Skip to content

Agent Adoption Lab

2. Use-case portfolio

3. Operating model

10. People and change

11. ROI and cost

2. Use-case portfolio

3. Operating model

10. People and change

11. ROI and cost

EU rules for agentic AI

Describe, never compute

Incident simulator

How mature is your organization?

Twelve questions, one per point of the framework, about your real organization rather than the lab's. It takes about three minutes, and your answers are not stored anywhere.

1. Strategy Can you say which business outcome agents should improve, and why autonomy is needed?

Not really; we want to use AI

We have ideas but no agreed thesis

A thesis for some use cases

Every use case has a value thesis and an agent-fit check

2. Portfolio How are agent use cases chosen and sequenced?

Whoever shouts loudest

A list, without scoring

Scored on impact and feasibility

Scored on impact, feasibility, risk and readiness, and sequenced in waves

3. Operating model Who owns tool integrations, permissions, model changes, safety policy and rollback?

Nobody in particular

Informally, whoever built the agent

Named owners for some decisions

A cross-functional team with a RACI for all five

4. Governance How is what agents may do decided?

It isn't; they do what the prompt says

Guidelines in a document

Autonomy levels for each agent

Autonomy levels, a permission per tool, and audit logs of every action

5. Platform How are agents built?

Each team in its own way

Some shared libraries

A shared platform for new agents

One platform: identity, connectors, knowledge, observability and evaluation

6. Boundaries What stops an agent that goes wrong mid-task?

Nothing but its prompt

A human watching

Step and time budgets

Budgets, typed tool schemas, approval gates and fallback to recommend

7. Data How governed is the knowledge agents retrieve?

Whatever is in the wiki

Some sources have owners

Owners and access filters

Owners, freshness, access filters at retrieval, versioning, and outside content treated as untrusted

8. AgentOps How are agents tested before release?

We try them by hand

Some test prompts

A regression suite of scenarios

Golden tasks, red-teaming, canaries and automatic rollback

9. Security How are agents secured against manipulation?

We rely on the model

Secrets are kept out of prompts

Least privilege and tool allow-lists

Least privilege, allow-lists, egress control, sanitized inputs and a threat model

10. Change How do people work with agents?

Everyone figures it out

Some training

Named approvers with SLAs

Approvers with SLAs, escalation paths, specific training and updated SOPs

11. ROI How do you know agents pay off?

People like them

We watch the token bill

Cycle-time metrics

Cost per business outcome, with budgets, routing and caching

12. Scale How does a new domain get its first agent?

It starts from scratch

It copies another team's code

It reuses the platform

Federated squads on a standard platform, with shared evaluation and governance as code

0 of 12 answered

Part of CloudWorrior Labs(opens in a new tab) · by Charles Sambo(opens in a new tab)

An unhandled error has occurred. Reload 🗙

Reconnecting…

Your connection to Agent Adoption Lab dropped, probably a Wi-Fi blip. Hold on while it reconnects. Your inputs are kept.

The connection could not be restored. The page will reload in a few seconds.