Lesson 5.1

AI at work

How jobs change, and what stays human

Wednesday 18 November · 16:20

Where we are

Lesson 5 of 12: AI at work and in teams

01 · Done

Introduction to Generative AI

Mon 9 Nov · 08:40, 10:20

02 · Done

AI for you: making it work in your day

Tue 10 Nov · 08:40, 10:20

03 · Done

Getting the most out of your AI tools

Wed 11 Nov · 08:40, 10:20

04 · Done

How AI systems work

Wed 18 Nov · 14:20

05 · You are here

AI at work and in teams

Wed 18 Nov · 16:20

06

Spotting AI opportunities

Fri 20 Nov · 08:40, 10:20

07

Practical business with AI

Mon 23 Nov · 08:40, 10:20

08

Ethical AI and what AI costs

Fri 27 Nov · 14:20, 16:20

09

Building with AI

Mon 30 Nov · 12:40, 14:20

10

Build and pitch prep

Fri 4 Dec · 08:40, 10:20, 12:40, 14:20

11

Final presentations

Tue 8 Dec · 08:40, 10:20

12

Reflection and final Q&A

Wed 9 Dec · 08:40, 10:20

Recap

In lesson 4, our main objective was to:

Understand how AI systems work

What we covered

Connect your tools

The connector lets AI use your notes and apps.

5 building blocks

LLM, prompt, context, skills and tools.

Memory

Short-term is this chat. Long-term is found with RAG.

Context engineering

Inputs + transformation = outputs.

Workflows

Each step’s output is the next step’s context.

Today · Lesson 5.1

What we will cover

01

How search has changed

Spotlight talk by Rokas

25 min · Present

02

AI + humans

From helper to robot

30 min · Present · Discuss

Anthropic Economic Index job explorer

03

What’s happening to jobs

Usage, jobs and secret cyborgs

35 min · Present · Discuss

After the break · 5.2Why team AI stallsAI is different in a teamPrivacy and trustBuy or buildThe future of workImplications for work

Lesson 5.1 · Part 1

How search has changed in the era of AI

We went from a book, to a list of links, to one answer.

BEFORE THE WEB

Yellow Pages

You look in a book by type of business. You read, phone and ask.

FROM 1998

Google Search

You type a few words. You get a list of links, then click and read.

FROM 2022

Ask an AI

You ask a question in your own words. The AI reads for you and gives one answer.

Rokas, smiling

Spotlight Talk by Rokas

SEO, GEO and how AI finds things

Lesson 5.1 · Part 2

AI + humans

How jobs change when people and AI work together

How we got here

The evolution of man

Apes

Early humans

Hunters

Office workers

Developers

What comes next · The 5 phases of AI for workers

The evolution of AI at work

01

Role augmentation

AI helps you do your job.

02

Task substitution

AI does some of your tasks.

03

AI agents

AI does whole pieces of work, step by step.

04

Digital employees

AI joins the team as a worker.

05

Embodied AI

AI gets a body and works in the real world.

Phase 1 of 5 · Role augmentation

Supporting your skills

AI helps with some tasks inside the job you already have.

  • AI helps you decide, look at data and do routine tasks.

  • You stay in control. AI makes you faster.

  • Example: customer service staff use AI to find answers fast.

ChatGPT ▾
A customer asks where order 4821 is. Can you check and write a reply?
Order 4821 left the warehouse on Monday. It should arrive on Thursday.

Here is a reply:
Hi Sam, your order is on its way and should arrive on Thursday. Thanks for waiting!
Ask anything
1Role augmentation2345

Phase 2 of 5 · Task substitution

Replacing a whole task

AI takes over repeated, rule-based tasks, so you can do harder work.

  • Routine tasks are handed over to AI.

  • People focus on new ideas, plans and solving problems.

  • Example: AI does data entry or answers simple customer questions.

AI chat
Add these 40 invoices to the sheet.
Using skill: invoice entry
Read 40 invoices
Find the date, amount and supplier
Add 38 rows to the Invoices sheet
Flag 2 invoices for you to check
12Task substitution345

Phase 3 of 5 · AI agents

Replacing a whole workflow

AI agents finish work on their own, with little or no checking by people.

