Your AI learning journey at Collingwood
How we build a demonstrably AI-savvy firm. This sets out how you grow your own capability with Claude, what good AI use looks like for your role, and how the programme supports you. Come back to it whenever you need it.
Everyone is expected to build fluency with Claude and keep progressing toward Power User level. It is a personal path with no boxes to tick. The point is simply to get good at using AI in your work.
The Capability Map sets out the AI competencies for each role type. It describes what good looks like at the level you are already at. It is a reference for performance conversations, not a promotion ladder.
Explore · tap any point to dive in
Built on a national standard
Our roles and competencies are not invented in-house. They are adapted from a recognised national standard, so the capability you build is benchmarked beyond Collingwood.
Everything here aligns to the AI Skills for Business Competency Framework v3.0.0, the UK national competency framework from The Alan Turing Institute and partners under the Innovate UK BridgeAI programme. We have taken what is useful for a firm of our size: four of its six personas, and three proficiency levels condensed from the framework's six. Each competency keeps the framework's reference code, so anyone who wants the detail can look it up.
How our roles map to the framework
| Collingwood role | Framework persona | What it means |
|---|---|---|
| All staff | AI Worker | Everyone at Collingwood. Uses AI in their role without being a specialist. Power users are staff working at the top of this tier. |
| AI Champion | AI Change Champion | The nominated, team-facing role. Trusted peers who model good practice, give first-line support, encourage sensible experimentation, and connect their team to the programme. |
| Technical team | AI Professional | The Head of AI and RevOps/BI, and champions as they take on build work. |
| Leadership | AI Leader | SMT and C-suite. The framework is clear that leaders are not expected to be technical experts. |
Senior and principal advisers sit within the all-staff role. Because their time goes to clients, champion duties sit with the associate tier, and they move toward the technical role only as AI features more in module delivery.
The framework overview
Read about the AI Skills for Business framework on the Turing Institute's site.
The full framework
The complete competency framework (v3.0.0) is published openly on Zenodo.
Power User vs AI Champion
Related, but deliberately different. One is a level of personal proficiency anyone can reach. The other is a nominated, team-facing role. All champions are power users. Not all power users are champions.
Power User
Someone who works fluently with Claude day to day and builds skills for their own work. Anyone in the firm can become one. It needs no time commitment beyond doing the job well.
A personal proficiency level, not a role. It describes working fluently with Claude and building skills for your own work.
- Work fluently with Claude across everyday tasks: drafting, analysis, research.
- Build skills that remove their own manual effort.
- Optionally, peer review a skill now and then (about 30 mins, ad hoc).
- Optionally, volunteer to demo skills or solutions they have built.
Only the optional asks above. Being a power user is just doing your job well with AI.
AI Champion
A nominated role with responsibilities to the team: modelling good practice, first-line support, building and sharing skills, and connecting the team to the programme.
A nominated role held by a power user who takes on team-facing duties. Every champion is a power user first.
- Build skills for their team, promoted org-wide where useful.
- Run show & tells and share knowledge through demos, posts and newsletter contributions.
- Take on complex, delegated work such as analysis, module creation and follow-ups.
- Give first-line support, triaging questions on prompting, skills and connectors so only genuinely technical issues escalate.
- Act as a feedback conduit, capturing what the team struggles with or quietly stops using.
- Spot adoption opportunities: repeated manual work that could become a skill.
- Provide light-touch quality and safe-use guardianship.
- Onboard new starters to the team's skills and ways of working.
In steady state, roughly 2 to 4 hours:
- A newsletter contribution where needed (30 to 60 mins).
- One show & tell or demo a quarter (about 40 mins a month, amortised).
- Ad hoc team posts and questions (30 to 60 mins a month).
- A short fortnightly champions sync.
Expect a heavier launch of about a day a month for the first four to six weeks. Skill creation sits outside this. It removes the team's own manual effort, so it is time-positive work a champion would want to do anyway.
How each is assessed
Both are assessed against the Collingwood AI Capability Map, our adaptation of the Turing Institute's framework. Each role type has an expected proficiency level, Aware, Applies or Leads, for every competency.
The Capability Map
Managers and staff assess against the expected level for the person's role.
Quarterly self-assessment
A short confidence pulse each quarter, rating yourself against your role's competencies. It builds a trend line over time, tracked by team.
Champions: shown in the work
Champions build a visible record through the programme itself: monthly demos, newsletter contributions and session recordings that feed their performance review.
