TBM 415: Demand Mix, Discovery, and AI as a (Dys)function Multiplier

John Cutler · email · 2026-04-09 source ↗ Reader 🖥️

AI as Amplifier Code Was Never the Bottleneck Product Discovery & Experimentation

How teams should operate is determined by the mix of demand they face and how they shape that demand — and AI will not fix your funnel but will reveal and amplify whatever version of it you are already running, accelerating both functional and dysfunctional patterns.

theoretical · expository — Theoretical: argues for understanding demand mix as the primary lens for organizational analysis, rather than prescribing specific actions. Expository in form: builds a conceptual framework (demand funnel, mix, mandate levels) and then applies it to AI. The AI section is the payoff, but the framework comes first. Classic Cutler: the diagrams carry as much argument as the text.

Structure

Part	Role
The demand funnel as helpful simplification	Framework introduction — establishes the funnel metaphor and the two key variables: (1) what's in the funnel and (2) how it's shaped
Mix and approach: the spectrum from interrupt-driven to self-sourced	Core taxonomy — shows that practices must be contextual to the demand mix, not universally applied
No team is an island: coupling patterns	Extends the framework to multi-team dynamics — tightly coupled (shared interrupts) vs. loosely coupled (shared priorities)
Demand and prioritization: four jobs	Connects demand mix to prioritization behavior — high-interrupt → efficiency/support, self-shaped → leverage/alignment
Demand and mandate levels	Adds vertical dimension — from 'build exactly this' to 'generate business outcomes', with different parts of the org at different altitudes
Controlled intake vs. distributed chaos: the two personas	Primary evidence — the efficient feature factory PM vs. the 'empowered' but fragmented PM. Both are rational responses to their demand context.
AI thoughts: the ingrained mental model and its alternative	Climactic argument — applies the entire framework to AI adoption. Two mental models: engineering-as-bottleneck (AI accelerates feeding the constraint) vs. learning-as-purpose (AI accelerates learning)
Questions	Reflective close — 8 diagnostic questions for the reader to apply the framework to their own organization

Key terms

demand mix Product Discovery & Experimentation
The composition of what enters a team's work funnel — where it came from, who is involved, and what type of work it is (support tickets, internal requests, production fires, strategic goals). The mix determines what practices are appropriate. Two teams with identical processes but different demand mixes will have completely different outcomes.
demand shaping Product Discovery & Experimentation
The activities a team and organization use to actively influence what enters the funnel — discovery, experimentation, back-pressure, selective intake, strategic alignment. Distinct from prioritization (which happens after demand arrives). Shaping happens upstream; prioritization happens downstream.
(dys)function multiplier AI as Amplifier
AI amplifies whatever organizational patterns already exist — if demand is well-shaped and teams are learning, AI accelerates learning; if demand is chaotic and teams are negotiating, AI accelerates chaos. The parenthetical signals that the same technology produces opposite effects depending on context.
the funnel Product Discovery & Experimentation
Not a sales funnel. The flow of all inputs (ideas, requests, fires, goals) through filtering, shaping, and execution toward change. Cutler's key move: the funnel can end at 'feeding a team' (bad) or at 'meaningful change for customers' (good). Where you think the funnel ends determines everything.
mandate levels
A vertical spectrum from 'build exactly this' (task execution) to 'generate long-term business outcomes' (strategic autonomy). Different parts of the organization operate at different altitudes simultaneously. The left side (prescriptive) looks simple but pushes complexity downstream; the right side (empowered) looks messy but engages complexity directly.
premature convergence Product Discovery & Experimentation
Making decisions about what to build too early in the process, before sufficient learning has occurred. Driven by the mental model that engineering is the bottleneck and work must be 'de-risked' before reaching it. AI accelerates this pattern if the mental model doesn't change.

Propositions

explicitHow teams should operate — their processes, prioritization, discovery practices — is determined by the mix of demand they face, not by abstract best practicesconf: high
explicitdevelopedAI will not fix the funnel — it reveals and amplifies whatever version of it you are already runningconf: high
extends ←TBM 430: Incubate, Compound, Refinance, Liquidate · p7 — Both state the same amplifier thesis -- AI does not fix underlying dysfunction, it accelerates whatever is already there -- but TBM 415 applies it to the product-discovery funnel while TBM 430 extends it to portfolio/technical-debt economics, generalizing the claim to a second domain.
explicitThe dominant mental model of product development treats engineering as the bottleneck, and the entire system organizes around feeding that constraint efficiently — this model is the root cause of most organizational dysfunctionconf: high
explicitIf you keep the engineering-as-bottleneck mental model, AI will accelerate all existing dysfunctions: more ideas generated, more artifacts produced, more negotiation, more overload — all fasterconf: high
explicitThe alternative mental model ends the funnel at meaningful change for customers, not at feeding a team — and in this model, AI accelerates learning rather than feeding a bottleneckconf: high
implicit'Empowered' product teams without coherent demand shaping are often just empowered to negotiate chaosconf: high
implicitAn efficient feature factory with controlled intake can outperform a theoretically 'empowered' organization with chaotic demand mixconf: high
explicitMany companies believe they were learning faster than they could ship, and see AI as an instant fix — but the real constraint was never shipping speedconf: high
explicitcorroboratedVibe-coded prototypes and vibe-written PRDs are just higher-velocity versions of premature convergence — new rituals replacing old ones without changing the underlying modelconf: high
supportsNo, AI Did Not Kill Agile · p5 — Cutler's vibe-coded prototypes and vibe-written PRDs as higher-velocity premature convergence is Kerievsky's 'waterfall with faster computers' observed in practice.
supports ←The hidden pattern behind successful products | Mark Pincus (Founder of Zynga) · p14 — Using AI to lovingly build one polished idea over three months is exactly the 'higher-velocity premature convergence' warning; Pincus prescribes the antidote — a hundred disposable failures to keep learning open.
implicitThe truly scarce resource in most organizations is not engineering capacity but focus, clarity, and the ability to make coherent decisionsconf: high

Arguments

AI applied to the engineering-as-bottleneck mental model will accelerate organizational dysfunction rather than resolve it strength: high

Explicit premises:

The dominant mental model treats engineering as the constraint and organizes around feeding it
This model produces premature convergence, performative certainty, large batches of pre-shaped work, and effort shifted from understanding customers to managing internal dynamics
AI makes it faster to generate ideas, write PRDs, prototype, and ship
If the mental model stays the same, more of everything enters the funnel faster

Implicit premises:

The bottleneck was never actually engineering capacity — it was the organization's ability to learn and make coherent decisions
Speed of execution without speed of learning is not just wasteful but actively harmful — it creates more sunk-cost attachment to bad ideas
The people who advocate for AI adoption in organizations are overwhelmingly operating within the engineering-as-bottleneck mental model

Inferences:

Organizations that adopt AI most eagerly (those who feel most constrained by engineering capacity) may be the ones most damaged by it — because their mental model ensures AI accelerates their dysfunction
The Turing Post 'Unsexy Truth' framework (legibility as prerequisite for AI) maps directly to Cutler's argument: you can't shape demand coherently if you can't describe your own workflows. Legibility is the infrastructure for demand shaping
The vibe coding debate is a surface manifestation of this deeper question: vibe coding is the engineering-as-bottleneck model applied to the individual — 'I can code faster' without 'I can learn faster'

Practices that look dysfunctional in one demand context can be perfectly rational in another — there are no universal best practices strength: high