Tau
Leading AI adoption · 12 lessons

AI adoption for managers: bring your team with you

Choose where AI will help your team, test it properly, pick tools you can trust and bring your people with you. For managers and anyone deciding which AI tools their organisation uses.

Who it’s for: Managers, team leads, executives and people who choose AI tools.

Buying AI licences is the easy part. The hard part is knowing where AI will help, proving it, and getting your people to use it well. Teams often roll out a tool, hear that people like it, and never find out whether it saved any time once checking and rework are counted.

This course takes a manager through it step by step. You'll score possible uses on value, feasibility and risk, then pick one or two worth testing. Before the test starts, you'll measure how long the task takes today, so there's a real number to beat. You'll then design a pilot of four to six weeks. This is a small trial with a rule, agreed in advance, for deciding whether to go ahead. You'll count rework and quality as well as minutes saved.

Later lessons cover what to ask an AI vendor about your data, how to write a one-page policy people actually follow, and how to set limits and approvals for AI agents. You'll sort AI uses into risk tiers using frameworks such as the EU AI Act and the NIST AI Risk Management Framework. Finally, you'll report value and risk to leadership on a single page.

By the end you’ll be able to…

  • Rank AI use cases by value, feasibility and risk
  • Run a 4–6 week pilot against a baseline with a go/no-go rule
  • Vet an AI vendor on data use, retention, audit logs and exit
  • Report AI value, incidents and risks to leadership on one page

Short on time? Start with the essentials

These 4 lessons cover the most useful ideas in this course. Each one takes about a minute.

Essential 1

Score AI use cases on value, feasibility and risk

Rank candidate AI use cases by business value, feasibility (data, tools, skills) and risk, and pick the one or two worth piloting first.

Essential 2

Design a 4–6 week AI pilot with a go/no-go rule

Set a pilot's scope, team, approved tool, success metric, guardrails and a go/no-go threshold agreed before it starts.

Essential 3

Count rework and quality, not just minutes saved

Calculate a pilot's net benefit from time saved minus checking and rework, alongside quality and error measures, and compare perceived with measured gains.

Essential 4

Write an AI use policy people actually follow

Write a one-page acceptable-use policy covering approved tools, data rules, required checks, disclosure, and a fast route to request new tools.

Every lesson in this course

This course has 12 video lessons of about a minute each, from the basics to expert habits. Each one ends with a question, three takeaways and its sources.

Foundations 2 lessons

  1. Score AI use cases on value, feasibility and riskEssential

    Rank candidate AI use cases by business value, feasibility (data, tools, skills) and risk, and pick the one or two worth piloting first.

    “Ten AI ideas, budget for two. Score them before the loudest voice wins.”

  2. Measure the baseline before an AI pilot starts

    Capture today's time, volume, cost, quality and error rate for a process so a pilot has a real number to beat.

    “'30% faster' than what? No baseline, no ROI.”

Building skills 4 lessons

  1. Design a 4–6 week AI pilot with a go/no-go ruleEssential

    Set a pilot's scope, team, approved tool, success metric, guardrails and a go/no-go threshold agreed before it starts.

    “Decide what 'success' means on day one, not after a demo impresses everyone.”

  2. Count rework and quality, not just minutes savedEssential

    Calculate a pilot's net benefit from time saved minus checking and rework, alongside quality and error measures, and compare perceived with measured gains.

    “Developers felt 20% faster with AI. Measured, they were 19% slower.”

  3. Check your data is ready before an AI rollout

    Test whether the data a use case needs is findable, current, clean and correctly permissioned, including overshared files an assistant could surface.

    “An AI assistant can surface every file a user can open, including the overshared ones.”

  4. Write an AI use policy people actually followEssential

    Write a one-page acceptable-use policy covering approved tools, data rules, required checks, disclosure, and a fast route to request new tools.

    “A 20-page ban gets ignored. A one-page 'yes, if' policy gets used.”

Advanced 5 lessons

  1. Vet an AI vendor on data, retention, logs and exit

    Question an AI vendor on training use of your data, retention, data location, sub-processors, audit logs, admin controls, certifications and how you get your data out.

    “'We don't train on your data.' Where is that in the contract, and for how long?”

  2. Sort your AI uses into risk tiers

    Classify AI uses by risk using frameworks as examples (EU AI Act risk categories, NIST AI RMF, ISO/IEC 42001), then check which rules apply to your organisation and markets.

    “Under the EU AI Act, AI that screens job applicants is high-risk. Is it on your list?”

  3. Set approvals, limits and logs for AI agents

    Define what an AI agent may do alone, which actions need human approval, its spending and data limits, what it logs, and how to stop it.

    “An agent that can send one email can send 500. Who approves the first?”

