Which jobs will shrink and which skills will be worth more in the company of 2029?

1. Analyze tasks before predicting job titles

A single “sales admin” position is really a dozen or more subtasks. Some are repetitive data entry; others involve calling an angry customer to smooth things over. If you judge the whole position at once as replaceable or not, you'll get it wrong either way.

A “sales admin” role might involve receiving leads, entering data into the CRM, preparing documents, scheduling appointments, chasing status updates and coordinating with customers. Some of these tasks can be heavily automated, while others depend on relationships. If a company assesses the whole position as either replaceable or not, it misses the chance to redesign the work.

Sort tasks into Routine Information, Variable Information, Physical, Relationship and Accountability. Reading, copying, categorizing, summarizing and producing documents are increasingly likely to be taken over by AI. Work that involves answering for the impact, negotiating, being physically present on site and building trust still needs people, though those people will get more input from AI.

2. Work likely to shrink or change shape significantly

The work that shrinks first tends to look alike: reading, copying, categorizing and passing things on. The work that stays is work that requires decisions in situations the manual doesn't cover.

Data administration, such as data entry, filing, format checks and compiling reports, will shrink as agents connect systems and handle a wider range of documents. Standardized content work, such as product descriptions, captions and summary reports, will take fewer people per piece, though someone still needs to set the direction and check the facts.

Status coordination, the job of asking who has got how far and passing information along, will increasingly be replaced by workflows. First-line support will shift from answering FAQs to handling cases that involve emotion, risk or exceptions. Junior analysis focused on gathering data and making slides may shrink in numbers, and newcomers will need new ways to learn through real work.

The shrinkage won't be the same in every industry. Work where the data isn't digital, the rules are complex or regulation applies may change more slowly. Owners need to use data from their own processes, and should never judge people against a list of job titles found on the internet.

3. Work and roles that will be worth more

Domain experts who can teach the system will be worth more than people who follow procedures without being able to explain why, because organizations need to turn knowledge into rules, examples and evaluations. Process designers connect business, technology and customer experience to eliminate handoffs.

Human skills such as analytical thinking, creativity, flexibility, leadership and collaboration still matter, because as drafts get cheaper to produce, the value moves to choosing which problems to solve, judging trade-offs and taking responsibility for outcomes.

4. Use a workforce matrix instead of across-the-board cuts

Have managers estimate how many hours each role spends in each group, then build scenarios in which 25, 50 and 70 percent of the work can be automated, and calculate the resulting capacity and the roles that need to be added. This shows the company both the positions that will shrink and the skill gaps that could block the change.

5. Key skills for people working in 2029

• Problem Framing: turning broad goals into tasks, criteria and constraints a system can understand
• Data Judgment: knowing which sources to trust, spotting missing data and never concluding more than the evidence supports
• Process Thinking: seeing work from start to finish and separating rules, AI and human decisions
• Evaluation: creating good examples, checking quality and classifying errors
• Exception Handling: making decisions when information is ambiguous or the stakes are high
• Customer Empathy: understanding context, trust and needs nobody has written down
• AI Safety Literacy: understanding confidential data, permissions, prompt injection and traceability

Typing prompts is only a basic skill, and tools will increasingly write prompts on their own. What lasts is understanding the business and judging whether a result “fits the context and creates value”.

6. Reskilling has to be tied to real workflows

A one-day general AI training gets people experimenting, but it doesn't change how they work. Pick a real workflow and have the team design the before and after, try out the system and take ownership of the KPI. Learners should use approved data, build evaluations and present the business results, so that what they learn becomes an asset for the company.

7. Build a scenario-based headcount plan

Set a hiring gate for routine positions: before an additional hire is approved, the team must show it has adjusted the process and assessed automation. For human-led positions, move quickly to retain people and build their AI skills, because this group will make the biggest difference.

8. A 12-month plan for owners and HR

1. Quarter 1: build a task inventory and workforce matrix for your key workflows
2. Quarter 2: pilot automation and augmentation in two processes, with training built on real work
3. Quarter 3: update job descriptions, KPIs, hiring gates and career paths
4. Quarter 4: build headcount scenarios for 2028 to 2029 from actual productivity results

In short: Routine information and coordination work is likely to shrink, while domain judgment, process design, customer trust and AI governance will be worth more. Companies should move from planning their workforce by position to planning it by task and workflow, with reskilling done on real work. This reduces both overhiring and the risk of running short of skills later on.