The Cascade Adoption Model
A point of view on how engineering teams actually adopt AI — and why most stall.
The core claim
AI adoption is a measurement problem, not a tooling problem.
Most engineering organizations don't fail at AI adoption because they picked the wrong tool. They fail because they can't see what's actually happening. Seat counts get reported to the board while nobody knows whether the tool is changing how work gets done. A few enthusiasts get great results, the rest quietly revert to old habits, and six months later the CTO is asked to justify spend with nothing but anecdotes.
The Cascade Adoption Model exists to name that failure mode and give teams a way out: adoption progresses in identifiable stages, and you cannot move to the next stage without measuring the one you're in.
The four stages
Adoption doesn't spread evenly. It cascades.
It starts with a trickle of individual use and, if managed deliberately, builds into something that flows through the entire organization. Most teams stall at Stage 1 or 2 and mistake it for success.
Stage 1 — Trickle
A few engineers, invisible impact.
A handful of early adopters are using AI tools daily and getting real value. Nobody else is, and leadership has no visibility into the difference. Usage data, if it exists at all, is seat activations — which measure procurement, not adoption.
The tell:
The CTO says “we have Copilot” when asked about AI strategy.
What's actually needed:
A baseline. Who is using what, for which tasks, and where the friction is. You cannot manage what you haven’t measured, and at this stage nothing has been measured.
Stage 2 — Channel
Deliberate workflows, uneven adoption.
The team has identified specific workflows where AI tools help — code review prep, test generation, migration work, documentation. Champions have emerged. But adoption is still concentrated: 20% of the team drives 80% of the usage, and the gap between the enthusiasts and everyone else is widening.
The tell:
Wildly different answers when you ask five engineers how they use the tools.
What's actually needed:
Channeling — codifying what the champions do well into team-level practices, and measuring workflow integration rather than raw usage. This is where most organizations plateau, because closing the enthusiast gap requires deliberate enablement, not more licenses.
Stage 3 — Cascade
Team-wide practice, measured impact.
AI-assisted work is the default, not the exception. New engineers are onboarded into AI-integrated workflows from day one. Leadership can connect tool usage to delivery signals — DORA metrics, cycle time, review throughput — and can distinguish real productivity effects from noise.
The tell:
The CTO can answer “what’s the ROI on our AI spend?” with data, in under two minutes.
What's actually needed:
Sustained instrumentation. Adoption at this stage is fragile — tool churn, team turnover, and workflow drift can quietly erode it. Measurement is what keeps the cascade flowing.
Stage 4 — Current
Self-improving, compounding.
The organization doesn’t just use AI tools well — it has a repeatable process for evaluating new ones, retiring ones that stop earning their keep, and folding new capabilities into existing workflows. Adoption is no longer a project. It’s an operating capability.
The tell:
When a new tool ships, the team has an answer within a quarter — adopt, pass, or watch — backed by their own evaluation data, not vendor benchmarks.
What's actually needed:
Governance without bureaucracy. A lightweight, recurring evaluation rhythm owned by the team, not imposed on it.
The three measurement dimensions
Seat counts appear on none of them.
At every stage, progress is measured on the same three dimensions.
Coverage
Who is actually using AI tools in their daily work? Not who has a license — who has integrated the tool into at least one recurring workflow. Coverage exposes the enthusiast gap.
Depth
How embedded are the tools in how work gets done? A team can have 100% coverage with everyone using AI for autocomplete and nothing else. Depth measures workflow integration: which tasks, how often, replacing what.
Impact
Is any of it changing delivery outcomes? Impact connects adoption to signals leadership already trusts: DORA metrics, cycle time, review load, and — critically — team-reported friction. Impact without the first two dimensions is unattributable; the first two without impact is activity theater.
A stage transition requires movement on all three. A team that improves coverage without depth has bought more autocomplete. A team that claims impact without measuring coverage is guessing.
How to use the model
Three ways it earns its keep.
Self-diagnosis
Most CTOs can place their team in about thirty seconds — and most place themselves a stage higher than the evidence supports. The honest test: can you answer the “tell” question for the stage you claim, with data?
Prioritization
The model is sequential on purpose. Teams that try to jump from Trickle to Cascade — usually via a mandate — skip the channeling work and generate resentment instead of adoption. The failure pattern is nearly universal: mandate, spike, quiet reversion.
Communication upward
The stages give CTOs a vocabulary for board conversations. “We’re at Stage 2, here’s the plan and the metrics to reach Stage 3 by Q3” is a fundable answer. “Adoption is going well” is not.
Where we come in
How our engagements map to the model.
The model is also how we scope work.
AI Readiness Assessment establishes which stage you're actually at, measured across coverage, depth, and impact — the baseline most teams skip.
Workflow Coaching is the Stage 2 → 3 engine: codifying champion practices into team-wide workflows and closing the enthusiast gap.
Adoption Reporting is the instrumentation that sustains Stage 3 and builds toward Stage 4 — ongoing measurement that keeps adoption from quietly eroding and gives leadership a defensible ROI narrative.
The Cascade Adoption Model is a working framework developed from hands-on engagements with engineering teams. It will evolve as the tooling landscape does — that's the point.
Find your stage
Where does your team actually stand?
Book a free 30-minute discovery call. We'll help you place your team honestly — and map out what it takes to reach the next stage.