Becoming a Power User
The one progression we expect of everyone. If you are not already a power user, this is the direction of travel, and the firm is set up to help you get there.
The expectation, plainly
We expect everyone to be progressing toward Power User level, if they are not already there. This is not about becoming a technical specialist. It is about using Claude fluently and sensibly in the work you already do.
What defines a Power User
- Works fluently with Claude day to day, as a natural part of how they draft, analyse and research.
- Builds skills for their own work, removing their own manual effort.
- Needs no defined time commitment beyond doing the job well.
How you get there
Capability is built through the everyday parts of the programme:
- Lunch & learns and the monthly session for training and live demos.
- The weekly AI newsletter for new tools, skills and ways of working.
- The skills library on the AI SharePoint site, with approved skills to use and learn from.
- The 4D skill framework, the firm's method for building and reviewing skills.
The capability baseline
These are the competencies expected of everyone at Collingwood. The list is also the curriculum for lunch & learns and the basis of the quarterly self-assessment. Becoming a power user means doing all of this comfortably, without prompting.
Protect sensitive & client data
Know what can and cannot be entered into AI, and use approved tools. (C16)
Use Claude confidently every day
Drafting, analysis and research as part of normal work. (C17)
Explore & experiment responsibly
Try new tools and features within firm policy. (C18)
Be open about AI's contribution
Be transparent about where AI helped with a piece of work. (C19, C21)
Use the firm's skills & automations
Work with approved skills in day-to-day workflows, in line with policy. (C22)
Judge AI output critically
Know when a quick check is not enough, and apply professional judgement before anything reaches a client. (C23)
Adapt as the role changes
Adjust as AI changes the work, and share what you learn with colleagues. (C24, C25 — Aware, moving to Applies)
The natural route to Champion
Power users take part as peer reviewers and trial users in the skill-building process. That is also the natural route by which a power user grows into a champion. If the team-facing role appeals, this is where it starts.
Champions & nominations
Champions are the nominated, team-facing backbone of the programme. Here is what the role involves, the time it takes, how it is recognised, and how to put a name forward.
What champions do
- Build skills for their team, promoted org-wide where useful.
- Run show & tells and share knowledge through demos, posts and the newsletter.
- Take on complex, delegated work: analysis, module creation and follow-ups.
- Give first-line support on prompting, skills and connectors.
- Act as a feedback conduit into the AI programme roadmap.
- Spot adoption opportunities: repeated manual work that could become a skill.
- Provide light-touch quality and safe-use guardianship.
- Onboard new starters to the team's skills and ways of working.
The time it takes
In steady state, roughly 2 to 4 hours.
- Newsletter contribution where needed: 30 to 60 mins.
- One show & tell or demo a quarter: about 40 mins a month.
- Ad hoc team posts and questions: 30 to 60 mins a month.
- A short fortnightly champions sync.
Heavier launch of about a day a month for the first four to six weeks. Skill creation sits outside this, as time-positive work.
How the role is recognised
Champions get visible credit for what they do, built into the programme rather than left to chance. Credit does not rely on someone remembering to give a shout-out.
Newsletter slot
A named "built this week" spot in the weekly newsletter.
Monthly demo slot
A standing slot for champion show & tells at the monthly session.
Named skills
Skills promoted to the team or org-wide carry their creator's name.
A record for reviews
Newsletter mentions and session recordings add up, so there is a clear record of a champion's work at review time.
Champions also have their own set of competencies in the Capability Map, several at Leads level.
Becoming a Champion
Champions are nominated. If the team-facing role appeals, or you can see it in a colleague, here is the route. There is no fixed intake, so put a name forward whenever it feels right.
Who it's for
Any power user who is up for the team-facing duties. You do not need to be the most technical person on the team. Modelling good practice and helping others matters more.
Put a name forward
Nominate yourself or a colleague. Managers and existing champions can nominate too. A short note on why is all it takes to start.
A quick conversation
We will have a short chat about fit and capacity, then get you set up with the champions group.
Ready to put a name forward?
Email Rory Dwyer and Amy Sheridan with a line on who you would like to nominate, including yourself, and why.
rory.dwyer@collingwood.group | amy.sheridan@collingwood.group
What good AI use looks like for your role
The competencies expected of each role type at Collingwood, in plain terms. Choose your role to see what is expected of you and how each capability is built. This is the reference point for AI in performance conversations. It describes the level you are at, not a ladder to climb.
