LeanDeck — Design Case Study

01 · Users and research

Who it's for, and what research showed

Who it's for

Agencies: Design and dev agencies that write proposals, agreements and invoices for their clients.
Agency teams: Work on the same documents in shared, multi-user workspaces.

What research showed

User interviews: I interviewed design and dev agencies about how they win work. A proposal took them 6–10 hours, mostly written by hand from scratch.
The real job: A proposal is only the start: once it is approved, it becomes an agreement and then an invoice.
The risk: Agreements and invoices are financially binding, so AI output had to be checkable and easy to correct.

Goals we designed against

Time to beta: A beta with real agencies in eight weeks.
First proposal: A first proposal in about ten minutes. Today a proposal or agreement takes about 5 minutes.
Output quality: AI output usable as-is, needing only light edits before it's sent.

02 · Structure

From the first map to today's navigation

I mapped the information architecture before the first wireframe, then grew it as agencies asked for more.

First information architecture

Dashboard
SearchFind anything
ClientsList, details, and the client's proposals, agreements and invoices
ProposalsCreate, view, edit, send, download, reminders, activity log
AgreementsSame actions
InvoicesSame actions
SettingsProfile, company, billings, upgrade plans, payment integration, reminder settings

Today's navigation (live app)

Dashboard
MainContacts, Proposals, Agreements
ResourcesTestimonials, Case Studies, Templates
EssentialsCompany profile, Reminder templates, Billings

Information architecture: the first map, and the live app's navigation today

User flow for creating a proposal: start from Clients or Proposals and select a client, then either pick a service template or start from scratch with an AI prompt, answer the questions, and the proposal is created; then preview and edit, and archive, download, set a reminder or send — User flow: creating a proposal, from a client or the proposal list, starting from a template or from scratch with AI, through the smart questions to review, then send, download, set a reminder or archive

03 · Design decisions

Where the AI helps, and where people decide

The problem behind each one, the decision, why, and what it cost.

Two ways to start, one set of questions

Create Proposal: the question — Start: pick an existing client or describe a new one

Smart questions: the AI asks what it needs, with one-tap service options. Sample answers blurred.

Problem: Agencies arrive in two situations: with a service they sell often, or with something new. A blank prompt box helps neither.
Decision: Start from a service template or from scratch with AI. Both paths go through the same short questions (who it's for, the project, the services), with suggested answers to tap.
Why: The AI gets the same structured input every time, so the draft is consistent, and nobody has to know how to write a prompt.
Trade-off: Fewer free-form inputs. The questions set what the AI can use.

AI returns structured, editable sections

Proposal editor with a table of contents, a proposal cover being edited, and team comments with the commenters' names and photos blurred — Proposal editor: table of contents, sections and team comments

Block settings beside a proposal cover: content, style and advanced tabs, with background fill, border, and typography settings such as font, size, weight and line height — Block settings: content, style and layout for each section, without a design tool

Problem: AI output had to be accurate and consistent every time, and a single block of generated text is hard to check and hard to fix.
Decision: The AI returns a proposal as structured, editable sections, with a section- and sentence-level editor that has AI assistance built in.
Why: Agencies can edit, reorder, restyle and price sections. A block of generated text is something they'd have to rewrite.
Trade-off: More prompt and evaluation work than free-text generation, slower generation, and less freedom for the AI, which has to fit a set section structure.

The AI assists; you decide

Problem: Financially binding documents need to be reviewable, and a plausible-but-wrong invoice is worse than a slower one.
Decision: Users can accept, edit or override any AI output without losing context. Generative AI drafts proposals and agreements; invoicing follows fixed rules.
Why: The product says it plainly: “The AI assists; you decide.”
Trade-off: Less automation where it would have been easiest to add.

Follow-ups built into sending

Send dialog on the Reminder Settings tab: a reminder switched on, with timing, repeat, subject and an optional custom message, and a button to add another reminder; names in the background blurred — Send: reminders set at the moment of sending, with timing, repeat and a custom message. Names blurred.

