The visible interface is only part of the work. Data readiness, existing systems, review requirements and the first release boundary all shape a useful estimate. A clearer brief makes the trade-offs easier to discuss.
Break down the delivery scope
Ask what the estimate includes: discovery, interface development, integrations, AI behaviour, testing and release. Each piece should have an output and a few clear dependencies.
A small enquiry assistant using one source is different from a system that updates several operational databases. Counting screens will not explain that difference. Ask which unknowns make the estimate most uncertain and how the team proposes to investigate them.
Identify data and supplier dependencies
Vendor permissions, sample data and test accounts can delay progress even when developers are ready. Put these needs into the plan and name who can resolve them.
A short investigation can remove a large uncertainty. Before planning a full integration, for example, test whether the existing platform supports the operation the product needs. Discovery should give you evidence for a decision, not just more slides.
Include ongoing operating costs
Think beyond model calls. Hosting, storage, indexing, monitoring, external services and human review may all contribute to ongoing cost. State the assumed users, requests and documents so the estimate can be revisited as the product grows.
For an AI-assisted task, ask about the cost of a useful completed outcome. Retries and corrections matter too. A pilot can help establish a realistic baseline; the first trial can make the resource requirements much clearer.
Define the first release and acceptance criteria
A good first milestone proves a complete but narrow journey for a defined user group. It should help answer the next question: expand, revise or stop.
Bring the current process, desired outcome, systems involved and any fixed dates to the initial conversation. We can then discuss a sensible starting scope and the assumptions behind it. The plan should reflect the needs and constraints of your particular project.
Before you start
- Separate one-time delivery work from ongoing operation.
- List assumptions about users, requests and source data.
- Identify access or vendor decisions that can block progress.
- Request a first milestone with a demonstrable outcome.
Related engineering work
Our Infinite Growth practice includes page audits, search analysis and prioritised reports. These distinct work packages show why the visible interface is only part of a product’s delivery scope.
View the project case