The AOaaS Thesis: Why $13T Demands Agentic Operating Systems
By Abhishek Barua, Founder, Astra Space AI Most AI companies are solving the wrong problem. They're building horizontal automation: helping SMB workers 10% faster, one feature at a time.…
By Abhishek Barua, Founder, Astra Space AI
Most AI companies are solving the wrong problem.
They're building horizontal automation: helping SMB workers 10% faster, one feature at a time. It's easy to understand. Easy to sell. Easy to build.
But underneath, there's a bigger category nobody's talking about: Agentic Operating Systems (AOaaS).
And the market is $13T, not $1.5T.
The $13T Problem
Enterprise labor budget: $3.7T annually.
But that's the revenue pool, not the problem pool.
The actual problem—the decision infrastructure that locks up 40% of that $3.7T—is closer to $13T when you account for:
- Compliance & audit overhead: $1.2T+ annually (people verifying other people's decisions)
- Process orchestration & handoffs: $2.1T+ in efficiency loss (waiting for approvals, rework, escalations)
- Data validation & reconciliation: $800B+ in labor (ensuring data quality for human decisions)
- Exception handling: 40% of human time spent on edge cases instead of strategy
- Artificial approval bottlenecks: Decisions that should be automated but must be human because "compliance" (really: legacy architecture)
Add it up. $13T in structural inefficiency. Not a technology problem. An operating-model problem.
That's the market AOaaS addresses.
Why Vertical Beats Horizontal
Look at who's winning in AI:
Cresta: Built an operating system for insurance, banking, telco. Not a chatbot. Not a feature. An OS.
Decagon: Insurance workflows end-to-end. Product/domain fit so tight that switching costs become structural, not incidental.
Sierra: Customer success OS. Not "help agents answer tickets faster." An automated operating model.
Why do they win?
Because they moved the decision boundary, not just the speed dial.
The Lock-in Difference
Horizontal AI:
- Locks through convenience (I use ChatGPT because it's available)
- Wins on breadth (every SMB knows about it)
- Loses on moat (anyone else can use ChatGPT too)
Vertical AOaaS:
- Locks through structural dependency (my entire operating model runs on this system)
- Wins on depth (30x customization per vertical)
- Builds real moat (switching costs are business-operational, not technical)
One is a product. One is infrastructure.
How AOaaS Scales to $13T
Start with one vertical: say, insurance underwriting.
Today: underwriters review cases, compliance team audits decisions, legal team checks for liability. 3 handoffs. 2-3 day cycle for policy approval.
With AOaaS: agents underwrite cases, assign risk, flag compliance issues. Humans audit the agents' reasoning (not the original case). 1 handoff. 2-3 hour cycle.
Result: $500M+ in operating leverage for a single $1B+ insurance customer.
Now do that across every vertical: insurance, banking, telco, healthtech, logistics, procurement. Each one unlocks $200M-$1B in labor efficiency.
That's $13T across the enterprise world.
The Timing
We're in Year 0 of AOaaS. Most capital is flowing to:
- Horizontal "AI agents" that do one thing well
- SMB automation (small revenue, no moat)
- Feature parity (ChatGPT + automation layer)
None of that will matter in 5 years.
What will matter: the companies that moved the decision boundary. The ones that replaced the entire operating model, not just assisted it.
That's AOaaS. That's the category Astra is building.
FAQ
Q: Isn't this just "enterprise automation" with a new name?
A: No. Automation speeds up humans (10% faster workers). AOaaS replaces the human decision entirely (agents decide + humans audit → 100x faster cycles). Insurance automation = "help underwriters work faster." Insurance AOaaS = "agents underwrite, humans verify." Different category, 10x different business outcomes.
Q: Why not build horizontal first, then go vertical?
A: Horizontal is commoditizing (GPT-4 is good enough). Everyone can use it. No defensibility. Vertical requires:
- Deep domain expertise
- Custom compliance architecture
- Operating-model redesign with the customer
- That's defensible. That's a $10B+ business. That's where the $13T is.
Q: How does Astra fit?
A: Astra Space AI is shipping the first true AOaaS for [vertical]. Our thesis:
- $13T market > $1.5T market
- Vertical defensibility > horizontal convenience
- Founder-led engineering > offshore feature shops
- That's how we win.
Follow the build: @abhishek_barua on LinkedIn | Astra Space AI