Three proficiency levels
Condensed from the national framework's six.
Understands the idea, why it matters and when it applies.
Uses it reliably in their own work, without being prompted.
Can teach it, sets the team standard, or is accountable for it.
Expectations are cumulative
Champions meet everything expected of all staff, plus their champion set. The technical team meets the professional set on top of the all-staff baseline. Nobody drops the basics as they take on more.
"Developed through"
Each competency shows the activity that builds it, so the map doubles as a guide to why each part of the programme exists.
Choose your role
| Competency | Ref | Expected | Developed through |
|---|---|---|---|
| Protect sensitive and client data when using AI. Know what can and cannot be entered, and use approved tools | C16 | Applies | Lunch & learnsWeekly AI newsletter |
| Use Claude confidently in everyday work, including drafting, analysis and research | C17 | Applies | Lunch & learnsSkills library |
| Explore and experiment with new tools and features responsibly, within firm policy | C18 | Applies | Monthly sessionSkills library |
| Be open about where AI contributed to a piece of work | C19, C21 | Applies | Lunch & learns |
| Use the firm's skills and automations in day-to-day workflows, in line with policy | C22 | Applies | Skills library4D skill framework |
| Judge AI output critically. Know when a quick check is not enough, and apply professional judgement before anything reaches a client | C23 | Applies | Lunch & learns4D skill framework |
| Adapt as AI changes the role, and share what is learned with colleagues | C24, C25 | Aware → Applies | Weekly AI newsletterMonthly session |
Current development priority: Associate Advisers (July 2026)
Following the July 2026 Associate Adviser AI audit, two "All staff" competencies are the current training priorities for Associate Advisers, addressed through dedicated modules in their development pathway:
- Judging AI output critically (C23): knowing when work needs full verification before it reaches a client, and catching assumptions the AI made that nobody asked for.
- Correct setup of prompts and tasks (C18, C22): defining scope, constraints and what "done and checked" means before the work starts.
| Competency | Ref | Expected | Developed through |
|---|---|---|---|
| Act as their team's visible advocate for effective, ethical AI use | C26 | Leads | Champions syncMonthly session |
| Provide first-line support on prompting, skills and connectors | C27 | Leads | Champions syncSkills library |
| Encourage compliance with firm AI policy, and escalate risks to the Head of AI | C28 | Applies | 4D skill frameworkChampions sync |
| Spot repeated manual effort in the team's work and propose skills or automations | C29 | Applies | Champions sync |
| Encourage participation in training, and share resources and best practice | C30 | Applies | Monthly sessionWeekly AI newsletter |
| Connect their team and leadership on adoption, barriers and successes | C31 | Applies | Champions syncPerformance review |
| Stay current on tools and firm policy, and keep the team informed | C32 | Applies | Weekly AI newsletter |
| Model good practice in their own daily work | C33 | Leads | Monthly sessionPerformance review |
| Own their personal development and keep learning | C34 | Applies | Self-assessmentPerformance review |
Champions meet every "All staff" competency in full, in addition to the above.
| Competency | Ref | Expected | Developed through |
|---|---|---|---|
| Build low-code and no-code solutions that address firm problems and improve workflows | C50 | Leads | 4D skill frameworkSkills library |
| Manage data responsibly across its lifecycle: governance, security and retention | C57 | Leads | 4D skill framework |
| Document what is built so it can be reviewed and audited | C58 | Leads | Skills library |
| Engage stakeholders through a build so solutions are fit for purpose | C64, C66 | Leads | Monthly sessionSMT updates |
| Evaluate and oversee external AI vendors and third-party tools | C73 | Leads | SMT updates |
Only the part of the framework's Professional persona relevant to Collingwood is used. It applies to the Head of AI and RevOps/BI at this stage.
| Competency | Ref | Expected | Developed through |
|---|---|---|---|
| Set and communicate an AI strategy aligned to firm goals, balancing innovation with delivery | C74, C75 | Applies | SMT updates |
| Sponsor governance that keeps AI use lawful, ethical and responsible | C76, C77 | Aware | SMT updates |
| Select vendors and platforms through fair, evidence-based evaluation | C78, C79 | Aware | SMT updates |
| Anticipate the workforce impact of AI and invest in upskilling and change | C80, C81 | Applies | SMT updatesPerformance review |
| Manage AI risk and be ready for audits and incidents | C82, C83 | Aware | SMT updates |
| Build a culture of responsible innovation and trust in AI use | C84, C85 | Aware | Monthly sessionWeekly AI newsletter |
How the map is used
Each quarter, staff self-assess against their role's competencies, tracked over time by team. Managers use the expected levels as the reference point for AI capability in performance conversations. Team-level development priorities are set periodically by the Head of People & Culture and the Head of AI, informed by audits, reviews and the self-assessment trend.
