Why is replacement the wrong frame?
Replacement is the wrong frame because a role is a bundle of different kinds of work, and an AI employee suits some kinds and not others. A finance controller's role contains reconciliation, which a Digital FTE does well, and accounting judgement, which it does not. Asking whether the AI replaces the controller misses the actual choice.
The actual choice is which work goes on which budget line. That framing produces better decisions and, not incidentally, better conversations with the people whose work is being discussed.
How do you split the work?
Split each function's work into two groups. The first is judgement and relationship work: decisions with defensible alternatives, conversations with customers and colleagues, and anything that changes a person's outcome. This stays on the headcount line, and the ftes.ai platform permanently gates it anyway.
The second is repetitive evidence-based work: matching, screening against criteria, drafting from sources, monitoring, and reporting. This is the work a Digital FTE can own, with approval, and it can move to a metered AI budget line.
- Headcount line: judgement, relationships, decisions about people, sign-off
- AI line: reconciliation, screening, drafting, monitoring, reporting
- Shared: approval, which is a few minutes of human time per proposal
How do you compare the two lines honestly?
Compare each line on what it buys, using recorded baselines. For the AI line, record the outcomes the Digital FTE produces and the tokens it consumes, visible per role in the workspace. For the headcount line, record what the same people now spend their time on, and whether that work is higher value.
Refuse assumed savings. Any analysis that starts from a percentage of hours saved without measuring them is a forecast dressed as a fact. Run a pilot, measure, and let the comparison come from the record.
What does the decision usually look like?
The usual result is not a smaller team but a redeployed one: the same people doing the judgement and relationship work that was crowded out by the repetitive work, plus a new metered line that pays for the repetitive work by usage. The headcount line changes shape more than size in the near term.
Where a role was going to be added for repetitive work, the decision is different: the AI line can absorb that work instead, and the hire may not be needed, or may be a different hire. That is the case where the AI budget genuinely substitutes for headcount, and it should be assessed on the same recorded basis.
What should the budget line contain?
The AI budget line should contain the plan cost, which on ftes.ai is a block of work metered in tokens for the whole organization, plus the human approval time, which is small per proposal but real, plus any integration or setup time in the first month. It should not contain assumed savings; those belong in the review, after they are measured.
Budget the line per function with a per-FTE budget in the workspace, so that spend is visible by role and can be governed the way any other line is. The spend governance guide on ftes.ai covers the controls.
What does the conversation with the team sound like?
The conversation with the team should be about the work sort, not the budget line. It starts with the two-week time record the team kept, sorted into work with a checkable right answer and work that needs judgement or a relationship, and it names what each person will own once the first column moves to a Digital FTE: the exceptions, the escalations, the interviews, the decisions. It says plainly that decisions about people stay with people, because the platform gates them permanently and because that is the organization's own commitment.
It also says what will be measured: the FTE on its targets and approval rate, the people on the judgement work they now own, and never the people on volume the FTE handles. A finance leader who frames the headcount question as allocation has an honest conversation available; one who frames it as replacement does not, and the people will hear the second framing whatever words are used. The change management guide on ftes.ai covers what to say and when, and the hybrid team design describes what the roles look like afterwards.