Every story on this blog is about some other founder's numbers.
This one is mine.
Last week Pat Walls had me and my co-founder Borja on Starter Story Build to walk through the weirdest thing we've done as founders: we decided to rebrand our startup with AI. We gave Claude 24 hours and agreed to actually do what it said.
The results:
→ MRR went from $7,000 to $11,000 a month.
→ Day-one churn dropped from 40% to 20%.
→ Trial-to-paid conversion went up 10 points.
One day of prompting. No agency, no rebrand consultant, no 6-week "brand discovery workshop".
Here's the full episode, and below it, the exact process so you can run it on your own product this weekend.
The setup: a stuck SaaS at $7K MRR
Borja spent five or six months building what was then called Rebel Growth, a SaaS that gets you customers from Google SEO and AI answers. He grew it to $7K MRR completely solo, mostly with ads and by using the product on itself.
Then it flattened.
Monthly churn was 31%. Day-one churn (people who cancel the same day they pay) was 40%. Growth was a leaky bucket with a hose taped to the side.
That's when we matched as co-founders. And instead of arguing about what to fix based on our opinions, we tried something else: dump everything about the business into an AI agent and let the data decide.
Borja said it best on the episode: your ego and your belief system contaminate your judgment. You built the thing, so you can't see it clearly. Claude has no ego about your product.
Step 1: Give the agent everything
An AI audit is only as good as its context. Most people paste their landing page into ChatGPT, get generic advice, and conclude AI can't help them.
We did the opposite. We gave the agent:
→ Our Stripe data, through a read-only API key. Read-only matters: the agent can see every invoice, subscription, and refund, but can't touch anything.
→ Transcripts from customer calls.
→ Our support emails, the raw ones, including the angry ones.
→ Anything else with signal: onboarding drop-off notes, cancellation reasons, past experiments.
Thousands of data points instead of one screenshot of a homepage.
If you read my breakdown of how consultants run AI audits as a business, this is the same principle. The audit is 80% data collection, 20% prompting.
Step 2: Build a knowledge base (the step everyone skips)
Context tells the agent what your business IS. A knowledge base tells it what good looks like.
We built ours in three parts.
1. Feed it Paul Graham
The best SaaS advisor alive has published over 200 essays for free. I sent them all to the agent and had it store the lot as its advisory brain.
The payoff came later, when the audit quoted "Make Something People Want" back at us next to our own churn number: "This 31% monthly churn is screaming one thing: users are not getting what they want."
Getting roasted by Paul Graham's essays, applied to your own Stripe data, is a special kind of humbling.
2. Use Grok to spy on competitors
Everyone in SaaS is building in public on X. Which means Grok, with its access to X, knows your competitors' growth channels, the walls they hit, and the features they shipped.
One prompt gave us a report per competitor: what worked, what stalled, what they built. Years of their trial and error, summarized, so we could skip the errors.
3. Send an agent to mystery-shop them
This is my favorite one. Give an agent a Gmail account and your competitors' URLs, and send it through their free trials. Landing page, onboarding, every feature. It writes up the friction points in a Notion report.
I used to record Loom videos of myself clicking through competitor products. Now an agent does it while I sleep. (If you want to see how far this goes, read how Ryan Carson ships 25 PRs a day managing AI agents solo.)
Step 3: Ask for the audit and the 90-day plan
With context plus knowledge base loaded, we asked for the actual audit. Four prompts, in order:
→ A general read on the SaaS, through the Paul Graham lens.
→ A customer and churn analysis: who cancels, when, and why.
→ A mystery-shop of OUR OWN product. We gave the agent a coupon, forced it through signup, onboarding, and first use, and had it screenshot every friction point.
→ A comparison against every competitor, with a plan to beat them.
Then the final prompt: "Give me a 90-day action plan to implement all of this, quick wins first."
The $9 mistake
The churn analysis found the killer. Out of our last 100 paid invoices, 31 were first-month payments. Only 13 of those people ever renewed.
Why? We were charging $9 for the first month, then $97 after.
