Why this lesson comes first
Your team will hear two loud stories about AI. One says it will change everything next quarter. The other says it will take everyone's job. Neither story tells you what to do on Monday. In this lesson you will learn a quick test for sorting evidence from hype and fear, see what AI adoption actually changes for a team, and fill in a one-page starting point.
Running case: You manage a nine-person customer service and admin team. Six people handle email, chat and phone queries. Three handle invoicing, scheduling and records. Senior leaders have said that the business is "adopting AI". Some of your people are already using free chatbots. Others are quietly worried.
Task 1: Check any AI claim with three questions
When someone shares a statistic in a meeting, a news article or a vendor pitch, ask these three questions in this order.
- Who published it? Was it a university, a government body, an independent researcher, or a company that sells AI tools? A vendor can still publish good data, but you should note its interest.
- When? AI tools change quickly. A 2023 finding may not describe today's tools. Always write the year next to the number.
- Measured or forecast? Did someone observe real work, or is this a prediction or a self-reported survey?
Then give the claim a label:
- Evidence: measured, dated and sourced, with its limits stated.
- Hype: promises big gains without measurement, or stretches a narrow result into a general rule.
- Fear: presents a forecast or a possibility as a certainty about job losses or harm.
Worked examples
- Evidence. Brynjolfsson, Li and Raymond studied 5,179 customer support agents who were given a generative AI assistant. Issues resolved per hour rose by 14% on average and by 34% for novice and lower-skilled agents, with minimal effect on the most experienced agents (NBER working paper 2023; Quarterly Journal of Economics, 2025). This is measured, but it covers one company and one tool, so it does not promise the same result for your team.
- Forecast, often turned into fear. Goldman Sachs economists estimated in March 2023 that generative AI could expose the equivalent of 300 million full-time jobs worldwide to automation. "Exposed" does not mean "lost". Repeating this figure as "AI will take 300 million jobs" turns it into fear.
- Forecasts that disagree. The World Economic Forum's Future of Jobs Report 2025 (January 2025, an employer survey) projected 170 million roles created and 92 million displaced by 2030, a net gain of 78 million. That is a different estimate from a different method. When forecasts disagree, say so. Do not pick the one that suits your argument.
- Survey from an interested publisher. Microsoft and LinkedIn's 2024 Work Trend Index (May 2024, 31,000 people in 31 countries) reported that 75% of knowledge workers used AI at work, and 78% of those users brought their own tools. This is self-reported and comes from a vendor, but it is a useful warning that unofficial use may already be happening in your team.
- Headline that needs its context. MIT NANDA's The GenAI Divide (2025, marked "preliminary") said 95% of organisations it studied saw no measurable return from generative AI. Its sample was small: interviews with 52 organisations and surveys of 153 leaders. Using it to prove "AI never works" is fear. Ignoring it is hype.
Do this: Collect three AI claims your team has recently heard. Run each through the three questions and label it.
Task 2: See adoption across four layers
Adopting AI is not the same as installing software. For your team, it changes four layers.
- Tools. Which AI assistants are approved: for example ChatGPT, Gemini, Claude, Microsoft Copilot, or AI features built into your helpdesk system? Who has access, and which tools are people already using unofficially?
- Tasks. Which pieces of work change? In the case team, that could include drafting email replies, summarising long complaint threads or tidying records. The job is still there, but how individual tasks get done is changing.
- People. Habits, skills, confidence and roles. A new agent might gain the most, as the 2023 study suggests. An experienced agent might feel their expertise is being undervalued. Both reactions need attention.
- Governance. Rules on customer data, checking AI output, telling customers when AI was used, and who is accountable when a reply is wrong. Trust depends on this layer.
A useful analogy is moving office. The furniture (tools) is the easy part. What really changes is who sits where, how people work together, and the house rules.
Do this: For the case team, write one sentence for each layer that describes what might change. Do not decide anything yet. You are only describing the change.
Task 3: Fill in the Adoption Starting Point
Copy this one-page template and complete it with your team.
Adoption Starting Point: [Team name], [Date]
- Current state
- Tools officially available:
- Tools people already use (ask, without blame):
- Main types of work and approximate weekly volume:
- How the team feels about AI today (in their own words):
- Four layers: what might change
- Tools:
- Tasks:
- People:
- Governance:
- Open questions (things nobody knows yet)
- e.g. "Are we allowed to paste customer emails into any AI tool?"
- Agreed sources of truth
- Company AI or IT policy (name and date):
- Official help pages for each approved tool:
- Named contacts (IT, data protection, HR):
- Rule: any statistic we quote includes its source, its date, and whether it was measured or forecast.
- Next review date:
Do this: Fill in sections 1 and 3 with your team in a 30-minute meeting. Agreeing sources of truth early stops rumours from turning into decisions.
Key takeaways
- Ask three questions of every claim: who published it, when, and was it measured or forecast.
- Label claims as evidence, hype or fear, and say openly when forecasts disagree.
- No statistic guarantees an outcome for your team.
- AI adoption changes four layers: tools, tasks, people and governance.
- A one-page starting point with open questions and agreed sources of truth builds trust from day one.