Design engineering for AI and B2B SaaS

You can build anything now.So why do users still get stuck?

We design one flow and ship it to your repo as code, without pulling your developers off the roadmap. $4,500 fixed.

01The problem

Most B2B products work. They just feel like a database with buttons.

Usually nothing is missing. People stall on one screen: the setup nobody finishes, the approval that waits, the upgrade nobody finds.

That screen is where we start.

02Selected work

Work we’re proud of.
One hard call in each.

Products we designed and prototyped. In each, someone has to make a call under time pressure, and the screen shows them everything except what to do. Every frame is recorded from the designs, and every number comes from the product’s own scenario.

Brace
Built with Prodular UILegal AIB2BWeb and mobile

Designed the contract review agent that finds the one clause that takes the liability cap away.

A customer's 52-page contract lands three days before quarter end. Brace reads it against the playbook in four minutes, redlines it in approved language and cites a clause for every change, and legal, finance and sales each decide their part.

6 days4 h

from the customer's paper to the first redline, with 35 of 38 quarter-end deals signed in the quarter.

03From the founder

Most products we see do not have a design problem. They have a decision problem.

The data is there, the feature is there, and the screen still makes someone hesitate. Prodular exists to fix that, one workflow at a time, on a clock short enough that nobody can hide in process. We design the flow and build it ourselves, so nothing is lost between a design file and the code. We lead every Sprint ourselves, and bring in senior specialists when a product needs them.

04How the week works

Five days.
Discovery takes a day, not a month.

Fixed scope, fixed price, fixed clock. You know before Monday what lands on Friday, and if Monday shows it is the wrong fix, you can stop there for $500.

  1. 01 · Monday

    Diagnosis

    The exact screen people drop off on, walked with you and confirmed.

  2. 02 · Tuesday

    Structure

    The flow rebuilt around the action it exists for, mapped end to end.

  3. 03 · Wednesday

    Design

    Every screen at full fidelity, with the empty, loading and error states.

  4. 04 · Thursday

    Build

    The new flow running in your codebase, on a preview link you can click through.

  5. 05 · Friday

    Ship

    A pull request your team can review and merge, with the reasoning behind it.

05Two ways to start

One week, fixed.
Or month to month.

One week · Fixed scope, fixed date

5-day Sprint: design + development

$4,500one week, one workflow

Pick the flow people give up on: onboarding, the trial, the upgrade, the approval. On Friday it is a pull request your team can review and merge.

  • Five working daysScoped in writing first
  • A pull requestReact front end, in your repo
  • A preview linkYour team can click through it
  • One round of changesThen five days of bug fixes
  • A prioritised listOf what we left alone

We agree the scope in writing, then payment holds your week. If Monday’s diagnosis shows this is the wrong fix, stop there: you keep the diagnosis and we refund everything above $500. Cancel more than five business days ahead for a full refund, and if we fail to deliver, you get a full refund too.

Embedded · Month to month

Embedded: design + development

from$9,999/ month

Designed and built by the same senior person every week, shipped to your repo. Not a request queue.

  • The same senior personEvery week
  • Frontend in your codebaseShipped to your repo
  • Backend workAgreed per feature
  • Thirty days noticeNo lock-in

Our longest engagement ran 30+ months.

07Fit

The Sprint is narrow,
on purpose.

One workflow, for one kind of company. Naming who that is not saves both of us a week.

Book this if

  1. 01

    You have a live product and real users.

    You can point at the screen where they drop off.

  2. 02

    It is AI or B2B SaaS.

    And the workflow carries a decision someone has to get right.

  3. 03

    Someone on your side can merge.

    A developer who reviews the pull request, ideally on a React front end.

  4. 04

    One or two people decide.

    You can agree a scope on a call and it stays agreed.

  5. 05

    You are buying judgment, not hands.

    What you pay for is the calls we make, not the hours we log.

Not a fit if

  • You are pre-product. There is no workflow to diagnose yet. Launchpad builds the first version instead.
  • You need a brand, a site or a deck. All real work. None of it ours.
  • You need volume. A queue of requests is a hire.
  • It has to clear six approvals. Nothing sharp survives consensus.

08Questions

Before you book.

  • Jacob, the founder, leads every Sprint and brings in senior specialists when a product needs them. No account-manager layer, and no juniors learning on your product.

09Next step

Name the screen where people give up.

For example

Twenty minutes, no deck. If it is not a fit, we will say so on the call.

Also from Prodular

Prodular UI
Our own design system.

Components, full screens and a Figma library that match token for token, made for SaaS products that ship AI.

  • Ten themesEvery text pair passes WCAG AA, light and dark
  • AI componentsComposer, reasoning, tool calls, sources, approvals
  • For coding agentsRules for every component, an llms.txt and an MCP server
  • Figma and codeThe same tokens, ten theme modes and three densities