Assessment: Adoption Maturity Scoring
What This Is For¶
This assessment tells you where your organization actually stands on AI adoption — not how many licenses you bought, but how deeply and safely AI is used. It is the measurement companion to the AI Adoption Framework (the argument) and the Designing an Adoption Program guide (the how-to). Score yourself before designing a program, to find the binding constraint; score again on a cadence, to see whether it is moving.
The reason a dedicated assessment is needed is the gap the framework documents: access is not adoption. As of early 2024 only 32% of desk workers had tried AI for work — two-thirds had not — and just 15% strongly agreed they had the training to use it effectively (Slack, 2024). Even where AI is used, regular use trails far behind trial: in 2025 about 60% of desk workers used AI at least occasionally, but only around 40% weekly and 20% daily (Slack, 2025). A maturity score that counts seats would call this success; a score that counts habitual, safe, workflow-integrated use tells the truth.
This page gives you six scoring dimensions, a 1–5 maturity ladder for each, the metrics that feed them, a resistance-mapping overlay, and a method for interpreting the result.
How to Use It¶
- Score each of the six dimensions 1–5 using the ladders below, evidenced by the metrics in the next section — not by impression.
- Plot a profile, not an average. A single low dimension is a bottleneck even when the mean looks healthy. Use the radar/profile to find it.
- Overlay the resistance map to see where in the workforce adoption is stuck.
- Interpret the pattern (final section) to choose the next intervention, then point fixes at the relevant phase of the adoption program guide.
The five-level structure follows established maturity-model convention — the same Initial → Managed → Defined → Quantitatively Managed → Optimizing progression as CMMI (CMMI Institute, ISACA) — so a score is comparable to how the organization already thinks about other capabilities.
The Six Dimensions¶
| # | Dimension | The question it answers |
|---|---|---|
| D1 | Tool Utilization & Engagement | Are people actually using it — deeply, broadly, and habitually? |
| D2 | Cultural Readiness | Is it safe to experiment, and is trust in AI calibrated? |
| D3 | Leadership Modeling & Sponsorship | Do leaders visibly use AI and actively sponsor adoption? |
| D4 | Enablement & Skills | Are people trained and supported to use AI well? |
| D5 | Governance & Acceptable-Use Clarity | Do people know what is allowed, with sanctioned tools that suffice? |
| D6 | Trust & Verification Practices | Are AI outputs verified appropriately — neither blindly trusted nor reflexively dismissed? |
The 1–5 Maturity Ladder¶
D1 — Tool Utilization & Engagement¶
| Level | What it looks like |
|---|---|
| 1 | Licenses provisioned but little real use; no usage tracking; activation well below ~30%. |
| 2 | Activation rising but engagement shallow; daily-active/monthly-active (DAU/MAU) stickiness in the low single digits; use concentrated in a few enthusiasts. |
| 3 | Majority of target users activated; weekly use common in target roles; DAU/MAU approaching ~20%. |
| 4 | Habitual daily use across most target roles; DAU/MAU around the ~31% B2B-SaaS benchmark; breadth across functions. |
| 5 | AI embedded in core workflows; usage tracked and tied to outcomes; sustained retention rather than novelty spikes. |
D2 — Cultural Readiness (psychological safety, experimentation, trust)¶
| Level | What it looks like |
|---|---|
| 1 | Fear dominates; AI errors are punished; experimentation absent; deep distrust of outputs. |
| 2 | Pockets of curiosity; experimentation tolerated informally but unsafe to admit failure. |
| 3 | Teams openly discuss AI use; failures treated as learning; calibrated trust beginning to form. |
| 4 | Psychological safety high; experimentation is normal and expected; trust calibrated to task risk. |
| 5 | Continuous, open experimentation is the norm; sharing wins and failures is routine across teams. |
D3 — Leadership Modeling & Sponsorship¶
| Level | What it looks like |
|---|---|
| 1 | No visible sponsor; leaders do not use AI themselves. |
| 2 | Verbal support only; no modeling, no resourcing. |
| 3 | A named sponsor exists; some leaders visibly use AI. |
| 4 | Active, visible, reachable sponsorship; leaders model use and remove barriers. |
| 5 | Sponsorship sustained and cascaded through management; leaders coach adoption and disclose their own AI mistakes. |
D4 — Enablement & Skills¶
| Level | What it looks like |
|---|---|
| 1 | No training; entirely self-taught. |
| 2 | Ad hoc resources; a minority trained; under five hours of learning is typical. |
| 3 | Formal, role-based onboarding for target roles; prompt-and-judgment skills taught. |
| 4 | Continuous enablement: champions network, office hours, just-in-time support, communities of practice. |
| 5 | Skills measured, refreshed, and adapted as capabilities change; enablement keeps pace with the tools. |
D5 — Governance & Acceptable-Use Clarity¶
| Level | What it looks like |
|---|---|
| 1 | No policy; high shadow-AI exposure; people guess at what is allowed. |
| 2 | A blanket ban or silence — which drives covert use rather than preventing it. |
| 3 | A published acceptable-use policy exists; sanctioned tools are available. |
| 4 | Clear data-handling rules; sanctioned tools cover real needs; shadow AI is low. |
| 5 | Governance enables rather than blocks; monitored and reviewed; demand is met by sanctioned tools. |
D6 — Trust & Verification Practices¶
| Level | What it looks like |
|---|---|
| 1 | Outputs either blindly accepted or reflexively dismissed; no shared norms. |
| 2 | Verification is ad hoc and personal; over- and under-trust both common. |
| 3 | Basic human-in-the-loop expectations exist for higher-risk uses. |
| 4 | Verification habits are explicit and matched to task risk; people know when to check. |
| 5 | Calibrated trust is cultural: AI relied on where reliable, verified where weak, and the difference is taught. |
Tool Utilization & Engagement Metrics (D1)¶
Score D1 on evidence, using product-analytics measures rather than seat counts:
- Activation rate — share of provisioned users who have completed a meaningful first action, not merely logged in (Amplitude, 2025).
