Case studies / BuildPersona
AI Content PlatformPartner product · LinkedInSep 2026

BuildPersona

LinkedIn profiles and posts for executives, written from their own voice and approved before anything goes out.

Built for
Executives, founders and ghostwriting agencies
Platform
Web app, WhatsApp idea capture, works from Claude
Status
Early access
The public BuildPersona landing page, with a waitlist form and a preview of the voice interview.
20 daysto a working staging app
9profile rubric categories
20voice interview questions
$49Executive plan per month
Overview

The short version.

BuildPersona helps executives show up on LinkedIn without sounding like everyone else's AI posts. It scores their profile, rewrites the parts that are holding it back, then interviews them about their experience and opinions. Posts, carousels and articles are drafted only from that approved voice, and nothing is published until a named person approves the exact version.

It's built in partnership with a marketing agency and has two editions: one for individual executives and one for ghostwriting agencies managing many clients.

The challenge

What was broken.

Executives know LinkedIn matters and don't have time for it. Ghostwritten posts read like generic AI content, and nobody can say whether a profile is actually working. Agencies managing a dozen executives were running approvals through WhatsApp threads and spreadsheets.

  • Generic AI posts damage credibility instead of building it.
  • No objective way to tell what's wrong with a profile.
  • Approvals lost in chat threads, and posts published without sign-off.
How it came together

From idea to production.

  1. 7 Sep

    Start

    Profile scoring, voice interview and content drafting designed and built.

  2. 11 Sep

    Partnership

    Product and revenue partnership signed with a marketing agency.

  3. 13 Sep

    Rebuild

    A focused rebuild with agency roles, approvals and scheduling.

  4. 24–26 Sep

    Writing quality

    Mandatory editorial review and 163 writing-quality checks, then staging for early access.

The solution

What we built.

BuildPersona starts from what the executive actually thinks. A nine-category rubric scores the LinkedIn profile, with evidence for each score. It suggests three headlines and rewrites the About and Experience sections. A 20-question interview, typed or spoken, becomes an approved voice brief. Only then can the Studio draft posts, image posts, PDF carousels and articles, each showing its sources. Facts the AI can't confirm stay as placeholders that block publishing until someone fills them in.

How it works
  1. Step 1Profile scanned and scored
  2. Step 2Voice interview recorded
  3. Step 3Voice brief approved
  4. Step 4Studio drafts in their voice
  5. Step 5Approved version scheduled
Key features

What we delivered.

Profile score

Nine categories scored with evidence. Anything the AI can't see is marked not assessed, never guessed.

Profile rewrites

Three headline options, About and Experience rewrites, and a keyword map.

Voice interview

20 questions in five groups, typed or recorded.

Studio

Text posts, image posts, PDF carousels and long-form articles.

Approval by version

Approval belongs to one exact version. Edit it afterwards and it needs approving again.

Idea capture on WhatsApp

Send a thought to a WhatsApp number and it lands in your private idea inbox.

Agency mode

Owners, editors and client approvers, from 1 to 50 seats.

Works from Claude

Draft from Claude Desktop or Claude Code. The assistant can never publish or approve.

Inside the product

A closer look.

Where the AI does the work

Intelligent, and trustworthy.

AI grades, code scores

The AI assigns rubric levels with evidence. The platform checks the evidence and does the scoring maths itself.

No invented facts

Unconfirmed facts stay as visible placeholders that block publishing.

Private stays private

Private interview answers and unreviewed transcripts are never used to write content.

The right model for each job

Drafting, news relevance, the assistant and image creation each use the model best suited to them.

Built with

The stack.

Platform
Next.js on Cloudflare Workers, D1, R2
AI
Gemini, DeepSeek and image models via OpenRouter; Tavily, Brave and Exa for research
Integrations
LinkedIn sign-in and publishing, WhatsApp, Apify, Dodo Payments, Resend
Assistants
MCP server for Claude Desktop and Claude Code
Results

The numbers.

Build proof

20 days
idea to working staging app
Verified
46
database migrations
Verified
119
automated test files
Verified
163
writing-quality checks
Verified

Expected impact for users

15 min
executive time per week for three posts
Projected
+48
profile score points after rewrites
Projected
4×
impressions against a ghostwritten baseline
Projected
5 → 1
approval steps for agencies
Projected

Verified from project repositories, test runs and release records. Projected modelled estimates of user impact, not measured client results.

Your product next

Got an idea? Let's ship it.

Tell us the workflow that's costing you customers or hours. We'll come back with a prototype plan, a timeline in weeks, and a fixed quote.