"AI-accelerated development" can sound abstract, or like marketing. So here is what a typical build week looks like in our studio, day by day. It is a composite of many weeks rather than one specific project, and it assumes a web application in the build stage, with discovery and design already approved.
Monday: plan the week in small pieces
The week starts with planning. We take the next set of approved features and break each into small, well-defined tasks: "add a filter by status to the bookings list, including the empty state and tests" rather than "improve bookings". Each task gets acceptance criteria that say what done looks like.
This step has become more important with AI, not less. AI tools work best on small, precise tasks with clear boundaries. Vague tasks produce vague code. The time we spend writing good task descriptions on Monday is repaid many times through the week.
We also flag which tasks need careful human design: anything touching permissions, payments, data migrations or unclear business rules. Those get an engineer's full attention from the start.
Tuesday: drafting at speed
Tuesday is usually the fastest day. Engineers work through the familiar tasks with AI tools such as Claude Code: new screens, forms, API endpoints, validation, database migrations. The AI drafts; the engineer reads every change, adjusts it and runs it.
A typical rhythm: describe the task and point the tool at the relevant code, review the proposed change, ask for adjustments, run the tests, fix what fails, then move on. Tasks that used to fill an afternoon often take an hour. The engineer's attention shifts from typing to judging.
Wednesday: the hard parts
By midweek, the familiar work is largely drafted and the harder tasks remain: the permission rule with three exceptions, the integration with a poorly documented API, the report that needs data from four tables. AI still helps here, by reading documentation, suggesting approaches or explaining unfamiliar code, but the engineer leads and the pace is slower.
This is also when questions for the client come up. "When a manager is on leave, who approves requests?" AI cannot answer that. We send questions in writing, with options, and keep working on other tasks while we wait.
Thursday: review, tests and security
Thursday is heavy on quality. Engineers review each other's work, focusing on logic, security and consistency with the rest of the codebase. AI helps draft additional tests, especially for edge cases, and engineers extend them with the cases that matter most for the business.
Automated checks run on every change: tests, code style and security scans. Anything touching sign-in, permissions or personal data gets a specific security review. Our review guide lists the checks.
Friday: demo and written update
The week's work is deployed to a test site, and we show it to the client, usually in a short call. The client clicks through the new features on their own devices. Feedback becomes tasks for next week.
We also send a written update: what was finished, what is next, any decisions needed and any risks. If something turned out harder than expected, it is in the update with options. AI helps draft the update from the week's completed tasks; a person edits and sends it.
What is different from a traditional week
- More gets built: familiar work moves much faster, so more features reach the demo.
- More gets tested: tests and edge cases are cheaper to write, so there are more of them.
- More time goes to judgment: engineers spend a larger share of their time on design, review and questions.
- Planning matters more: clear, small tasks are what make AI effective.
What is the same
- Unclear requirements still slow everything down.
- Waiting on third parties, such as API access or app store review, takes as long as it always did.
- Hard problems are still hard.
- Clients still need to see working software regularly and make decisions.
That is why we never promise a single speed multiplier. Some weeks are dominated by familiar work and fly. Others are dominated by hard questions and move at a normal pace. See where AI helps in development and where it does not.
What the client does in a week
A client typically spends an hour or two a week with us: answering questions, attending the demo and approving what is ready. Fast answers make a big difference. When decisions arrive quickly, the speed of AI-assisted building carries through to delivery. When they wait, the whole project waits with them.
Read more
For the full process, see how we work and AI-accelerated development. For how this affects budgets, read how AI changes software development cost and how long it takes to build an app. Planning a project? Tell us about it.
