Maker? Manager? - The AI Acceleration is Here, But It Is Uneven
Why the bottleneck is moving from makers to managers
The “velocity drives growth” story is playing out over and over, the latest banger being Lovable (70M ARR at month 7 with 35 people). 1
I recently heard Andrew Ng talk about a 1 PM per 0.5 engineer ratio at one of his companies. Similarly, Brian Balfour spoke about velocity as a moat stacking function that creates strategic advantages by compressing roadmaps from quarters to days. The PM-Eng ratio flip is wild, a hook - definitely atypical. But it is an interesting framing of the lopsided acceleration enabled by AI.
The Uneven Compression
To paraphrase William Gibson -
"The
futurecompression is already here, it's just not evenly distributed".
AI has accelerated the "maker layer" - coding using Cursor, designing prototypes with Figma’s AI-assisted workflows, writing copy with Claude by orders of magnitude.
But managers - who own outcomes, risk, and cross-functional coherence - remain underpowered. By “managers,” I don’t mean people managers, I mean anyone accountable for high-leverage, stacked decisions with cross-org impact. These are:
PMs translating market signal to strategy
Engineers designing architectures, reviewing code, making trade-offs
Legal, brand, and design reviewers ensuring coherence and safety
Current AI investment heavily favors individual contributor productivity. But, the tooling that empowers strategic decision-making and risk management remains underpowered. Yes, we're building toward agents and agentic systems. But it's unclear how much of this is gated by human review.
Either way, due to a mix of systems, culture and org design, there is a maturity curve in how teams are leveraging the acceleration that AI affords.
The Gap
The acceleration paradox is clear when we examine how teams actually operate. Consider how this plays out in practice in product development. There are three types of orgs:
Legacy - This is most companies today
AI Enabled - Due to tight budgets + efficiency mandates + downward pressure on headcount + people navigating integrating AI into workflows, these are teams of AI enabled IC’s “figuring things out”
AI Native - AI isn't bolted onto existing processes - the entire operational model is redesigned around AI's strengths. Decision-making, coordination, and execution happen at machine speed while humans focus on strategic outcomes and course correction.
Like at Perplexity AI, teams use AI before consulting colleagues. Their "Slime Mold" model with 2-3 person teams eliminates traditional coordination overhead. The default isn't "schedule a meeting" or "create a doc" - it's "ask AI first."
The AI Native Org is not on the radar yet for most enterprise though. Why? Because it requires trust in system-level decisions enabled by:
Explainability: Trace decisions or debug outcomes with confidence.
Guardrails: Robust "do not cross" systems to prevent brand, legal, operational or UX risk.
So What
"This seems like a natural diffusion of innovation. Designers get AI tools first, decision-makers get them later. Why is this a problem?"
Three things:
The Efficiency Bar Is Raised: ZIRP-era growth by headcount is over. Lovable achieved $70M ARR with 35 people - 10 engineers, 3 growth people, zero salespeople. Ramp’s AI agents achieve 99% accuracy in expense approvals with complete audit trails, 90% of transactions auto-code before ERP sync, while underwriting reduced approval time from days to hours. Efficiency determines valuations and outcomes for startups.
Decision Density Creates Compound Advantages: It's not just "faster shipping" - it's hundreds of micro-decisions together unlocking velocity.
This one blew my mind - when GitHub flagged Lovable for API abuse (creating 1 repository every 2 seconds), they implemented an AWS S3-based custom Git backend within hours. This goes against everything we understand about investing in undifferentiated work, but shows what’s possible as R&D costs fall.Pressure For Capital Efficiency: High velocity organizations achieve better unit economics - higher output per dollar invested. When investors can back companies achieving 3x the output with 1/3 the team size, capital flows. GTM velocity determines funding outcomes.
The bottleneck has moved from making to managing
Until organizations trust AI systems to operate without constant human oversight, AI acceleration remains a local unlock, not a system level acceleration. Every review checkpoint, every manual validation, every 'human in the loop' requirement maintains the status quo pace.
High-output pods need traceable, high-fidelity outcomes with low cost of course correction.
Overall, delivery velocity is becoming table stakes. Sustainable advantage lies in systems, culture, and org design that enables velocity with explainability.
The question isn't whether simply being AI Enabled is "good enough" - it's whether we, as Product Managers, and the broader product teams are “AI native” - that we have the systems, culture and org in place to unlock strategic advantages due to resource efficiency, market responsiveness, risk distribution, and competitive positioning.