fixed fee. New development goes through the backlog, and it is clear up front what the desired result is and what investment it takes. We use three simple forms for this.
One form per backlog item
For every backlog item we choose one form of agreement in advance, never a mix:
| Form | When | What is fixed | When is it done |
|---|---|---|---|
| Blue: fixed price | the result can be described precisely in advance | the result and the price | when it meets what was agreed |
| Green: fixed budget with a review point | the goal is clear, the route to it is not | the budget | when the business finds the result usable |
| Yellow: timebox | nobody knows in advance whether it will work | the maximum time and investment | after the timebox: there is an analysis, prototype or advice, and you decide whether we continue or stop |
How do we choose? When we describe a backlog item, we propose a form; you decide. A timebox can lead to a new backlog item with a fixed price.
Blue: fixed price
When the work is predictable enough, we agree a fixed price in advance. How many hours we need internally then makes no difference. Does it go faster than expected thanks to AI or an existing approach? Fine. You buy the agreed result, not our hours.
Green: fixed budget with a review point
Development with AI is not always fully predictable. Take having AI recognise and classify documents: the goal is clear, but nobody knows in advance exactly how much testing and adjustment it will take. We agree a budget and decide together when the result is good enough. The business decides whether it is usable, not a developer or a theoretical score. If you want to optimise further after that, it becomes a new backlog item.
Yellow: timebox
Sometimes nobody knows in advance whether something will work, especially with new applications of AI. Then a full scope and price is false certainty. We agree a limited research investment, for example a few hours, with a maximum known in advance. After that there is something usable: an analysis, prototype, advice or investment proposal. And you decide again: continue or stop?
Why this works
A small experiment that shows an idea is not sensible can be far more valuable than a large project that should never have started. Big uncertainties become small decisions: research first, then decide, build a first version, review it, and only then invest further.
Would you like advice on agreeing on software development? Request a free advice conversation.
Updated on