A Bot Scored Us 91% Replaceable. Here's What It Got Right.

The Roast

Someone ran our website through deathbyclawd.com, a satirical tool that scores how replaceable your SaaS is by AI. We scored 91 out of 100. Basically dead on arrival.

The scanner's verdict: "Calmworks sells 'calm' but their investors should be panicking. This is literally a motivational poster with a Calendly link."

It generated a SKILL.md file, a system prompt that supposedly replaces our entire business. Total replacement cost: $0.003 per run.

The eulogy it wrote: "They didn't die. They were gently refactored into a system prompt."

We laughed. Then we took notes.

What It Got Right

The scanner saw our old homepage and read: generic pain statement, three process steps, three time-boxed services, greyed-out logos, and a booking button. That's it.

Fair criticism. If you describe what you do as "we diagnose, then we build, then we partner" without showing what that actually produced, you sound like every other consulting page on the internet. A well-written prompt can say the same thing.

The scanner scored us:

• Moat Depth: 96/100 (100 = no moat). It couldn't find anything unique.
• Markdown Replaceable: 94/100. It thought everything on the site could be a prompt.
• Secret AI Wrapper: 72/100. It thought we might just be putting a human face on AI.

And you know what? Looking at the old site through a bot's eyes, those scores make sense. We were describing process, not showing proof. The scanner had no way to know what we actually do.

What It Couldn't See

Here's where the $0.003 replacement falls apart.

A prompt didn't build Parliament's remote voting system in 24 hours. When COVID shut down the Finnish Parliament and 200 MPs needed to continue legislating, someone had to be in the room. Understand the procedural requirements. Design something that respected both democratic process and extreme time pressure. Ship it before the next morning. That was Vitali and a team of humans, not a system prompt.

A prompt didn't spend three years embedded at Visma Sign. Growing a product from regional tool to the #1 Nordic e-signature with 2M+ users takes understanding market dynamics across four countries, building design capability inside the team, and iterating on real usage data month after month. You can't paste that into a context window.

A prompt didn't redesign 60 consultants' operating model at Miltton. Cutting costs 40% while making delivery 45% faster requires sitting in rooms with people, understanding why they resist change, and building systems they'll actually use. Tarek did that over years, not tokens.

A prompt didn't improve TB detection 48% in Tanzania. Working with USAID in East Africa, processing 900,000+ samples, training healthcare workers to trust a new system, reducing costs 94%. That's fieldwork. That's relationships. That's showing up.

The scanner looked at our website and saw words. It couldn't see the 30+ years of combined experience, the 100+ workshops, the products used by 100,000+ people, the relationships with CFOs and COOs who pick up the phone when we call.

The Real Question

The scanner is asking something interesting under the satire: can your service survive AI?

Our answer: yes, because we build with it. We don't compete with AI. We use AI to compress weeks into days, to turn 40-page PDFs into interactive reports in 48 hours, to prototype faster than anyone expects. AI is in our delivery pipeline, not something we sell against.

The things that are replaceable by AI, we already replaced. The things that aren't, Vitali builds and Tarek manages adoption for. Two people, full lifecycle, no handoff gap.

A prompt can write a sprint plan. It can't sit across from a nervous CFO and know which question to ask. It can't build trust with 8,500 employees across a global manufacturing operation. It can't tell you that your "AI strategy" is actually a procurement problem, and here's the dashboard that proves it, running inside your existing tools, built last week.

What We Did About It

We rewrote the website.

The old homepage was generic because it described what we do instead of showing what we've done. Now it leads with proof. The numbers, the clients, the outcomes. If a scanner runs it again, it should choke on specifics it can't generate.

But honestly, the best response to "you're replaceable" isn't a better website. It's another client who got results they couldn't get from a prompt.

So if you're reading this and wondering whether to book a call or just ask ChatGPT: try the prompt first. Seriously. See how far it gets you. Then call us when you need something that actually ships.