  4. Redesign roles task by task, and say so honestly

    Break a role into tasks, mark which become AI-assisted and which stay human-led, and explain the change to staff without promises you cannot keep.

    “'No one will lose their job' is a promise. Say what you know, and what you don't.”

  5. Plan AI training by role, not one course for all

    Build a role-based AI-literacy plan (all staff, power users, managers, owners of higher-risk uses) with records of who completed what, supporting duties such as EU AI Act Article 4.

    “Same AI course for the CFO and the contact centre? Neither gets what they need.”

Expert 1 lessons

  1. Report AI value and risk to your board on one page

    Present AI to a board or leadership team on one page: use cases and owners, measured value, incidents, top risks and control status, and decisions needed.

    “40 slides of demos → one page: value, risk, incidents, decisions needed.”

Apps we recommend

These are the tools this course uses, and what each one is best for. We have no partnership with any of them.

Credo AI

Best for: AI registry, risk tiering and policy packs

Catalogues AI systems, vendors and agents, flags shadow AI and maps controls to the EU AI Act, NIST AI RMF and ISO 42001.

Paid

OneTrust AI Governance

Best for: AI inventory tied to privacy and vendor risk

Inventories in-house and third-party AI, automates risk tiering and maps policies to EU AI Act, NIST AI RMF and ISO 42001.

Paid

Vanta

Best for: Getting ready for an ISO/IEC 42001 audit

Templates, mapped controls and automated evidence collection for an AI management system, reusable for other frameworks.

Paid
Hands-on project

One-page AI pilot plan for a real team process

Choose a process your team runs today. Score it on value, feasibility and risk, measure its baseline, and write a pilot plan an executive could approve: scope, approved tool, data rules, success metric, risks and controls, owners, timeline and a go/no-go rule. Use anonymised figures where needed.

What you’ll make: A one-page pilot plan with baseline, success metric, risk and control table, owners, timeline and a pre-agreed go/no-go rule

Check your own work

  • The baseline gives current time, volume and error rate, and says how and when it was measured
  • One primary success metric has a target number and counts checking and rework time
  • Every named risk has a control and an owner, and the data classification and tool approval are stated
  • The go/no-go rule is set before the pilot starts and says what happens after a 'no'
  • Applicable rules (e.g. EU AI Act risk tier, sector or privacy rules) are checked and noted, not assumed

Further reading and viewing

We’ve chosen these articles, research papers, videos and guides from other trusted sources.

Guide · NIST · 30 min

AI Risk Management Framework

The US voluntary framework (Govern, Map, Measure, Manage), plus its Generative AI Profile.

Guide · European Commission · 15 min

AI Act: regulatory framework on AI

Official summary of the EU AI Act's risk categories and the current dates obligations apply.

Guide · ISO · 10 min

ISO/IEC 42001:2023 AI management systems

The certifiable international standard for running AI governance as a management system.

Research · METR · 15 min

Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity

A randomised trial where felt speed-up and measured speed went opposite ways: why you need a baseline.

Guide · UK National Cyber Security Centre · 15 min

AI and cyber security: what you need to know

Plain-English questions on AI risk for managers, boards and senior leaders.

Questions about this course

How do you start AI adoption in a team?

Start small and specific. List the tasks where AI might help, then score each on value, feasibility and risk. Pick one or two, measure how they're done today, and run a short pilot with an approved tool. Agree in advance what result would mean carrying on. Bring your team into the choices early, and be honest about what will change.

How do you measure whether AI is saving time?

Measure the task before you start, including time, volume, quality and errors. During the pilot, record time spent drafting, checking and fixing separately. The real saving is time saved minus checking and rework. Compare that with what people say they saved, because perceived and measured gains can differ. This course has lessons on both steps.

What should an AI acceptable use policy include?

Keep it to one page so people read it. Cover which tools are approved, what data can go into each, the checks people must do before AI-assisted work goes out, when to disclose AI use, and a quick route to request new tools. A policy that blocks everything tends to push people towards unapproved tools, so make the approved path easy.

How do I keep company data safe when choosing an AI tool?

Ask the vendor whether your data trains their models, how long they keep it and where it's stored. Ask what audit logs and admin controls you get, and how you get your data out if you leave. Then check which laws apply to your organisation and markets, such as the EU AI Act, and set data rules staff can follow.

Do I need a technical background for this course?

No. It's written for managers, team leads and anyone choosing AI tools, not engineers. Each lesson is a short video of about a minute, followed by one question, three takeaways and the sources. If you're new to AI yourself, the AI at work essentials course is a good first step. The first 5 lessons are free.

Get early access to this course

Tau opens soon. Join the waitlist and we’ll email you as soon as you can start this course.