  • Agents run a process from start to end, within set limits.

  • People step in less for routine work.

  • Example: an AI assistant books meetings and sends simple messages by itself.

Workflow: book meetingsOn
New emailSomeone asks to meet
AI agent reads itFinds who, what and when
Check the calendarPicks a free time
Send the inviteAnd a short reply
123AI agents45

Phase 4 of 5 · Digital employees

Replacing an employee

AI takes on set roles that people used to do.

  • A digital employee has a role in the team, like a virtual colleague.

  • People and digital employees do tasks together.

  • Example: an AI assistant gives customer support day and night.

Our team

# general

# support

# sales

# support
JL

Jonas 09:12

A client asks about a refund. Can someone answer?

Ava APP 09:12

Done. I sent the client our refund steps and added a note in the CRM. ✅

Message #support
1234Digital employees5

Phase 5 of 5 · Embodied AI

Giving AI a body

AI moves into robots that act in the physical world on their own.

  • AI comes into the workplace as robots or smart devices.

  • It can do tasks that need moving or handling things.

  • Example: robots in factories, hospitals and warehouses.

12345Embodied AI

Lesson 5.1 · Part 3

What’s happening to jobs

Who uses AI, which jobs it touches, and where it beats people

AI usage

Where people use AI the most

How much each country uses Claude, compared with its number of people. 1× is what you would expect.

1× expected

Australia

6.40×

Singapore

5.81×

Switzerland

5.02×

Luxembourg

4.85×

New Zealand

4.84×

Canada

4.13×

Norway

4.04×

Iceland

4.02×

Malta

4.00×

France

3.97×

Source: Anthropic Economic Index, Claude.ai conversations, updated 26 June 2026. anthropic.com/economic-index (opens in a new tab)

Jobs that use AI the most

AI is already part of these jobs

People hand AI routine parts, like data entry, and work with it on ideas. Share of all Claude.ai conversations.

Document management specialists

5.10%

Librarians

4.22%

Counter and rental clerks

2.72%

Editors

2.58%

Writers and authors

2.22%

Computer support specialists

2.17%

Computer systems analysts

1.91%

Web and interface designers

1.70%

Source: Anthropic Economic Index, Claude.ai conversations, updated 26 June 2026. anthropic.com/economic-index (opens in a new tab)

Which roles?

Where AI fits most and least

Microsoft Research (2025) matched 200,000 Copilot chats to the tasks in each job.

Fits most

Interpreters and translators

Writers and authors

Customer service staff

Fits least

Nursing assistants

Roofers

Dishwashers

Humans vs AI

AI outcompeting humans

Each bar runs from the first test to the year AI did as well as people. The bars get shorter.

200020052010201520202025

Handwriting

2018 · 20 years

Speech

2018 · 20 years

Images

2015 · 6 years

Reading

2018 · 2 years

Language

2020 · 2 years

Common sense

2023 · 4 years

School maths

Close, not yet

Writing code

Close, not yet

Years read from a chart by Will Henshall for TIME, with data from Contextual AI, so they are close, not exact. Tests: MNIST, Switchboard, ImageNet, SQuAD, GLUE, HellaSwag, GSM8K and HumanEval.

The reality

AI as the great leveller

It lifts those who start lower the most…

…but it fails quietly on tasks it cannot do.

758 consultants at Boston Consulting Group worked on real tasks, with and without GPT-4.

+43%

Lower half

Quality of work rose 43% for those who scored lowest on their own.

+17%

Top half

The strongest performers gained much less.

Source: Dell’Acqua et al., Navigating the Jagged Technological Frontier, Harvard Business School, 2023 (opens in a new tab)
Recap · Lesson 2

What people really do

Secret cyborgs

Many people already use AI at work and do not tell anyone. They may be unsure of the rules, or afraid of looking lazy.

Be open about how you use it.