- DAU/MAU stickiness — daily-active over monthly-active users, the standard embeddedness proxy. Useful reference points: B2B SaaS sits around 31%, with 25–35% considered broadly healthy (Mixpanel, 2026).
- Breadth — percentage of target roles with any regular use, to catch adoption that is real but confined to one team.
- Retention — 30- and 90-day sustained use, which separates habit from a novelty spike. (Daily use grew rapidly through this period — Slack measured daily AI use up 233% in the six months to mid-2025, versus November 2024 — so trend matters as much as level (Slack, 2025).)
- License waste — provisioned-but-inactive seats. Across SaaS generally, organizations actively use only about 49% of provisioned licenses (Zylo, 2024); AI rollouts are not exempt, and the unused half is both wasted spend and a false adoption signal.
A useful framing: the daily users are where value concentrates. Slack found daily AI users were markedly more likely to report "very good" productivity, focus, and job satisfaction than occasional users (Slack, 2025) — which is why D1 rewards depth and habit, not breadth of access alone.
Cultural Readiness Indicators (D2)¶
Culture is scored from signals, not vibes:
- Psychological-safety survey — adapt Edmondson's validated items; psychological safety is the strongest enabler of the team learning behavior that experimentation requires (Edmondson, 1999), and was the top dynamic of effective teams in Google's large internal study (Google, 2015).
- Disclosure comfort — whether people will admit AI use to a manager. This is a leading indicator of real use: those comfortable disclosing are 67% more likely to have used AI for work (Slack, 2024).
- Experimentation throughput — count of shared experiments, prompts contributed, wins and failures posted. Low throughput with high usage signals adoption being driven underground.
- Trust disposition — calibrated against a wary baseline: only 46% of people globally are willing to trust AI (KPMG, 2025), so neither blanket enthusiasm nor blanket suspicion is the target.
Resistance Mapping¶
Adoption maturity is not uniform across a workforce, so a single score hides the real problem. Overlay the population onto Rogers' adopter segments — innovators 2.5%, early adopters 13.5%, early majority 34%, late majority 34%, laggards 16% (Rogers, 2003) — and locate where adoption has stalled.
The decisive boundary is between the early adopters and the early majority. The enthusiasts (the first ~16%) adopt on novelty; the majority adopt only on proof, and the transition between them is where most programs stall — the "chasm" (Moore, 1991). Practical use of the map:
- Stuck before ~16% penetration (Rogers, 2003) → you have reached the enthusiasts and no further. The constraint is usually proof and relevance, not access — invest in role-based use cases and visible peer wins (D4).
- Stuck at the majority → the constraint is usually safety and manager engagement (D2, D3) — pragmatists adopt when their manager and peers visibly do.
- High usage but concentrated → breadth problem; check D1 breadth and the champion coverage feeding it.
Segment the resistance, too: fear of job loss, fear of looking incompetent, distrust of outputs, and loss of autonomy call for different responses (see the framework). Mapping which fear dominates which segment tells the program where to aim.