How it works
The rhythm that keeps capability building: how we communicate, how skills get built and promoted, and how we know it is working.
Courses that build this
Every course and tutorial behind the programme lives on Claude Academy, free, and is filtered to your role. Worth working through alongside the day-to-day rhythm below.
Communications cadence
Weekly bulletin
AI developments relevant to Collingwood: Claude and the tools we use, ways of working, and changes affecting the firm, its sector or clients. Plus new tools and skills, and early sight of what is in flight.
Monthly session
Claude lunch-and-learn content, champion show & tells, and workshops such as ideation, a hackathon or group learning.
Ad hoc posts & sync
Champions post to their team channels whenever it is relevant, creating multiple touchpoints. A short fortnightly champions sync keeps the group aligned.
How the monthly session evolves
Months one to three are Claude-led lunch & learns. From month four the slot becomes champion demos and workshops, with a Claude lunch & learn only when a release warrants one. That gives the audience one habit rather than competing meetings, and sets the succession path as the Head of AI moves toward client-facing work.
How a skill gets built
Skill builds follow the firm's 4D Framework and its review chain: the SME builds, a peer reviews, the Head of AI signs off. Write-connector skills always need Head of AI review.
The lifecycle around the build
Scope
Walk the manual process end to end with the person who owns the pain: inputs, outputs, judgement points, edge cases.
Champion + pain owner (+ Head of AI where complex)Build
Champion drafts the skill under the 4D approach, reusing existing versions. Head of AI reviews structure and data handling, not domain content.
Champion (SME)Calibrate
Run the skill against completed real work where the answer is known, and compare. This is Discern with a known benchmark.
Champion + power user (peer review)Trial
A two-week live run with two or three team members, collecting friction points and refining.
Champion + power users in the teamPromote
Head of AI sign-off, demo at the monthly session, newsletter mention, filed on the AI SharePoint site. Org-wide only once proven at team level.
Champion + Head of AIOne cycle is roughly three to four weeks per skill, mostly champion time. Power users take part as peer reviewers and trial users, the natural route into the champion role.
Where knowledge lives
- Loom video recordings, with tools such as Guidde for step-by-step manuals.
- The AI SharePoint site as the single store for company-wide materials: the skills library, recordings, release summaries and switchover guides.
- Champions own their team's knowledge flow, making the team aware of changes, why they happened, and what they mean for the wider firm.
How adoption is measured
- Self-assessment: a quarterly confidence pulse, giving a trend line per tier.
- Usage data: Claude Team admin analytics, split by team.
- Skill metrics: live skills per team, and how much they are used.
- Time saved: a per-skill estimate captured at promotion, feeding the firm-wide baseline.
- Engagement: session attendance, newsletter readership, volume of champion posts.
- Qualitative: one line per champion each month on what their team stopped doing manually.
Reporting: a monthly view for the Head of AI and Head of People & Culture, and a quarterly one-pager to the SMT aligned to the programme's ROI indicators, so adoption and ROI tell one story.
The activities behind the map
These are the activities the Capability Map's "developed through" column points to.
Lunch & learns
Monthly training on Claude and AI ways of working, open to all.
Monthly session
Training, champion demos and workshops.
Weekly AI newsletter
The internal bulletin on AI developments, tools and skills.
Champions sync
The regular meeting of AI Champions with the Head of AI.
Skills library
The firm's library of approved Claude skills, with guides and recordings.
4D skill framework
Delegate, Describe, Discern, Diligence, with peer review and Head of AI sign-off.
Self-assessment
A short quarterly survey against your role's competencies.
Performance review
The standard review process, into which AI capability evidence feeds.
SMT updates
Roadmap, prioritisation and resourcing updates to the SMT.
Courses to complete
Free courses and tutorials from Anthropic's own Claude Academy, mapped to the Capability Map roles. Choose your role to see your list, or view everything at once.