Quiet use means no shared learning

It also means no shared rules

Open use lets a team get better together

A job is a bundle of tasks

AI takes tasks, not whole jobs

Look at your own future job as a list of tasks.

01

Tasks AI takes

Drafting, summarising, sorting, searching and first checks.

02

Tasks that grow

Judgement, talking to people, deciding, and caring.

03

New tasks

Checking AI work, setting it up, and teaching others.

Lesson 5.2

Driving AI adoption in teams

How to get a whole team using AI, and keep it going

Wednesday 18 November · after the break

Driving AI adoption

Six barriers to adoption

Teams stall on people, data and rules, not on the technology. Clear each one.

01

Internal data

The information the AI needs is scattered, private or out of date.

02

Adoption

People do not trust it, do not have time to learn it, or fear for their job.

03

Customisation

A general tool does not know your company, your words or your rules.

04

Privacy and security

People paste private data into tools the company has not checked.

05

Accountability

When AI gets it wrong, nobody knows who is responsible.

06

Unclear value

Nobody measures what it saves, so the project loses support.

Barrier 1 · Internal data

Connect AI to your work tools

A connector lets your AI read the tools you already use.

  • Ask about your own work: “What did the team decide on Tuesday?”

  • The AI looks in Slack, Drive or Linear and answers.

  • It only sees what you are allowed to see.

Barrier 2 · Adoption

Why people do not use AI

Buying the tool is easy. Getting people to use it is the hard part.

Skills

They do not know where to start, or how to ask well.

Digital confidence

Some people are less at home with new tools.

Time

Learning feels like extra work on top of the day job.

Incentives

Nothing rewards them for trying. The old way still counts.

Fear

They worry AI will replace them, or make them look lazy.

Trust and rules

One wrong answer, or unclear rules, and they stop.

Barrier 3 · Customisation

Teach AI how your company works

  • Give it your data, know-how, processes, examples and rules.

  • Save them as a skill or project instructions.

  • Answers then fit your company, not the internet.

Our way of workingSkill

Used in every chat by the whole team

Internal dataPrice list, client list, last year’s reports.
Expert domain knowledgeWhat our senior people know, written down.
Process understandingHow a refund moves from request to payment.
ExamplesThree great proposals, and one bad one.
Business rulesNever promise a delivery date. Say “clients”.

More than a style guide

Show it, don’t just tell it

Barrier 4 · Privacy and security

Protecting your A$$ Data & IP

  • Agree which AI tools the company has checked.

  • Sort data into three groups everyone can remember.

  • More on this later, in Privacy and trust.

Fine to usePublic facts, your own drafts, made-up examples.
Approved tools onlyInternal documents, plans, numbers.
Never pastePersonal data, health data, passwords, client contracts.

Barrier 5 · Accountability

A person owns every AI answer

  • Name one owner for each AI task.

  • Check before you send or decide.

  • Keep a record of what the AI did.

AI draftsA reply to a client about a refund.
JLJonas checksHe fixes one wrong date.
JLJonas sendsHis name is on it, so he owns it.

It’s the only step that matters

Barrier 6 · Unclear value

Measure what it saves or gains

  • Pick one number before you start, such as time per task.

  • Measure it before and after.

  • Share the result with the team.

Increased revenueDeals closed each month811
Decreased costsSupport costs each month€4k€2.5k
Decreased timeTime to reply to a client4 hours20 min
Reduced riskMistakes found in checks93

Lesson 5.2 · Part 1

AI is different in a team

One person and a chat window is easy. A team is harder.

Multiplayer mode

What changes when it is not just you

On your own

You choose the tool

Your notes are your own

You decide what is good enough

Mistakes cost you

In a team

Everyone needs the same tools and rules

Information has to be shared safely

Someone sets the standard

Mistakes reach customers

One slide, four words

Privacy, security, governance, precision

Ask all four before you roll anything out.

01

Privacy

Whose information is it, and is the AI allowed to see it?

02

Security

Who can get in, and where does the data go?

03

Governance

Who decides the rules, and who checks they are followed?

04

Precision

How often is it right? How do we know?

A simple test

Where does your data go?