Sounds clever. It wasn't. People who join at $9 decide your product is worth $9. When month two hits at $97, they feel scammed and leave. We weren't discounting the first month, we were training customers to undervalue us by 10x.
The fix was almost insulting in its simplicity: charge $97 from day one.
Fewer signups, sure. But the people who pay $97 upfront are serious, they actually use the product, and they stay. That single change is most of why day-one churn fell from 40% to 20%.
The other part was tightening what happens after signup. A lot of "churn" was really trials going quiet because nobody followed up with them properly. If that sounds familiar, this breakdown of lead nurturing strategies covers the follow-up sequences we borrowed from.
The rebrand: Rebel Growth becomes Distribb
The plan also said the brand and the landing page were working against us. So on March 12 we relaunched as Distribb.
The landing page changes, straight from the audit:
→ Hero section: one paragraph instead of a sub-headline plus bullet points trying to answer every objection at once. One CTA. A video instead of a static image. Customer logos underneath.
→ A three-step "how it works" section instead of a wall of text.
→ Features reframed as results. I had been listing what the tool does instead of what you get.
→ 20 feature cards cut down to the few that matter.
→ A section answering the #1 fear from our support emails (losing control over content quality), with real writing examples.
→ Case studies, testimonials, a simple pricing table, an expectations section.
→ An FAQ written for the questions AI assistants actually research, so when someone asks ChatGPT about us, there's a clean answer to land on.
Nothing on that list is revolutionary. Marketers have preached most of it for decades. The difference is the agent tied every change to OUR data: this section exists because of that support email, this card got cut because zero customers ever mentioned the feature. Generic advice never survives contact with your product. This wasn't generic.
The results, with real numbers
Since the March 12 relaunch:
→ $7,000/month → $11,000/month MRR.
→ Day-one churn: 40% → 20%.
→ Trial-to-paid conversion: up 10 points, at 10x the price.
Read that last one again. We raised the first-month price from $9 to $97 and MORE trials converted. Every instinct I had said cheaper converts better. I would have argued against the change in any brainstorm.
Which is exactly why this process works. It surfaced a move we would never have made on our own, because it looked wrong from the inside.
Steal this: the 24-hour AI rebrand checklist
Run this on your own SaaS this weekend:
1. Connect a read-only Stripe key and dump in your call transcripts and support emails.
2. Load a knowledge base: Paul Graham's essays, Grok reports on your competitors, an agent's mystery-shop notes on their trials.
3. Prompt the audit: general review, churn analysis, mystery-shop of your own product, competitor comparison.
4. Ask for a 90-day action plan, quick wins first.
5. Actually implement it. This is where most people fail. The plan only counts if you ship it.
And when the audit tells you something that offends you, that's usually the thing to fix first.
FAQ
Can AI really rebrand a startup?
Yes, if you feed it real data. An AI agent with read-only Stripe access, support emails, and call transcripts found the pricing mistake and churn cause we'd missed for months. It won't design your logo taste for you, but it will tell you what's broken and in what order to fix it. We went from $7K to $11K MRR off its 90-day plan.
How long does an AI business audit take?
Ours took about 24 hours end to end: a few hours collecting data (Stripe, emails, transcripts), a few hours building the knowledge base, then the audit prompts themselves. Implementing the 90-day plan took longer, but the diagnosis and roadmap fit in one focused day.
Is it safe to give an AI agent your Stripe data?
Use a read-only restricted API key. The agent can analyze every invoice and subscription but can't move money, issue refunds, or change anything. Never hand an agent a full-access key.
Why did raising prices reduce churn?
Price anchors perceived value. At $9 for the first month, customers decided the product was worth $9 and cancelled when the real $97 price hit. Charging $97 from day one filtered for serious users: our day-one churn fell from 40% to 20% and trial-to-paid conversion went UP 10 points.
Want more stories like this?
I interview bootstrapped SaaS founders making $100K to $10M a year on the Profitable Founder Podcast, and I share my own numbers, wins and faceplants included.
