The smartest companies have stopped pretending every team needs frontier models — they're running two parallel AI strategies, and the budget gap between them tells you everything about what actually works.
The Summary
- Allie Miller, CEO of Open Machine, breaks down how tech companies are structuring AI investments: baseline teams with subscription budgets vs. frontier teams with experimental budgets
- The divide reveals which AI applications deliver ROI today versus which are still R&D bets
- Most enterprise value comes from the boring stuff — finance, legal, marketing automation — not the moonshots
The Signal
The big AI spending story isn't about who's raising another billion. It's about how companies that already committed are carving up their budgets. Miller's framework maps the actual division happening inside enterprises right now: baseline AI teams getting subscription-tier access to proven models, frontier teams getting blank checks to experiment with cutting-edge capabilities.
The baseline teams are where the money gets made back. Finance departments using AI to close books faster. Legal teams automating contract review. Marketing teams generating the seventieth variation of an email campaign. These aren't sexy deployments, but they're the ones that pencil out on a quarterly P&L. The tools are stable, the use cases are proven, the budget is predictable.
"Frontier companies divide AI efforts into baseline capabilities and experimental teams — the budget gap between them tells you what actually delivers."
The frontier teams are different animals. Smaller headcount, bigger budgets, permission to fail. These are the groups testing whether an AI agent can run an entire customer support operation, or whether a model can replace a research analyst, or whether you can automate parts of software development that still require senior engineers. The success rate is lower, but the potential efficiency gains are measured in orders of magnitude, not percentage points.
What Miller's describing is the maturation curve of every infrastructure shift:
- Early adopters throw money everywhere hoping something sticks
- Winners emerge when companies figure out which applications pay for themselves
- The market splits into production tools (reliable, boring, profitable) and research bets (expensive, risky, potentially transformative)
We're hitting that middle phase now. The companies that survive the next two years won't be the ones spending the most on AI. They'll be the ones who figured out which 80% of their AI budget should go to proven, incremental wins, and which 20% should fund the swings that might redefine their business model.
The Implication
If you're running AI strategy inside a company, this is your template. Stop treating every AI initiative like it deserves the same budget or timeline. Separate the proven automation plays from the experimental agent deployments. The baseline work funds itself and buys you credibility. The frontier work is where you find competitive moats, but only if you can afford to fail three times before you succeed once.
For investors and builders, watch which companies are honest about this split. The ones still pretending every AI project is a strategic priority are burning capital. The ones running two-speed AI programs are the ones that will still be standing when the hype cycle turns.