# Chapter 09: Roles, skills and new ways of working

> The technology is ready before the people are. Companies that get value from agents invest as much in roles, skills and habits as in the agents themselves.

The AI-Native Playbook for Enterprises · Twistag · 7 min read · Updated: 2026-10-09

## In one minute

1. **Every role changes a little, some change a lot.** Most people will work alongside agents. A smaller group will own, run and improve them. Both need different training.
2. **Teach delegation along with tools.** The core skill is handing work to an agent well. That means a clear brief, a way to check the result and the judgement to know when to step in.
3. **Leaders go first.** Teams adopt new ways of working when their managers visibly use them, measure them and reward them.

## People are the slower part

An agent can be built in weeks. Changing how a team works takes longer. People need to trust the agent, learn to check its work, and find a new rhythm for their day. Managers need to rethink targets and roles.

The scale of change is large. The World Economic Forum's Future of Jobs Report 2025, based on more than 1,000 employers in 55 economies, expects 39% of workers' core skills to change by 2030. About half of employers plan to reorient their business in response to AI. BCG's 2026 index expects dedicated AI roles to grow from 7% of the workforce in 2026 to 22% by 2030, and companies in its survey expect a workforce reduction of roughly 10% to 15% by the same year.

How a company handles this decides whether its agents get used. An agent nobody trusts is an agent nobody uses.

## Three groups, three needs

Agents change work in different ways for different people.

**People who work alongside agents.** Most of the workforce. They review drafts, handle flagged items, ask agents for help and spend more time on the parts of the job that need judgement or a relationship. They need practical training on their own tools and processes, and confidence that the agent is there to help.

**People who own, run and improve agents.** A smaller group: business owners, operators and the engineers described in Chapter 7. They need deeper skills: writing clear instructions, building test cases, reading the numbers, deciding when an agent is ready for more room.

**Leaders.** They decide where agents go, approve the business cases and answer for the results. They need enough hands-on experience to judge what agents can and cannot do, and the vocabulary to hold teams to account.

## The skills that matter

The skills that make people effective with agents are less technical than most expect.

- **Delegation.** Giving an agent a clear task, with the context, the constraints and what good looks like. The same skill that makes a good manager.
- **Checking.** Reading an agent's output quickly and critically, spotting what is wrong or missing, and knowing which parts to verify against sources.
- **Judgement on when to step in.** Recognising the cases an agent should not handle alone, and escalating them.
- **Process thinking.** Seeing a piece of work as steps that can be split between people and agents, as in Chapter 2.
- **Working with data.** Asking good questions of data, and knowing when an answer looks wrong.

Microsoft's 2025 Work Trend Index, based on a survey of 31,000 people in 31 countries, describes this new role as an "agent boss": someone who builds, delegates to and manages agents to increase their own impact. McKinsey's research suggests that people without technical backgrounds can learn to manage agent workflows about as quickly as trained engineers.

Good tools lower the bar further. Datatalks, the customer data platform Twistag has engineered for six years, gives marketers at more than 50 sports organisations a visual segment builder, a Segmentation Assistant that suggests segment logic from goals written in plain language, and an assistant that answers data questions without SQL. People who never wrote a query now work with more than 6 million unified fan profiles.

Where we saw it: [Unifying fan data across 50+ sports organisations](https://twistag.com/case-studies/datatalks-customer-data-platform-sports) (Datatalks)

## Roles that appear

As agents spread, a few roles appear in most companies. In smaller companies, one person often holds several of them.

| Role | What it does | Often comes from |
|---|---|---|
| Agent owner | Answers for an agent's results, budget and controls | The manager of the process |
| Agent operator | Watches the numbers, reviews samples, handles feedback | Experienced people from the process team |
| AI engineer | Builds and changes agents, prompts, rules and integrations | Software and data engineering |
| AI governance lead | Runs the register, the tiers and the reviews | Risk, compliance or legal |
| AI programme lead | Keeps the short list, the business cases and the reporting | Strategy, transformation or operations |

The best agent operators are usually people who know the process well. They spot the wrong answers fastest. Moving them from routine handling into operating the agent keeps their knowledge in the business.

