Investor Relations
We don't need your money.
But we might want your partnership.
Airlock is bootstrapped, profitable-by-design, and building the operating system for enterprise data operations. We're not raising out of desperation. We're looking for partners who understand what we're building and want to accelerate it.
A note on how we think about capital
Most AI startups burn venture capital to subsidize inference costs, pray for scale, and figure out unit economics later. Our unit economics already work. We ran 18,000+ LLM calls across 17 personas and 15 models and found that enterprise-grade multi-agent deliberation costs $0.0036 per run. Our Pro tier has 91% gross margin. Our BYOK tier is 100% margin.
We're not looking for money to keep the lights on. We're looking for capital that accelerates distribution — enterprise sales, compliance certifications (SOC 2, FedRAMP), and ecosystem partnerships.
If your term sheet includes 2x liquidation preferences, full-ratchet anti-dilution, or board seats that outnumber founders — thanks, but we're good. We'd rather grow slower and own what we build.
The numbers speak for themselves
ConstellationBench + Sovereign Triads — 18,000+ LLM calls across 17 personas, 15 models, and 3 adversarial stress layers.
Unit economics that actually work
Every number below comes from real ConstellationBench data. No projections. No “at scale” hand-waving.
| Tier | Price | COGS / Council | Monthly COGS | Gross Margin |
|---|---|---|---|---|
| Free Gemini 2.5 Flash | $0 | $0.0036 | $0.11 | Subsidized Acquisition cost lower than a Google Ad click |
| Pro Sonnet 4.6 | $9.99/mo | $0.030 | $0.90 | 91.0% 335 councils/mo before break-even |
| Constellation Opus 4.6 | $39.99/mo | $0.171 | $5.14 | 87.1% Premium tier, deepest analysis |
| BYOK Customer keys | $29.99/mo | $0.00 | $0.00 | 100% Pure platform revenue |
Monthly COGS assumes 30 council runs/month per user. Actual usage varies. All costs from OpenRouter API pricing as of March 2026.
Why this moment matters
Model costs are collapsing
In March 2026 we found 9 models under $0.01/council run. A year ago the cheapest viable option was $0.15. The cost curve makes multi-agent AI economically inevitable.
MCP is becoming standard
Model Context Protocol connects AI to enterprise tools. Airlock already orchestrates 7 MCP servers. As MCP adoption grows, our multi-repo architecture becomes a competitive moat.
Single-agent AI is hitting a ceiling
One model, one perspective, one blind spot. Multi-persona councils with quantified behavioral profiles (DECF drives) produce more robust analysis. The market is shifting from chatbots to deliberation systems.
Enterprise buyers want platforms, not plugins
Contract lifecycle management, CRM, project management, and AI orchestration — currently 4+ separate vendors. Airlock collapses the stack into one unified interface.
The product
Airlock is an enterprise data operations platform with a unified interface. Contracts, CRM, project management, calendar, and documents — unified in one workspace with AI councils that deliberate on business decisions.
The founder
Zac Holwerda
Military Intelligence Analyst turned investment researcher turned community builder turned platform founder. The through-line: due diligence and data lifecycle integrity.
Built and led the Brain Brigade community (Discord, 2020+). Published investment DD research on Substack. Wrote about ethical community design on Medium. Now applying that same obsession with structured analysis to enterprise operations.
Every feature in Airlock exists because Zac needed it and couldn't find it. The platform is opinionated because the founder is opinionated. That's not a bug.
If we raised, here's where it goes
Interested?
We're open to conversations with investors who bring more than capital — industry expertise, enterprise distribution, or technical depth in AI infrastructure. No pitch deck on request. We prefer real conversations.
invest@brainbrigade.xyzBrain Brigade Inc. · This page is informational only and does not constitute an offer to sell securities. All financial data reflects actual benchmark results and current pricing, not projections.