They will replace tasks far more often than whole roles. Gathering, keying, matching and first drafts move to agents. Reviewing, deciding exceptions, owning outcomes and dealing with customers in difficulty stay with people, because accountability does. Roles built on routine handling will shrink. How far depends on the depth agents earn, and that is decided on evidence, not forecasts.
What decides how much work moves?
Depth. An agent at depth 1 reads and prepares; people do the rest. At depth 2 it drafts and a person sends every item, so people still touch every case, but faster. At depth 3 it completes routine work within bounds, and people handle what is held, escalated or sampled. Only then does the volume of work for people fall sharply, and only for the work whose depth the agent has earned.
The second factor is accountability. Regulated decisions, customer harm and money movement all need someone who answers for them. An agent can prepare the case for a credit decision or a complaint outcome, but a named person decides, and regulators expect that. Work that carries accountability does not disappear; it concentrates.
Which tasks move, and which stay?
| Task in an operations team | Likely to move to agents | Stays with people |
|---|---|---|
| Collecting documents and data for a case | Yes | Checking unusual items |
| Re-keying between systems | Yes | No |
| Matching, reconciling, listing breaks | Largely | Investigating and posting complex breaks |
| First drafts of letters and notes | Yes, at depth 2 | Sending, and anything sensitive |
| Routine updates within limits | At depth 3, once certified | Sampling and held items |
| Exceptions and judgement calls | Preparation only | Deciding |
| Customers in difficulty or vulnerable | Preparation only | The conversation and the decision |
| Owning the process and the agent | No | Always |
The right column grows in importance as the left column moves. Teams spend less time typing and more time judging. That shift is real, and so is the fact that a team doing mostly judgement needs fewer people than one doing mostly typing.
Who answers for the outcome?
Management decides staffing; the process owner answers for what the agent does. Those are separate decisions and should be made separately. Reducing a team because an agent was expected to reach depth 3, before it has, leaves nobody to absorb the work when a certification run clears no depth and the agent drops to depth 1 the same day. Staffing should follow evidence of the depth an agent actually holds, not the business case's hopes.
Resistance in teams is often less about job loss than about who carries the blame when an agent gets something wrong. Our view is that staff resistance is usually about accountability, not jobs, and it is answered by making ownership and the held queue clear.
What does the change look like in a real team?
Say a bank's payments operations team of 30 handles payment investigations, repairs and customer queries. An agent takes on gathering payment records and drafting investigation replies at depth 2, then completes simple repairs at depth 3 a few months later.
Over the next year, volume rises with new business, and the team absorbs it without hiring. Three people move into a new role reviewing held repairs and sampling the agent's work. Two become owners of the team's agents and briefs. Natural leavers in routine roles are not replaced, and twelve months on the team is somewhat smaller. Nobody's job was taken by an agent on a single day; the shape of the team changed.
The same pattern shows up in finance and support, covered in whether AI agents will replace accountants and whether an AI support agent will replace human support staff.
What should leaders do differently?
- Tie staffing changes to depth held, not depth planned.
- Create the new roles early: reviewers, agent owners, exception specialists.
- Train people for judgement work, since that is where the team's value moves.
- Keep the fallback staffed, so people can take work back when an agent is reduced.
The full deployment sequence is in our guide to deploying an AI agent.
Questions people ask
Will entry-level operations roles disappear?
Many entry-level tasks will, which changes how people learn the job. Firms will need to train newcomers on review and exceptions deliberately, rather than by years of routine handling.
Do agents reduce the need for experienced staff?
Usually the reverse. Experienced staff write the briefs, set answer keys, review hard cases and own agents. Their knowledge becomes more valuable, not less.
Should we tell staff about agent plans early?
Yes. Explain which tasks will move, which roles will grow and how depth will be decided. Uncertainty does more damage than a clear plan.