Most organisations still discuss the future of work through the language of job titles: which roles will grow, which will shrink, and which are most exposed to AI. Titles are useful for pay, accountability, and organisation design. They are a poor unit for understanding how technology changes work. AI does not encounter an “HR Business Partner”, a “Financial Analyst”, or a “Legal Counsel” as a single object. It encounters activities: gathering information, comparing documents, drafting a response, checking a policy, explaining an option, making a judgment, and coordinating with another person. A job is a portfolio of those activities, performed in a particular context and held together by accountability.

That distinction matters because two roles with the same title can have very different AI futures. One analyst may spend much of the week preparing recurring reports from structured data. Another may spend most of the week interpreting ambiguous evidence with senior stakeholders. The first has a significant automation opportunity; the second may be strongly augmented by better search, synthesis,and drafting tools, while remaining deeply human-led. Calling both roles“analyst” tells us almost nothing useful about the work that will change.

A practicalactivity map normally separates three broad forms of work:

• **Transactional activity:** entering, validating, routing, reconciling, scheduling, andproducing standard outputs.
• **Cognitive activity:** researching, reading, analysing, drafting, comparing, diagnosing,and deciding.
• **Relational activity:** influencing, coaching, negotiating, reassuring, building trust, and aligning people.

The categories are not a hierarchy. All three can be valuable. But they respond differently to technology. Transactional work is often the most suitable for rules-based automation. Cognitive work is where generative AI is already becoming embedded:it can prepare a first draft, retrieve relevant information, identify patterns,or create options for review. Relational work may be supported by better preparation, but the human relationship and accountability generally remain central.

This is why headlines about “jobs at risk” are not enough for workforce decisions. A role may contain a few highly automatable activities that are peripheral to its value. Removing or reducing those activities might make the role more valuable,not less. Conversely, a role with no single dramatic automation candidate may still change substantially when many small cognitive activities become faster or easier.

For HR, thepractical questions are therefore more precise:

• Whichactivities consume time today?
• Whichactivities will be automated, augmented, or largely unchanged?
• Whichactivities become more important when routine effort falls?
• Which skills,capacity assumptions, and career paths follow from that change?

This approachalso leads to better conversations with employees. “Your job is 40% exposed” isvague and threatening. “The first drafting, information retrieval, and routinechecking parts of your work will change; your judgment, stakeholder advice, andfinal accountability will become more important” is specific enough to discuss,test, and prepare for.

The future ofwork is not an event that arrives at the edge of the organisation. It isalready arriving inside ordinary activities. The organisations that understandthose activities first will be better placed to redesign work, build skills,and choose technology with intent.

**Next article:**If AI changes activities, why will so much of the early impact be augmentationrather than end-to-end automation?