How to use this page
Everything below is free on academy.claude.com and takes minutes to hours, not days. Expectations are cumulative, the same as the Capability Map: a Champion's list includes everything under All staff, and so on. Where a course also has a companion tutorial, both are listed, the tutorial is the quick version.
Choose your role
All staff
Claude 101 ↗
Everyday use of Claude: your first conversation, prompting, projects, artifacts and skills.
AI Fluency: Framework & Foundations ↗
The 4D framework Collingwood's own skill process is built on: Delegate, Describe, Discern, Diligence.
AI Capabilities and Limitations ↗
A working mental model of what Claude can and cannot do, and why.
Introduction to Claude Cowork ↗
Delegating full tasks to Claude: workspaces, standing context, plugins and the task loop.
Teach Claude your way of working using skills ↗
How to package what you know into a reusable Claude skill.
The 4 Properties of AI ↗
A seven minute run through the four properties that shape what Claude can do for you.
Delegating and scheduling tasks in Claude Cowork ↗
Setting Cowork to repeat a task on a schedule, without starting it yourself each time.
AI Champion
Customize Claude Cowork ↗
Adjusting Cowork's settings, permissions and defaults to fit how your team works.
How to customize plugins in Claude Cowork ↗
Adapting an existing Cowork plugin to your team's own workflow.
How to build a plugin from scratch in Claude Cowork ↗
Building a new Cowork plugin from nothing, for a workflow no existing plugin covers.
Technical team
The AI-Native SDLC Playbook ↗
A technical guide to rebuilding the software development lifecycle around AI.
Building Effective Human Agent Teams ↗
Moving from one person plus AI to a whole team working alongside agents.
Introduction to Subagents ↗
Breaking a complex Claude Code task into parallel subagents, run and checked independently.
Scaling workflows with Claude Cowork at your organization ↗
The playbook for rolling Cowork out organisation-wide: strategy, champion selection and enablement.
Leadership
Scaling workflows with Claude Cowork at your organization ↗
Written for admins, transformation leads and department leads: strategy, champion selection and enablement, the exact model this programme already runs.
Building Effective Human Agent Teams ↗
Moving from one person plus AI to a whole team working alongside agents.
The 4 Properties of AI ↗
The core mental model of what AI can and cannot do, without the hands-on exercises aimed at daily users.
Claude 101 ↗
Everyday use of Claude: your first conversation, prompting, projects, artifacts and skills.
AI Fluency: Framework & Foundations ↗
The 4D framework Collingwood's own skill process is built on.
AI Capabilities and Limitations ↗
A working mental model of what Claude can and cannot do, and why.
Introduction to Claude Cowork ↗
Delegating full tasks to Claude: workspaces, standing context, plugins and the task loop.
Teach Claude your way of working using skills ↗
How to package what you know into a reusable Claude skill.
The 4 Properties of AI ↗
A seven minute run through the four properties that shape what Claude can do for you.
Delegating and scheduling tasks in Claude Cowork ↗
Setting Cowork to repeat a task on a schedule, without starting it yourself each time.
Customize Claude Cowork ↗
Adjusting Cowork's settings, permissions and defaults to fit how your team works.
How to customize plugins in Claude Cowork ↗
Adapting an existing Cowork plugin to your team's own workflow.
How to build a plugin from scratch in Claude Cowork ↗
Building a new Cowork plugin from nothing, for a workflow no existing plugin covers.
Plus every course and tutorial under All staff.
The AI-Native SDLC Playbook ↗
A technical guide to rebuilding the software development lifecycle around AI.
Building Effective Human Agent Teams ↗
Moving from one person plus AI to a whole team working alongside agents.
Introduction to Subagents ↗
Breaking a complex Claude Code task into parallel subagents, run and checked independently.
Scaling workflows with Claude Cowork at your organization ↗
The playbook for rolling Cowork out organisation-wide: strategy, champion selection and enablement.
Plus every course and tutorial under All staff and AI Champion.
Scaling workflows with Claude Cowork at your organization ↗
Written for admins, transformation leads and department leads: strategy, champion selection and enablement, the exact model this programme already runs.
Building Effective Human Agent Teams ↗
Moving from one person plus AI to a whole team working alongside agents.
The 4 Properties of AI ↗
The core mental model of what AI can and cannot do, without the hands-on exercises aimed at daily users.
Not required, but the All staff list is open to any leader who wants to build personal fluency too.