## Bringing people with the change

Adoption usually stalls for human reasons. People do not understand why the change is happening, fear what it means for them, or find the new tool harder than the old way.

What works, in Twistag's experience across client programmes:

- **Involve the team from the first week.** The people who run the process help map it, test the agent and shape the review screen. They own part of the result.
- **Be honest about what changes.** Say clearly which tasks move to agents, what people will do instead, and what support exists for those whose roles change most.
- **Train on real work.** Short sessions on the team's own process, with their own documents, beat general AI courses.
- **Make the review fast.** If checking an agent's work takes longer than doing it, people will stop checking or stop using it.
- **Share the results.** Hours saved, errors avoided and better outcomes, shown to the team that made it happen.

## Leaders go first

Teams copy what their leaders do. When executives use agents in their own work, talk about it, and ask for AI in every proposal they review, adoption follows. When AI is something leaders approve for others, it stays on the edge.

Three habits help:

- every member of the leadership team uses an agent in their own weekly work;
- every business case for a new process or system asks where agents fit;
- targets and reviews recognise work done with agents, including the time spent improving them.

Twistag works this way internally. Its engineers work with delivery agents every day, and the review layer around them records every override, so leadership can see where people trust the agents and where they step in.

## AI literacy as a baseline

Under the EU AI Act, companies that use AI systems have had a duty since February 2025 to support AI literacy among the staff who work with them. The 2026 changes softened the wording, but the direction is clear. A short, role-based baseline for everyone, covering what AI can and cannot do, how to check its work and how to handle data safely, is both a good practice and a sensible way to meet that duty.

## Common mistakes

- **Training on tools only.** People learn which buttons to press, but not how to delegate or check.
- **No time to learn.** Training added on top of a full workload, so nobody does it.
- **Owners in name only.** A manager listed as an agent owner who never looks at its numbers.
- **Silence about jobs.** People hear about changes to their roles through rumour, and resist the agents that could help them.
- **Leaders who delegate AI.** The leadership team approves agents for others and never uses one.

## What to do this month

1. For each agent planned or live, write down how the roles around it change, and share it with the team.
2. Add agent owner, operator and engineer responsibilities to the relevant job descriptions.
3. Run a short, role-based AI literacy session for every team that works with an agent.
4. Ask each member of the leadership team to use an agent in their own work for a month, and report back.

## Questions for the leadership team

1. Which roles in the company will change most as agents take on routine work, and what will those people do instead?
2. Who owns, runs and improves each agent today, and are those responsibilities in their job descriptions?
3. What share of the leadership team uses AI agents in their own work each week?
4. How will the company support people whose work changes the most?
5. How do performance reviews and targets reflect work done with agents?

## Sources

1. [Applied AI Index 2026](https://www.bcg.com/press/30september2026-ai-starting-to-pay-off-companies-generate-value), Boston Consulting Group, 2026-09
2. [Future of Jobs Report 2025](https://www.businesstoday.in/wef-2025/story/39-of-skills-outdated-by-2030-wef-warns-of-ais-impact-on-jobs-460306-2025-01-10), World Economic Forum, 2025-01
3. [2025 Work Trend Index: The year the Frontier Firm is born](https://www.uctoday.com/collaboration/frontier-firms-and-agent-bosses-microsoft-outlines-its-future-of-work-vision/), Microsoft, 2025-04
4. [The agentic organization: Contours of the next paradigm for the AI era](https://www.mckinsey.com/capabilities/people-and-organizational-performance/our-insights/the-agentic-organization-contours-of-the-next-paradigm-for-the-ai-era), McKinsey & Company, 2025-09

---

Source: https://twistag.com/guides-and-playbooks/ai-native-playbook/chapter-09