Jobs, skills and work redesign in the AI era

0:00 Jobs, skills and work redesign

Few topics attract more confident claims with less evidence than AI and jobs. One headline says mass unemployment. The next says nothing will change. In this lesson you'll get a grounded view: what credible research suggests, what's genuinely uncertain, and a practical, task-based method to redesign work and build skills responsibly.

0:22 Tasks, not jobs

Start with the most important idea. Tasks, not whole jobs, are the unit of change. Research from organisations like the International Labour Organization looks at exposure task by task. Most jobs contain some tasks AI can help with and many it can't. The ILO's analyses have found clerical and administrative work among the most exposed, and have emphasised that augmentation, AI helping with parts of a job, is more common than full automation for most occupations.

0:55 Why it matters

Why does this matter? Because your people are already asking what AI means for their jobs, and silence fills with rumours. Leaders who respond with hype or with fear lose trust. Leaders who respond with a clear, evidence-based approach, looking at tasks, redesigning roles together, investing in skills and sharing the gains, keep their best people and get more value from AI. How you handle this conversation will shape adoption, morale and retention far more than which tool you choose.

1:30 The spreadsheet lesson

Here's an analogy. When spreadsheets arrived, they didn't eliminate accountants. They eliminated a lot of manual arithmetic, and accountants spent more time on analysis and advice. Some roles shrank, new ones appeared, and the people who learned the tool early had an advantage. AI is a bigger and faster change, and the outcomes are genuinely uncertain, but the lesson from spreadsheets still helps: think in tasks, not jobs, and invest in the skills that grow in value.

2:03 Simple example: a support role

A simple example of task mapping. A customer-support agent's week includes answering order-status questions, handling refund requests, dealing with angry complaints and writing weekly feedback summaries. Order status is high-volume and rule-based: automate with checks. Refunds need policy judgement: augment. Angry complaints need empathy and authority: keep human, with AI suggesting a draft. Feedback summaries: automate, with a human reviewing themes. One role, four different answers.

2:32 Productivity evidence

What about productivity? Controlled studies of customer support agents, writers, consultants and developers found real gains in speed and quality on suitable tasks, often bigger for less experienced workers. But they also found worse outcomes when people used AI on tasks beyond its capabilities. The lesson: gains are real, uneven, and depend on knowing where AI works and where it doesn't.

2:59 Surveys and usage data

Employer surveys and usage data add more colour. The World Economic Forum's Future of Jobs Report 2025 found employers expecting substantial changes in required skills, with AI and data, technological literacy, analytical and creative thinking, resilience and curiosity growing. And analyses of real AI usage, like the Anthropic Economic Index, show heavy use in software, writing and analysis, with a mix of augmentation and automation. Remember: surveys capture expectations, not outcomes.

3:30 What's uncertain

And be honest about uncertainty. We don't know the pace of adoption across industries and countries. We don't know how much productivity turns into lower costs versus more output, or what happens to entry-level roles, career ladders and wages, especially as agents mature. That's why good planning uses scenarios, which we'll cover in the next module, rather than a single forecast.

3:57 Task-based redesign

Now the method. Map a role's main tasks and the time each takes. Classify each one. Automate, where AI does it and a human spot-checks: high-volume, rule-based, verifiable work. Augment, where AI drafts or analyses and a human decides. Or keep human: relationships, accountability, sensitive judgement, physical presence, creative direction. Then redesign the role around the freed time, and define the new skills people need.

4:25 Protect the pipeline, lead well

Don't forget the talent pipeline. If AI absorbs the junior tasks that used to teach people the job, where will tomorrow's experts come from? Design deliberate learning paths: juniors reviewing AI output against expert standards, working on real cases with AI as a coach, and rotating through the judgement-heavy tasks. And lead the change well: be transparent, involve staff in the mapping, follow consultation duties where they exist, and share the gains visibly.

4:57 Worked example: service centre

A customer-service operation in Riyadh introduced AI-drafted replies and call summaries in Arabic and English. With team leaders, it classified routine status queries as automate with checks, complaints as augment, and escalations as human. Freed time went into complex cases and a new proactive outreach service. Junior agents got a learning path using AI as a coach on past cases. And the team tracked quality, customer satisfaction and agent satisfaction, not just handling time.

5:29 Skills worth building

Which skills are worth building now? Beyond tool skills, three stand out. Evaluation: judging whether AI output is correct, which needs domain expertise. Delegation: breaking work into tasks an AI or agent can do well, with clear instructions and checks. And the human skills that grow in value when routine work shrinks: relationships, persuasion, leadership, ethics and creative direction. Build all three, not just prompt tricks.

5:58 Three mistakes

Three common mistakes. First, announcing headcount changes before understanding what actually changes at the task level. Second, quoting dramatic, unsourced statistics about job losses or gains, which damages credibility and trust. Third, ignoring entry-level learning: if AI absorbs the tasks that used to train juniors, you need to design new ways for them to build expertise.

6:22 Try this now

Try this now. Choose one role you know well, ideally your own. With the people who do it, list its main tasks and roughly what share of the week each takes. Label each task automate with checks, augment, or keep human, and write one sentence about what people would do with the time freed. Then use the skills-plan prompt in the lesson text to draft a ninety-day learning plan, with weekly two-hour blocks tied to real work. Review the plan together and adjust it. The conversation is as valuable as the document.

7:02 Recap

To recap. Analyse work task by task; augmentation is more common than full automation for most roles. Productivity gains are real but uneven. Treat surveys as expectations and plan with scenarios. Redesign roles, define new skills, protect junior learning and share gains. Your next step: complete the task-redesign worksheet in the lesson text for one role, with the people who do it, and build a ninety-day skills plan. Next module: building your horizon-scanning practice.