Ask us a question

Riverbank · SDLC for AI-first development

AI builds the code. Riverbank organizes the work around it.

Riverbank is an SDLC management tool that embraces AI-first coding — and doesn't forget all the other roles a successful implementation still needs. It guides you through the steps, and it doesn't forget all the people who still have to be in the loop.

Ask us about Riverbank Watch the 1:39 film →

Riverbank runs Bluefish's own work today — fifteen phases of it were built and remembered inside itself, and it is carrying four products on its board. We're looking for the first teams outside Bluefish to run it too.

The case

We still need a lifecycle. It is just a different one.

  1. AI can build code quickly — and needs far less detailed specification than it used to. Nobody seriously disputes this any more. Working software in front of somebody in days, where it used to be quarters, from a few lines rather than a signed-off hundred-page document.
  2. But the work still has to be organized. Without that, the code just sits in repositories with no way forward. Which of these is ready? Which was tested? Which was approved? Which reached a customer? Nobody can say.
  3. So we still need a system to document, manage and deliver software. Just not the current one. The tools we have organize a world where engineering time was the scarce thing and the hand-off was the process. That was the right design for that problem. It is not the shape of this one.

What it doesn't forget

Building the software was never the whole job.

So the tool that organizes the work has to keep asking the people who are still in the loop, at the moment it needs them — and never at any other moment.

Before anything is built

It doesn't forget that we still need to build software people will buy

The business problem, what it is worth, and how we will know it worked — asked before a line is written, not after.

While it is tried, and when it goes out

It doesn't forget that people have to test and deploy

A screenshot and a few sentences is the whole of what a tester is asked for. On the way out: what went out, to whom, what changed, and whether anybody actually took it.

Before it ships

It doesn't forget that QA, regulatory and compliance all have to review and approve

Five sign-offs — safety, security, regulatory, coherence, release readiness — each with its evidence attached. Nothing signs itself, and nothing is chased.

Long after it ships

It doesn't forget that features have to be documented and supported

Written while it was fresh: one for the tester, one for the engineer who keeps it running, one for the person who has to sell it.

The steps

One piece of work, all the way through.

Nothing is handed on, nothing is written twice, and everyone still in the loop is drawn on the same line. Opening a piece of work shows the actual material — the sketch somebody drew, the plan the build wrote, the screenshot a tester posted, the sign-off somebody put their name to — in the order it happened.

  • It holds the material, not a pointer to itThe sketch, the plan, the screenshot, the sign-off — the real things, in the order they happened, not an attachment box somebody forgot to fill.
  • It never splits into jobs for peopleOne piece of work has a step, an owner for that step, and two clocks. How the building gets broken up is the build's business, and it is visible to everyone.
  • Clocks, not estimatesTwo clocks per piece of work, and a color when one is about to blow. Nothing to estimate, nothing to re-estimate, nothing to chase.
  • It survives the thing it builtWhen the work ships, what it learned stays attached to the code it touched — so the next person to open that code is not starting cold.

Modeled screen · example data · no customer data appears anywhere.

Memory

No session starts as a new day.

The biggest tax on building with AI is forgetting. Every session opens cold, every project is a new world, and what the last one learned is gone. Riverbank keeps it — and hands it back at the start of the next session, without anybody having to remember to.

Every session leaves a note

What was tried, what was decided, what changed, what was left — written at the end of a session in seconds, by the build, for the next build.

Lessons become rules

When the same mistake shows up twice, the system says so and proposes the rule that would have caught it. A person ratifies it; then it rides along on every build after.

The house rules are served, not repeated

How your organization works, what it doesn't do, what it learned the hard way — handed to the AI at the start of every session, so nobody has to say it again.

Answers by meaning, not keyword

Ask what was decided about a piece of code and get the decision, who made it, when, and what it cost — with a standing instruction to check it against the live code before acting.

On purpose

What it doesn't have.

Every absence below is deliberate. A tool that offers a field for something is a tool that will grow that something back, so Riverbank doesn't offer the field.

  • No estimatesTwo clocks per piece of work, and nothing to argue about in a room.
  • No sprints or pointsWork moves when it is ready, not when a fortnight ends.
  • No queues or assignmentA piece of work has an owner for the step it is in. It never fragments into a list of jobs for named people.
  • No second reporting surfaceOne page shows every piece of work, its step and its clocks. There is no burndown, no velocity and no dashboard builder, and the system politely refuses to grow one.
  • No test-case managementA finding is a screenshot and a few sentences. Mechanical pass/fail belongs to the build now.

Riverbank gives you the tool to organize around AI-first development.

  • Transparent
  • Organized
  • Documented
  • Memorialized

And none of it is a second job — the record is what the work leaves behind.

Building with AI and losing the thread of it?

Tell us how your team works today and we'll show you what Riverbank would hold for you — and say plainly what it doesn't do yet. No obligation.

Ask us a question