I lost $17,000 last year doing free work for clients. So I built an AI to stop it. (Pre-launch, landing page live)

Hey IH 👋

Long post warning. But I think a lot of people here will relate to this.

I've been freelancing for 4 years — web design and branding mostly. I'm not bad at it. I have good clients, decent rates, repeat business. By most measures, things are going well.

But at the end of last year I did something I'd been putting off for a while.

I opened my time tracker and actually compared billed hours to real hours across every project.

The gap was $17,000.

Not money a client refused to pay. Not an invoice dispute. Money I never even asked for — because I did the work without ever flagging that it was outside the original scope.

Here's how it happens, if you've never tracked it yourself:

Client hires you. You agree on a scope. 5-page website, 2 revision rounds, delivered in 3 weeks. Everything is clear.

Then week 2 hits.

"Hey can you also add a contact form with file upload?"

"Can we make a quick animation on the hero section?"

"Actually we need a blog. Just a simple one."

"Can you add a newsletter signup?"

"Oh and we'll need social media banners for launch."

Each request sounds small in isolation. And in the moment — under the social pressure of a client conversation — you say yes. Because saying no feels difficult. Because you can't quickly find the exact clause in the contract that covers this. Because you don't want to seem like the difficult one. Because the relationship matters more than the $200 you're about to give away for free.

So you do it.

And then you do the next one. And the one after that.

By the end of the project you've done 60% more work than agreed and you're resentful of a client who, honestly, probably had no idea any of this was extra. Because you never told them.

I sat with that $17,000 number for a while and tried to figure out what the actual root cause was.

It wasn't bad contracts. I have detailed contracts.

It wasn't bad clients. Most of mine are decent people.

It wasn't my inability to say no. I know what's out of scope.

It was that I had no system to say it — professionally, quickly, without it turning into an awkward negotiation every single time.

What I'm Building

I've been working on something called GARDIAN for the past couple of months.

The core idea is simple:

You define your project scope once when you start — deliverables, what's included, and critically, what's explicitly NOT included. GARDIAN stores this as your locked scope.

Every time a client sends a new request, you paste it in. The AI checks it against the locked scope and gives you a verdict in seconds:

🟢 IN SCOPE — this is covered, no action needed

🔴 OUT OF SCOPE — this is new work, here's the relevant clause

🟡 BORDERLINE — genuinely ambiguous, here's both interpretations

If it's out of scope, one click generates a professional change order message — in whatever tone fits the relationship. Friendly, firm, or diplomatic.

The client gets a warm, clear message.

You get documentation of everything.

Nobody has to be the bad guy.

There's also a client portal — a link you send your client where they can see their agreed scope, submit requests, and respond to change orders. All in one place instead of scattered across email and WhatsApp.