Services · AI Adoption

AI adoption, measured by telemetry — not surveys.

Most organisations measure AI by counting licenses, sending surveys, and tallying chatbot logins. That measures exposure — not transformation. We measure the real thing: every prompt, skill, agent, MCP tool call and multi-agent flow.

Sample telemetry · enterprise AI gateway
Events / 60s · ≈40
  • 09:42:11promptuser.Asummarize Q3 deck and flag riskL2
  • 09:42:14tooluser.Asearch → 14 issuesL3
  • 09:42:16agentreview-agentsub-agent invoked: fact-checkerL4
  • 09:42:18mcpcrm-mcpreport.fetch (id=Q3-NA)L3
  • 09:42:21skilluser.Bskill: weekly-board-prepL2
  • 09:42:23promptuser.Cdraft RFP response, tone formalL2
  • 09:42:24toolreview-agentvector.search → 9 chunksL3
  • 09:42:27agentops-orchestratorspawned 3 agents (intake / triage / route)L5
  • 09:42:29mcpgit-mcppr.review (repo=core)L3

The problem, reframed

Without telemetry, AI adoption stays invisible.

How it’s usually measured

Exposure metrics

License counts, surveys, training completions. They tell you who could use AI — never who actually does.

  • How many employees have access to AI?
  • How many ChatGPT licenses are active?
  • How many trainings were completed?
  • How many people answered 'yes' in the survey?

How we measure it

Behavioral telemetry

Objective signals from the systems people actually use, model- and infrastructure-agnostic.

  • Who is using AI daily — and how intensely?
  • Who is building skills and reusable workflows?
  • Who is orchestrating multi-agent systems?
  • Where is AI creating measurable business impact?

What we track

Eight signal families. One telemetry layer.

If a user prompts a model, runs a skill, invokes a tool, calls an MCP server, or spawns an agent — that interaction generates measurable adoption data.

01
PROMPTS

Prompt-level AI usage

PROMPTS

Every model call leaves a fingerprint. Frequency, structure and context-awareness reveal who is actually thinking with AI — and who is one-lining it.

  • Frequency and intensity
  • Prompt structure complexity
  • Context-aware interactions
  • Model switching behavior
Sample telemetry · schematic · last 24h
Daily prompts / user↗ trending
Avg tokens / prompt≈ hundreds
Multi-turn sharemajority

From telemetry to maturity

Five levels of AI maturity.

Raw telemetry becomes structured intelligence. Each user, team and department gets scored on a five-level ladder, by behavior — not self-report.

LEVEL 3

Integrated AI workflows

Tool-connected AI is embedded into business processes — ticketing, ERP, source control, internal APIs.

tool callsMCP serversprocess triggers

Model- and infrastructure-agnostic

If it produces inference, orchestration, or tool-calling telemetry — we can measure it.

  • GPT family
  • Claude
  • Llama
  • Mistral
  • Qwen
  • DeepSeek
  • Gemma
  • Open-source LLMs
  • On-prem AI
  • Enterprise AI gateways

The chain

Behavior → telemetry → maturity → transformation.

True AI transformation happens when AI is embedded into workflows, composed into systems, delegated as agents and orchestrated across teams. Those behaviors generate telemetry. Telemetry reveals maturity. Maturity drives transformation.

  1. — Behavior

    Real interactions

    Embedded into workflows, composed into systems, delegated as agents, orchestrated across teams.

  2. — Telemetry

    Objective signals

    Prompts, skills, agents, tool calls, MCP traffic, model-switching, inference events — captured continuously.

  3. — Maturity → Transformation

    Scored capability

    Telemetry rolls up to L1–L5 maturity per user, team, department — the only honest measure of AI transformation.

What enterprises gain

AI adoption becomes a data discipline — not a narrative.

  • Quantify real AI usage

    Replace anecdotes with intensity scores per user, team, function.

  • Identify power users

    Surface champions and shadow practitioners worth elevating.

  • Detect shadow AI

    Spot unauthorised tools, models, or data flows in real time.

  • Benchmark departments

    Compare maturity across functions and over time.

  • Track transformation

    See progression L1 → L5 by team, role, and quarter.

  • Measure ROI

    Tie AI usage to business outcomes — output, cycle time, cost.

  • Optimize enablement

    Direct training where telemetry shows it actually moves the needle.

  • Strengthen governance

    Continuous compliance, not annual surveys.

Next step

Stop guessing. Start measuring.

45-minute readiness review: we map your AI surfaces — gateways, IDEs, MCP servers, model providers — and leave a 4-week telemetry pilot plan with a maturity baseline.

What you leave with

Telemetry source map
Maturity baseline (L1–L5)
4-week pilot plan