The Survey
Between October and December 2025, I surveyed 50 early-stage startups (pre-seed through Series A) about their AI tooling. The companies ranged from 2 to 35 employees, operated across 12 countries, and covered sectors including fintech, healthtech, developer tools, and consumer apps. I asked about every AI tool they use, how much they pay, how often they use it, and what they'd cut if they had to reduce their AI spend by 50%.
The results challenged several assumptions I held going in.
What They're Actually Using
The most common AI tools, by adoption rate across the 50 companies:
- ChatGPT (GPT-4o): 94% — the near-universal default for general AI tasks
- GitHub Copilot or Cursor: 88% — essentially table stakes for any technical team
- Claude: 72% — used alongside ChatGPT, not instead of it
- Notion AI: 54% — high among teams already using Notion
- Midjourney or DALL-E: 48% — primarily for marketing assets
- Perplexity: 42% — for research and competitive intelligence
- Whisper (via API): 36% — for meeting transcription and voice features
- Eleven Labs: 28% — for product demos and marketing videos
The Surprising Findings
Three findings surprised me. First, the average AI spend was much lower than I expected: $340/month at pre-seed, $890/month at seed, $2,100/month at Series A. These are not the eye-watering AI bills that make headlines. Most startups are being quite disciplined.
Second, the "what would you cut" question revealed that ChatGPT and coding assistants are considered non-negotiable by virtually everyone. The products most likely to be cut were specialist tools with narrow use cases — image generators, voice tools, and category-specific AI features. The general-purpose tools are sticky; the specialists are expendable.
Third, there was almost no correlation between AI spend and perceived productivity gain. The companies spending $3,000/month on AI tools did not report meaningfully better outcomes than those spending $500/month. The difference was in how intentionally they used the products, not how many they had.
The Lean AI Stack
Based on the survey data and my advisory experience, here's the minimum viable AI stack for an early-stage startup:
- ChatGPT Team ($25/user/mo): General-purpose AI for the whole team
- Cursor Pro ($20/mo) or GitHub Copilot ($19/user/mo): For any technical work
- Perplexity Pro ($20/mo): For research and competitive intelligence
Total: roughly $65–$85/month per person. Everything else is optional and should be added only when you have a specific, measurable use case that the core stack doesn't cover.
The temptation to build a comprehensive AI stack early is real — the products are exciting and the FOMO is genuine. Resist it. The startups in my survey that were most productive with AI were the ones that had mastered a small number of tools deeply, not the ones with the longest list of subscriptions.
Share this article