Problem: After a proposal is sent, following up is manual and easy to forget.
Decision: Reminders sit inside the Send step, and every proposal, agreement and invoice can carry its own, with reusable reminder templates.
Why: The follow-up is decided while the document is still in mind, not remembered days later.
Trade-off: One more thing to decide while sending. It is optional.

A short sign-up; company details where they matter

Live app: a prompt to complete the company profile, saying why it matters

Problem: The AI needs the agency's company details to write accurate proposals, but a long sign-up form loses people before they see the product.
Decision: Sign up with Google, Apple or a name and email. The company profile is asked for afterwards, on the dashboard, with the reason given: accurate company information in every proposal and agreement.
Why: People reach the product first, and they give their company details when they can see what those details are for.
Trade-off: Until the profile is complete, drafts have less company information in them.

04 · The dashboard

Where every deal stands

Dashboard with sample data: total revenue, active proposals, pending agreements, win rate and average deal size; a revenue trend; the pipeline by stage from sent to invoiced; hot leads with how often each proposal was viewed; and recent activity — Dashboard: revenue, active proposals, pending agreements and win rate; the pipeline from sent to invoiced; hot leads by how often a proposal was viewed. Sample data.

The pipeline runs from sent to viewed, negotiating, agreement sent, signed and invoiced, so an agency can see where deals stall, and the leads whose proposals are being read most.

05 · Templates, case studies and testimonials

More designs, and proof agencies reuse

After launch, agencies asked for more design options. The answer was a template library, and a place for the proof they put in proposals: case studies and testimonials.

Templates: A library of premade proposal designs: 126 template pages and 18 proposal covers, with cover, case study, process, scope of work, estimation and testimonial pages.
Resources: Testimonials, Case Studies and Templates now sit together under Resources in the navigation, so an agency adds its proof once and works with it in every proposal.
From conversation: Paste a client conversation, and the proposal, agreement and invoice are drafted from it. Invoice amounts still come from the work and prices agreed in the proposal.

Four premade template pages side by side: a proposal cover, a success stories page with sample figures, a solution steps page, and a testimonials page, with the contact details and sample text blurred — Premade template pages: cover, success stories, process and testimonials. Sample content; contact details and sample text blurred.

06 · Brand and design system

One identity, from logo to components

Logo: A geometric “L” with a diagonal for momentum, a base for stability and a spark for the AI.
Colour: Teal green (#236458) as the primary colour, deep slate (#121414) and white.
Type and icons: Inter Tight for type, and Hugeicons as the icon set.
Components: The web app is built on Ant Design 5, themed with the brand colour and type, so every screen shares the same parts.
Beyond the product: The brand guideline also covers favicons and social media covers. I also made the pitch deck and the product positioning.

Brand guideline: the logomark on light and teal

Brand guideline: the colour palette

07 · Testing and changes

From beta to what agencies asked for

Beta: The beta launched in early January 2026 with 6 agencies, 8 weeks after kickoff on 9 November 2025.
AI quality: I wrote and tested the prompts and defined the criteria for judging AI output, so changes were checked against the same standard.

After launch: what we changed

Feedback: Agencies asked for more design options, and to paste a client conversation and get their documents from it.
Change: More proposal and agreement designs, and conversation-to-documents: paste the client conversation, and the proposal, agreement and invoice are drafted from it. Invoice amounts still come from the work and prices agreed in the proposal.
Why it mattered: It removed the last manual step: retyping what the client already said.

Results

~5 min

To create a proposal or agreement (6–10 hours by hand)

8wk

From kickoff to beta

6

Agencies in the beta

500+

Design and dev businesses using it

3 → 1

Proposal, agreement and invoice in one connected workflow

Customer count as published on leandeck.ai.

The design in numbers

100+

Screens and states designed

126

Premade template pages

18

Proposal cover designs

2

User flows mapped (proposal and agreement)

7 → 10

Navigation areas, first map to today

Counted from the Figma files and the information architecture.

08 · What I'd do differently

Build the AI test set on day one

I'd build a fixed set of test proposals in week one and score every prompt change and every editor change against it, instead of judging AI output case by case.