Last month, a Series B fintech company offered us $200k to build their platform. We said no, and it was one of the hardest business decisions I've made at Autor. Here's the full story — the call, the spreadsheet I built at 11pm to justify it, and what it taught us about running a boutique AI development studio.
The pitch sounded perfect
The founder reached out through a warm intro. Smart guy, well-funded, clear vision. His company had raised $14M and needed a platform rebuilt from scratch — a customer-facing fintech tool with AI-powered risk assessment, automated document processing, and a dashboard for their ops team.
Timeline: 6 months. Budget: $200k. Our rate, our stack, our call on architecture.
For a studio our size — senior engineers only, typically running 2-3 concurrent projects — $200k over 6 months is a significant chunk of annual revenue. My co-founder and I spent a week scoping it. We had three calls with their team. We started sketching the architecture.
Then I built the spreadsheet that killed the deal.
The 85/15 problem
When I broke down the actual work, the numbers told a different story. Out of an estimated 2,400 engineering hours:
- ~360 hours (15%) was genuine AI work — the risk models, document processing pipeline, and the inference layer
- ~2,040 hours (85%) was standard web development — auth flows, admin panels, CRUD operations, payment integrations, responsive frontend, notification system, user management
We're an AI development studio. We charge $150/hr because our engineers have built production AI systems — voice agents handling thousands of calls, RAG pipelines processing medical documents, real-time inference serving healthcare providers 24/7. That's what clients pay a premium for.
Spending 85% of a 6-month engagement on work that any competent web dev shop could handle at $80-100/hr? That's not what we built Autor to do.
But it gets worse.
The opportunity cost spreadsheet
I opened a new tab in the spreadsheet and modeled what saying yes would actually cost us. Here's what I found:
With our current team size, taking this project meant turning away roughly 3-4 focused AI engagements over the same period. Based on our pipeline at the time, those projects looked like:
- A dental group wanting voice AI for 12 clinics (similar to our Loquent platform — we could deliver in 8 weeks, not 6 months)
- A healthtech startup needing a clinical document processing pipeline
- Two smaller AI agent builds, each in the $40-60k range
Total revenue from those projects: approximately $180-220k. So financially, it was roughly a wash. The $200k project wasn't even a revenue premium — it was just one big number that felt safer than four smaller ones.
But the real cost wasn't financial. It was positioning.
What happened last time
I've been here before. In 2024, we took a large project that was maybe 30% AI and 70% general platform work. The client was great, the money was good, and we told ourselves the AI component justified it.
Six months later, here's what actually happened:
Our dev.to posts dried up because nobody had time to write about AI work — we were debugging CSS grid issues and Stripe webhook edge cases. Our case studies from that period were weak because the AI component shipped in month 5 and we couldn't talk about the other client's business.
Three prospects during that period asked what we'd been working on. "A fintech platform" doesn't land the same as "We built a voice AI system that handles 3,000 automated calls per month for healthcare providers." Two of those prospects went with other studios.
The worst part: when the project ended, our inbound pipeline had cooled. It took us almost two months to rebuild momentum. The total cost of that positioning drift was probably $150k in delayed or lost deals. I can't prove that number exactly, but I can tell you our Q1 2025 was our weakest quarter since founding, and it traced directly back to that decision.
The conversation with the founder
Telling the founder was uncomfortable. He'd invested time with us. The warm intro was a friend I respect. And his reaction was exactly what I expected: "Can't you just staff up for this?"
I explained our model. We don't subcontract. We don't hire junior devs and mark them up. Every engineer at Autor has built production AI systems. That's not a recruiting pitch — it's a structural decision that limits our capacity on purpose.
He pushed back: "So hire another senior engineer."
Here's the thing about hiring senior AI engineers in Toronto in 2026: it takes 3-4 months to find someone good, another month to onboard them into our systems and standards, and the project needed to start in 3 weeks. Even if I found someone tomorrow, they wouldn't be productive on this codebase for 6 weeks. The math doesn't work.
We offered an alternative: we'd build the AI components — the risk models, the document processing pipeline, the inference infrastructure — as a standalone engagement. Eight weeks, $65k. They could hire a separate web dev team for the platform work at a lower rate, and we'd architect the integration points.
He thought about it for a week and went with a full-service agency instead. No hard feelings. It was the right call for his timeline.
What we did with those 6 months instead
Since turning down that project in June, here's what shipped:
We onboarded two new Loquent clients — a multi-location dental group in Ontario and a physiotherapy chain in BC. Combined, they're running about 1,800 automated calls per month through our platform. The dental group went live in 6 weeks.
We built a clinical intake AI agent for a healthtech company that processes patient forms in three languages. That project was 10 weeks and every hour was genuine AI engineering — multilingual NLP, structured data extraction from messy PDFs, PHIPA-compliant data handling.
We published four technical articles (including this one) that generated inbound leads. Our dev.to posts collectively pulled 8,000+ views. One of them — the healthcare voice AI analysis piece — got picked up by two AI newsletters.
Total revenue from these projects is tracking ahead of the $200k we turned down. More importantly, every project reinforced our positioning. Every case study we can write from this period is an AI story, not a "we also do general web dev" story.
The 5 things I'd tell any boutique AI studio about saying no
Model the real work breakdown before you say yes. If more than 40% of the hours are outside your core expertise, you're not being hired for what makes you valuable. You're being hired because you're available. That's a commodity position, and it will price you like one eventually.
Calculate opportunity cost in positioning, not just revenue. Lost positioning compounds. A quarter spent on off-brand work doesn't just cost you that quarter — it weakens your pipeline for the next two. I've lived this twice now and the pattern is clear.
"Just hire more people" is not a strategy for boutique studios. Our value proposition is that every engineer is senior and has shipped production AI. The moment we dilute that to service a larger project, we become a body shop with a nicer website. Clients who want 15 developers can find them. Clients who want 4 engineers who've each built voice AI systems handling thousands of calls? That's a shorter list.
Offer the scoped alternative. Don't just say no — propose the engagement that fits your strengths. We lost this particular deal, but the "we'll build the AI layer, you hire a web team" pitch has landed with other clients. It respects their needs while protecting your positioning.
Track what you ship instead. After saying no to a big project, I now keep a running doc of what we delivered in that same window. It's the best antidote to the "what if we'd taken it" doubt. Six months later, the evidence is always overwhelming.
The uncomfortable truth
Turning down $200k when you're a small studio with real payroll is genuinely scary. There's no venture fund backstopping us. Every dollar matters. And I won't pretend the decision was easy or that I didn't second-guess it for weeks.
But every time we've said yes to a large project that diluted our focus, we've paid for it in ways that don't show up on the invoice. Weaker case studies, slower content output, cooled inbound pipeline, and that creeping feeling that we're becoming generalists who happen to know some AI — instead of AI specialists who happen to be excellent engineers.
Boutique doesn't mean small. It means focused. And focus requires saying no to good money when it pulls you off course.
If you're building something similar, we'd love to hear about it. Reach out at hello@autor.ca or visit autor.ca.
Top comments (0)