Digital collar careers: moving into directing and checking AI work

A digital collar career rarely starts with a new job title. It usually starts inside the job someone already has, when they begin handing parts of their work to AI agents and become the person who sets those tasks, checks the results and answers for them. This page covers how to make that move on purpose.

The short answer: start from the work you know best. Find the tasks an AI agent can do in your field, learn to brief it well and check it rigorously, and keep a record of what you built and what it changed. Domain knowledge plus a track record of supervising AI on real work is what makes the move credible.

Why the shift is happening

The International Labour Organization finds that one in four workers globally is in an occupation with some exposure to generative AI, with clerical jobs most exposed. Because most occupations still contain tasks that need human input, it concludes that transforming jobs is the most likely impact (ILO Working Paper 140, 2025).

Employers in the World Economic Forum's 2025 survey say the same in their own way: half plan to reorient their business in response to AI, two-thirds plan to hire talent with specific AI skills, and 40% expect to reduce staff where AI can automate tasks (WEF Future of Jobs Report 2025). The work moving toward people is the work of directing and checking.

Three common routes in

Inside your current role

The most common route. A bookkeeper starts supervising AI-prepared reconciliations, a support agent starts reviewing AI-drafted replies, a marketer starts running an agent that produces first drafts. The job title stays the same while the work changes.

Into a coordinating role

Someone who has set up agents for one team becomes the person other teams ask for help: writing agent jobs, setting review rules, tracking errors. This is often the step where the work becomes visible and gets a name.

Into oversight and quality

Roles focused on checking AI output, testing systems before launch, investigating incidents and making sure rules are followed. Under the EU AI Act, organizations using high-risk AI systems must assign oversight to people with the necessary competence, training and authority (EU AI Act, Article 26), which makes these skills a requirement in some fields rather than a nice extra.

The skills that matter

In the WEF survey, analytical thinking is the most sought-after core skill, considered essential by seven in ten companies. AI and big data top the list of fastest-growing skills, followed by networks and cybersecurity and technology literacy. Creative thinking, resilience, flexibility and agility, and curiosity and lifelong learning are also expected to rise in importance to 2030 (WEF Future of Jobs Report 2025).

For digital collar work these break down into:

• Domain depth: you cannot catch an AI's error in work you do not understand.
• Briefing: turning a goal into clear instructions, context and limits.
• Checking: verifying facts, numbers and edge cases, and knowing where a system tends to fail.
• Process design: deciding which steps run alone and which need a person.
• Quality control: the WEF lists it among the skills that most separate growing jobs from declining ones.
• Judgment and accountability: knowing when not to use AI, and standing behind the result.

Our page on digital collar skills goes into each in more detail.

Build a portfolio of real work

Certificates show that you studied. A portfolio shows that you can do the job. For each project, record:

1. The task: what the work was before AI was involved.
2. The setup: the agent's instructions, inputs and what it was allowed to do.
3. The checks: how you reviewed the output and what you caught.
4. The result: time saved, error rates, or quality changes you actually measured.
5. What you changed: the fixes you made after it went wrong.

Leave out anything confidential. A clear description of the method, with honest numbers, is worth more than a polished demo. Include the failures you caught: they show the judgment employers are hiring for.

Training: use what is offered

The WEF estimates that if the world's workforce were 100 people, 59 would need training by 2030. Employers expect 29 of them could be upskilled in their current roles and 19 upskilled and moved elsewhere in their organization, while 11 would be unlikely to receive the training they need. 85% of employers plan to prioritize upskilling their workforce, and 50% plan to move staff from declining to growing roles (WEF Future of Jobs Report 2025).

In the EU, employers that deploy AI systems must also support AI literacy among staff who use them (EU AI Act, Article 4). Ask what your employer offers, and ask to join the first AI project in your team.

A 90-day plan

• Weeks 1 to 4: list your tasks and mark the ones an AI tool could draft. Try one, and check every output carefully.
• Weeks 5 to 8: write a short job for that use: instructions, limits, how you check it. Measure how often it is wrong.
• Weeks 9 to 12: share the setup and results with your team, help one colleague adopt it, and write it up as your first portfolio entry.