Before you paste anything into an AI tool, know where it goes. This matters most for work and client information.

Use tools your organisation has approved

Check the data and training settings

Keep names, secrets and client data out of prompts

Lesson 5.2 · Part 2

Buy or build

Four ways to get an AI system, with trade-offs.

Choose with care

Four options

Option

What it is

Good for

Watch out for

AI-first software

A new product built around AI.

Fast results

Newer, can change fast

Old software plus AI

A tool you use already adds AI features.

Low effort to start

Features may be basic

A platform

A service you configure and build on.

Flexible

Needs skill and time

Build in-house

Your team builds it itself.

Fits you exactly

Costly to build and keep

Discuss · 10 minutes

Buy or build your project?

01

Does a tool like this already exist?

02

What would make ours different?

03

Do we have the skill and time to build it?

04

What if the supplier changes the price or closes?

The 5 phases, in a team

Every phase needs an owner

In 5.1 you placed your own role on the 5 phases. On your own, you are the owner. In a team, say who owns what.

Who sets the instructions or skills?

Who checks the output?

Who is told when it goes wrong?

A career thought

Your project is your portfolio

People get hired by showing what they built, not only what they studied.

A live prototype beats a list of skills

Write down what you decided, and why

Keep it in a place you can share

Lesson 5.2 · Part 3

The future of work

Bigger questions, with no easy answers.

Follow the money

Incentives and ethics

Companies that make AI also need to make money. That shapes what they build and how it behaves.

Ask who pays for the tool

Ask what they gain from your use

Ask what happens to your data

Discuss · 10 minutes

What data is valuable?

01

What do you know that nobody else has written down?

02

Whose data would a company pay for?

03

Who should own the data you create?

04

What would you never want sold?

A real deal · August 2026

Old company data is now worth money

Spirit Airlines went bankrupt. Google paid $10 million for its old business data, to help train AI.

Millions of staff emails and Teams chats, plus pricing data

A rival bidder, Mercor, offered $7.5 million

Personal details are removed before Google gets it

The flight attendants’ union objected

Source: CNN, Forbes and Cybernews, 18 August 2026. The sale needed court approval.

Lesson 5.2 · Part 4

Implications for work

What AI means for finding a job, and for the teams that hire

Recap

Four ways to deal with a problem

Resolve

Do what worked before. It is fast and good enough.

Solve

Find the best answer, with data and tests.

Absolve

Hand the problem to someone else, so it is no longer yours.

Dissolve

Change the way you work, so the problem goes away.

How will you grow?

Four ways to get the work done

Hire people

More of the same team.

Resolve

Automate the role

Build a system that does the work.

Solve

Outsource the work

To people or to AI, outside your team.

Absolve

Hire an AI worker

A new kind of team member.

Dissolve

Before you choose

What happens to your current team?

Your current team

What do you do with the people you have?
How do you teach them new skills?
Does it still make sense to pay people in these roles?
Do you still hire people, or only AI systems?

What does it mean to hire an AI worker?

Planning your AI workforce

Hire an AI worker

What do you need to do to train an AI worker?
What will it cost?
How does it change how fast you can grow?

Hands-on · 5 minutes

Save your takeaways

Use your AI tool and the Humble connector to keep what you learned today.

Save a note in Humble called “Lesson 5: AI at work and in teams”, with my three biggest takeaways from today: [type them here]. File it in a folder called “ISM Essentials of AI”, and make the folder if it does not exist yet.

Lesson 5: AI at work and in teams

New note · ISM Essentials of AI

Wrap-up

What we covered today

Jobs change

Tasks change, not whole jobs.

Humans and AI

Use both well.

The future

Questions we will keep asking.

Teams

Agreements, owners and barriers.

Trust

Privacy, security, governance, precision.

Buy or build

Four options.

Up next

Lesson 6: Spotting AI opportunities

Start with the business problem, then use the evaluation framework, your options and the five interfaces of AI.

Friday 20 November, 08:40 and 10:20

Open the Lesson 6 slides

You need the class password.

← Lesson

1 of 51