Plan for Scenarios, Not One AI Future

TL;DR

• AI is changing tasks faster than job titles. Plan around the work you expect to do, not a confident forecast that one role will vanish or dominate.
• Build a portfolio that performs across several futures: product judgement, AI fluency, domain depth, evidence of recent shipping, and the ability to redesign workflows.
• Review your direction quarterly. Tool capability moves quickly; organisational adoption, regulation, incentives, and career markets move at different speeds.

---

The dangerous career plan is the one that depends on a single AI forecast being right.

“PMs disappear.” “Designers win.” “Everyone becomes a builder.” “Agents replace management.” Each claim may describe part of the market. None is stable enough to anchor a five-year career.

AI changes tasks, workflows, team shapes, and hiring signals at different rates. Career strategy should prepare you for several plausible operating models while building capabilities that remain useful across them.

Start with tasks, not titles

A job is a bundle of tasks held together by an organisation. AI affects the tasks unevenly.

Break your role into four categories:

Writing a first draft may compress. Deciding which market to enter may become better informed but remains accountable human work. Operating agent permissions and evals may become a new responsibility. Persuading a regulated customer may change slowly because institutions adopt differently from tools.

Titles hide this redistribution. Task analysis exposes it.

Model adoption, not only capability

Frontier capability is one input into career change. Deployment depends on data, integration, trust, economics, regulation, management quality, and whether the workflow is worth redesigning.

That creates uneven adoption:

• An AI-native startup may rebuild roles within months
• A large enterprise may automate isolated tasks while preserving its org chart
• A regulated business may add review and governance work before reducing any role
• A team with weak management may increase output without changing outcomes

Do not assume the most advanced company describes the median employer next year. Do not assume the median employer protects you for five years either.

Career optionality comes from understanding both the frontier and the adoption system around it.

Use three planning horizons

Immediate horizon: the next six months

Improve the work you do now. Identify one repeated task to compress, one judgement-heavy responsibility to deepen, and one visible system to build for others.

The aim is evidence, not tool familiarity.

Next horizon: six months to two years

Choose roles that increase the density of useful learning. Look for direct customer contact, access to production systems, accountable decisions, strong peers, and room to ship end to end.

Title and compensation matter. They should be weighed against whether the role produces current capability and credible evidence.

Scenario horizon: three to five years

Do not define one skip role as destiny. Define three plausible scenarios and the capability portfolio each rewards.

For example:

1. AI-native compression: small generalist teams own broad outcomes.
2. Enterprise augmentation: existing functions remain but use agents, platforms, and governance extensively.
3. Specialist resurgence: cheap general output makes deep design, domain, systems, or risk expertise more valuable.

Choose development moves that perform in at least two scenarios. The best moves often perform in all three.

Build a durable capability portfolio

Five capabilities survive a wide range of futures.

Product judgement

Knowing which problem, standard, form, and trade-off deserve commitment becomes more valuable when implementation options multiply. Taste makes this practice concrete.

AI fluency

Use models and agents on real work. Build systems others depend on. Understand evaluation, context, permissions, cost, and failure. The AI fluency spectrum distinguishes personal usage from organisational impact.

Domain depth

Property, finance, health, legal, logistics, construction, and other domains contain constraints that generic generation does not erase. Combining current AI practice with real domain judgement is scarcer than either alone.

Recent evidence

Maintain examples from the last year that show what you framed, built, evaluated, shipped, and learned. Evidence decays as tools and practices move.

This is not a demand to code every week. Senior leaders can show recent operating changes, eval systems, product decisions, platform improvements, or production outcomes they directly shaped.

Workflow redesign

Personal productivity is useful. The larger career signal is the ability to change how a team or function works while managing adoption, risk, workload, and accountability.

Audit the skill that fills your calendar

Senior people accumulate work that once made them valuable: status synthesis, roadmap coordination, polished documents, executive alignment, or repeated review.

Some of that work remains important. Some persists because the organisation is familiar with it.

Run a calendar audit:

• Which repeated task could be compressed materially?
• Which meeting exists to transfer context that a better system could preserve?
• Which approval depends on your judgement, and which depends only on your position?
• Which activity develops future capability?
• Which work are you defending because you are good at it?

Remove or redesign one low-value obligation before adding skill development. A career plan that depends on permanent nights-and-weekends effort is not durable.

Modernity and pedigree are both evidence

Recent AI-native work is a stronger signal than brand alone. Brand and tenure can still represent scale, craft, operating maturity, and domain exposure.

Treat both as evidence, not verdicts.

When evaluating a candidate or your own positioning, ask:

• What did they personally learn and change recently?
• Can they explain a current production system in detail?
• Have they operated through failure, cost, adoption, and governance?
• Does older experience supply depth that the current market still needs?
• Is their practice moving, or did it plateau?

The useful distinction is not old company versus new company. It is current capability versus inherited reputation.

Make career decisions with scenarios

Stay or leave

Stay when the role produces learning, evidence, relationships, and domain depth relevant across your scenarios. Leave when the role repeatedly blocks those things and you cannot redesign it from inside.

Do not use a universal percentage-of-time threshold. A quarter spent on a difficult regulated launch may build more durable capability than a quarter spent producing many low-risk prototypes.

Take a title cut

A title cut can be rational when it buys direct exposure to stronger practice, production ownership, or a valuable domain. It is not automatically brave or strategic. Model the work, manager, runway, and path back to appropriate scope.

Pivot industries

Combine AI fluency with a domain where consequences and workflows matter. Avoid pivoting only because an industry is fashionable. Ask whether you are willing to learn its data, customers, regulation, economics, and operational reality.

Build independently

Building a product can compress learning across discovery, implementation, pricing, support, and operations. It also carries financial and personal risk.

Use it when the downside is acceptable and the work creates evidence you cannot obtain in your current role. A failed business can still produce valuable capability. It should not be romanticised as the only path to modernity.

Run a quarterly career review

Every quarter, answer:

1. Which tasks in my role changed materially?
2. Which scenario became more or less plausible?
3. What did I ship, operate, or redesign?
4. Which judgement did I deepen?
5. Where did my practice depend on one vendor or tool?
6. What low-value work did I remove?
7. Is my pace sustainable?
8. Which next move would improve at least two scenarios?

Keep the answers concrete. “Used AI more” is not movement. “Built a shared evaluation workflow used by three teams and reduced manual review on low-risk cases” is.

Anti-pattern: career planning by headline

A leader reads that product roles are collapsing. They accept a smaller role at an AI-native company without examining the work, manager, economics, or learning environment.

The company uses newer tools but offers little customer contact, weak production discipline, and no specialist depth. The move optimised for the headline rather than the capability portfolio.

The opposite mistake is staying in a comfortable role because adoption appears slow. Capability eventually reaches the workflow, and the person has no recent evidence when the market moves.

Scenario planning avoids both. Stay close to the frontier, understand the adoption lag, and build capabilities that remain valuable when the forecast changes.