From a LinkedIn PDF and a Logo to a Working Agency: What Building Basirah-AI With Claude Actually Looked Like
A few weeks ago, everything I had for Basirah-AI was two files: a PDF export of my LinkedIn profile, and a logo my designer had put together. No website. No business plan. No leads. No idea, honestly, which direction the agency should even point in.
I want to walk through what happened between that starting point and where things stand now — a live site, a full set of business documents, a working leads database, and the early shape of an operations "command center" — not to sell anyone on Basirah-AI, but because the process itself is a useful answer to a question I get asked a lot: what can you actually do with Claude, beyond writing emails and summarizing documents?
Starting from almost nothing
The brief I gave Claude was thin on purpose. I knew I wanted to run a service-based agency — no physical location — built around GoHighLevel, a CRM and marketing automation platform. Beyond that, I hadn't picked a niche, a market, or even a name for the offer. I dropped in my LinkedIn resume and the logo and asked Claude to help me figure out the rest.
The first thing worth noting is what Claude didn't do: it didn't take my resume at face value and run with it. An early draft of the business plan, built off the resume, implied I'd been working hands-on with GoHighLevel since mid-2025. That wasn't accurate — my actual background was a stretch of self-directed learning in data science, machine learning, cloud computing, and cybersecurity, with AI exploration starting around September 2025, and GoHighLevel specifically only entering the picture in August 2026. I flagged the discrepancy, and every document got corrected to reflect the real timeline. No inflated expertise, no invented track record.
That correction ended up shaping how I think about working this way: Claude is very good at producing something coherent quickly, but the coherence is only as trustworthy as the facts it's fed and the review it gets. Treat it like a capable collaborator who needs its work checked, not an oracle, and the output holds up.
Turning a vague idea into a positioned business
Once the facts were straight, we worked through the actual business design together. Rather than me guessing at a niche, Claude laid out options and reasoning, and we landed on a staged approach: chiropractic and physical therapy clinics as the beachhead market, dental practices next, medspas and aesthetics as a premium tier later. The core offer became a "cadence system" built in GoHighLevel — sold as a one-time setup fee plus a monthly retainer, benchmarked against realistic market pricing rather than numbers pulled from nowhere.
From there, Claude produced the actual paperwork a real agency needs to operate: a business plan, an offer and pricing sheet, a 30/60/90-day client acquisition plan, outreach scripts for both warm and cold channels, a sales SOP, and a service agreement template. These weren't generic placeholders — the cold email sequence, for example, went through a specific accuracy pass where an unsourced retention statistic got replaced with a properly attributed Bain & Company / Frederick Reichheld figure, and a CAN-SPAM-compliant footer was added before anything was called "ready."
We also picked a market. Claude recommended the Charlotte, NC metro area, and we built out a real lead list — 100 actual chiropractic and PT clinics across Charlotte and its suburbs, not sample data.
Building the site itself
With the positioning settled, Claude built a single-page marketing site for basirah-ai.com — hero, services, about, and contact sections. The homepage leads with the clinic-focused Cadence System offer, with a "Beyond the Clinic" section underneath for the broader SQL Server, consulting, and applied-AI work. The visual identity — a monochrome palette with a single teal accent — was pulled directly from the logo I'd started with, so the site actually looked like it belonged to the brand rather than a generic template.
Getting it live meant real infrastructure work, not just a static export: deploying the site to Cloudflare, migrating the domain's nameservers over, and splitting the domain deliberately so the root and www serve the new site while an existing blog stays put on a subdomain — all without breaking the email routing already configured on the domain. That's the kind of change where one wrong record breaks something, and it didn't.
The contact section went through a few honest iterations, too. It launched with a placeholder "Book a Call" link because there was no real booking tool yet. Once I decided to use Google Calendar as a free stand-in — my GoHighLevel trial had expired, so its calendar wasn't usable yet — that placeholder became a working link, and the contact details got updated to the business email and number actually meant for clients.
From spreadsheet to database
The lead list started life in a Google Sheet, which is where most people would leave it. Instead, we moved it into Supabase — a hosted Postgres database — so the 100 Charlotte-area leads live in an actual leads table rather than a spreadsheet tab. Alongside it sit two more tables, projects and tasks, the early skeleton of what's becoming a proper operations command center for running the agency: not just where leads are stored, but where the work of following up on them, tracking deals, and managing delivery will eventually live.
That's a meaningfully different thing from "AI wrote me a webpage." It's AI helping stand up the actual data layer a small operation needs, with room to grow into it.
What this is actually demonstrating
None of this required me to know how to write SQL, configure DNS, or structure a sales process from scratch. What it required was being specific about what I wanted, supplying the real material I had (even if it was just two files), and staying in the loop to correct things when they drifted from the truth.
That's the honest version of "what Claude can do": not a magic business-in-a-box, but a collaborator that can move from a vague idea to real infrastructure — documents, a live website, a working database — as long as a person is still checking the facts and making the calls. If you're sitting on your own vague idea and a couple of source files, that gap is smaller than it looks.

Comments
Post a Comment