Case study01 · OBSS

Estimation Agentfrom client brief to UX/UI estimate

An agent that turns RFPs and feature lists into UI/UX effort estimates, using past projects to generate proposal-ready Excel sheets. It cut the typical estimation turnaround from 4–5 working days to 1–2 days, while keeping every assumption visible and every estimate open to human review.

RoleStaff Product Designer at OBSS, designed and built end to end

TeamIndependent project

Duration2 days

UsersPre-sales, solutions consultants and design reviewers

1–2 daysestimation turnaround, down from 4–5 working days

3 → 1revision rounds per estimate

2 daysto design and build the agent

1 clickfrom reviewed estimate to a proposal-ready Excel file

Context

At OBSS, design effort is part of most client proposals. Turning a client RFP into a UI/UX estimate meant interpreting requirements, identifying potential flows and screens, and assigning separate UX and UI person-days, then assembling the results in Excel.

The process typically took 4–5 days, and estimates prepared by different people could vary in detail and structure. The output feeds the sales team's proposal preparation: a structured starting point, not a final price or commitment.

The challenge

01SlowA single estimate typically took 4–5 working days of interpreting requirements and building sheets by hand.

02InconsistentEstimates prepared by different people varied in detail and structure, and often needed several revision rounds.

03Must stay reviewableSpeed couldn't come at the cost of trust: assumptions had to stay visible and every estimate open to human review.

Problem statementHow might we make UI/UX estimation faster and more consistent, while keeping assumptions visible and every estimate open to human review?

Key decisions

Estimate by function, not by screen count

The agent breaks a brief into functions, lists the likely flows and screens under each, and estimates UX and UI separately per function. The unit of estimation matches how designers actually scope work.

Anchor every number to history

I used our past estimation sheets as the agent's reference base. Each function is matched to similar functions from past projects, and the row shows the reference it drew on: the average UX and UI days and how many projects that average is based on.

References library of historical function patterns — The references library: similar functions, how many past projects they come from, and their average UX/UI days.

Make uncertainty explicit

Every row carries its assumptions. When the brief leaves something open, such as approval rules or whether a feature is in scope, the row is flagged Assumption · Needs verification, and the summary counts how many functions still need an answer.

Show the work while it runs

Instead of a spinner, the analysis runs through five visible steps: scanning the document, extracting functions and requirements, mapping flows and screens, comparing against historical effort, and calculating the estimate.

Analysis steps in progress — Five named steps replace a generic spinner, so reviewers know what the agent is doing.

Keep people in control

The estimate is a draft. Every UX and UI value is editable, totals update as you go, and the Excel export is generated from the edited table with formulas that stay editable for sales.

Private by default

Client documents are sensitive. In the demo, uploaded files are processed only in the browser and never sent to a server, and every sample is clearly marked as fictional.

The solution

A focused three-step flow: upload or pick a document, watch the analysis, then edit and export. Supporting pages give the team the reference library and the people who use it.

Start from any brief

Step 1

Drop in an RFP, function list or requirements document up to 10 MB, or try one of three fictional sample briefs covering mobile banking, an HR self-service portal and a B2B ordering panel.

Estimation Agent start screen — Upload a document or start from a sample.

Review the estimate

Step 2

The result opens with the totals (function count, UX days, UI days, grand total) and a clear count of functions that need verification. Below, each function lists its flows, assumptions and the historical reference behind its numbers.

Estimate summary and function table — Summary totals, flags and a per-function breakdown.

Edit and export

Step 3

Adjust any value in place and download the table as an Excel file for the proposal, with formulas intact.

Editable estimate table with Excel export — Editable totals and one-click Excel export.

Built for the pre-sales team

Users

The people who turn RFPs into estimates, including presales leads, solutions consultants, UX and UI designers and design ops, work from the same tool and the same reference library.

Demo Prototype

Interactive

Try it yourself: pick one of the fictional sample briefs, watch the analysis run, then edit the estimate and export it to Excel.

Interactive demo prototype. Demo data is fictional. Open in a new tab ↗

Impact

The agent reduced the typical UI/UX estimation turnaround from 4–5 working days to 1–2 days. By turning incoming client documents into a consistent breakdown of features with separate UX and UI person-day estimates, it gave the sales team a clearer starting point for proposals.

1–2 daysestimation turnaround, down from 4–5 working days

3 → 1revision rounds per estimate

Same formatevery estimate follows one structure, whoever runs it

Estimates remain open to human review.

Takeaways

Explainability is the feature

People act on numbers they can trace. Showing the reference behind each estimate mattered more than making the estimate itself cleverer.

Design the uncertainty

Flagging what a brief leaves out makes the estimate more honest and gives sales concrete questions to take back to the client.

Designers can build the tool

Designing and building it myself meant iterating on the real product instead of on mockups of it, and getting from idea to a working tool in two days.

Next case · 02
Super App Ecosystem

helloonuroral@gmail.com LinkedIn

Updated last, 2026.