Scoring and Interpretation¶
Record a 1–5 for each dimension and read the profile, not just the mean. The shape is diagnostic:
| Pattern | What it means | Where to act |
|---|---|---|
| High D1, low D2/D5 | Usage is real but unsafe and ungoverned — the shadow-AI pattern: people are using AI, often unsanctioned, and hiding it | Build safety and channel demand into sanctioned tools (program guide, Phases 4–5) |
| High D3/D4, low D1 | Investment is not converting — the access-vs-use gap: leadership and training are in place but habitual use is missing | Check relevance and friction; role-based use cases, remove workflow blockers |
| High everything except D6 | Adoption is broad but trust is uncalibrated — over- or under-trust risk | Teach verification habits matched to task risk (D6) |
| Uniformly low | Early stage | Start at Phase 0 — sponsorship and governance — before anything else |
Two interpretation rules hold across patterns. First, a high mean does not redeem a low bottleneck dimension — adoption is gated by its weakest link, much as the framework argues culture gates the whole transformation. Second, trend beats level: re-score on a regular cadence, because a rising score that plateaus at the majority is the signal to change tactics, not to keep doing more of what reached the enthusiasts.
Key Takeaways¶
- Score use, not access. Trial and licenses overstate adoption; habitual, deep, workflow-integrated use is the real measure (Slack, 2024; Slack, 2025).
- Six dimensions, scored as a profile. Utilization, culture, leadership, enablement, governance, and verification — read the shape, because the weakest dimension is the bottleneck.
- Engagement is measurable. Activation, DAU/MAU (~31% B2B-SaaS benchmark), retention, breadth, and license waste give D1 an evidence base (Mixpanel, 2026; Zylo, 2024).
- Map the resistance. Use Rogers' segments to find where adoption stalls — most often at the chasm between enthusiasts and the pragmatic majority (Rogers, 2003; Moore, 1991).
- Re-score on a cadence. Maturity is a trend, not a one-time grade; the profile tells you which program phase to invest in next.
Sources¶
- Slack Workforce Lab — Workforce Index (June 2024), 2024. 32% of desk workers have experimented with AI tools; only 15% of global desk workers strongly agree that they have the education and training necessary to use AI effectively. View source · verified 2026-06-20 · primary
- Slack Workforce Lab — The New AI Advantage, 2025. Three in five desk workers use AI at least occasionally, two in five use it weekly, and one in five use it daily. View source · verified 2026-06-20 · primary
- CMMI Institute (ISACA) — Capability Maturity Model Integration, 2018. CMMI defines five maturity levels: Initial, Managed, Defined, Quantitatively Managed, and Optimizing. View source · verified 2026-06-20 · primary
- Amplitude — Top Digital Product Adoption Metrics, 2025. The activation rate gauges the percentage of users who complete a predefined action that signifies meaningful engagement with the product. View source · verified 2026-06-20 · primary
- Mixpanel — MAU: Definition, Formula, and Benchmarks, 2026. B2B SaaS stickiness averages about 31% across North America and EMEA; a product sitting between 25% and 35% is broadly in line with the market. View source · verified 2026-06-20 · primary
- Slack Workforce Lab — The New AI Advantage, 2025. The share of workers using AI every day has more than doubled in six months and is now 233% higher than it was in November 2024. View source · verified 2026-06-20 · primary
- Zylo — SaaS Management Index, 2024. Companies are only using half (49%) of their provisioned SaaS licenses. View source · verified 2026-06-20 · primary
- Slack Workforce Lab — The New AI Advantage, 2025. Casual users (less than once a week) show little to no difference in outcomes from non-users; weekly users see gains, but the biggest impact comes from integrating AI into the daily workflow. View source · verified 2026-06-20 · primary
- Amy C. Edmondson — Psychological Safety and Learning Behavior in Work Teams (Administrative Science Quarterly, 44(2), 350-383), 1999. A shared belief held by members of a team that the team is safe for interpersonal risk taking; team psychological safety is associated with learning behavior, which mediates between psychological safety and team performance. View source · verified 2026-06-20 · primary
- Google re:Work — Project Aristotle: Understand Team Effectiveness, 2015. Psychological safety was far and away the most important of the five dynamics we found; it is the underpinning of the other four. View source · verified 2026-06-20 · primary
- Slack Workforce Lab — Fall 2024 Workforce Index, 2024. Those who are comfortable sharing that they used AI for work tasks are 67% more likely to have used AI for work than those who would not be comfortable admitting AI use. View source · verified 2026-06-20 · primary
- KPMG & University of Melbourne — Trust, Attitudes and Use of AI: A Global Study 2025, 2025. Only 46% of people globally are willing to trust AI systems. View source · verified 2026-06-20 · primary
- Everett M. Rogers — Diffusion of Innovations (5th ed.; ISBN 978-0743222099), 2003. Adopter categories: innovators 2.5%, early adopters 13.5%, early majority 34%, late majority 34%, laggards 16%, obtained by partitioning the normal adoption distribution at the mean plus or minus one and two standard deviations. View source · verified 2026-06-20 · ⚠ secondary mirror
- Geoffrey A. Moore — Crossing the Chasm, 1991. Moore identifies a chasm between the early adopters (visionaries) and the early majority (pragmatists): the gap where most high-tech adoption efforts stall. View source · verified 2026-06-20 · primary