What does the myth say?
That the word employee is marketing, and underneath it is the same chat interface everyone has used: type a question, get an answer, copy it somewhere. The belief is reasonable because most AI products people have touched are exactly that, and because a Digital FTE does use a language model. Leaders who hold it expect a pilot to produce some helpful text and no change to how work gets done, and they decline on that basis.
The examination is about what a chatbot and a Digital FTE share, and what one has that the other does not.
Where does the belief come from?
From experience. The first generation of business AI was chat: a window, a prompt, an answer, and the human doing everything before and after. Vendors then attached the word agent or employee to the same window and changed little, which taught buyers that the words mean nothing. It also comes from the honest fact that a Digital FTE and a chatbot can share the same underlying model, so a buyer reasonably asks what else could be different.
The answer is that the model is the smallest part of an employee, in the way that a person's vocabulary is the smallest part of a job.
What is true in it?
That both use a language model to read and write. That a Digital FTE can be talked to, and that a lazy deployment of one, a remit that is never written, tools that are never connected, would behave much like a chatbot. That the word employee is a metaphor, and metaphors can mislead. And that some products sold as AI employees are in fact chatbots with a schedule, which is why the test at the end of this piece matters.
The myth is not stupid. It is a correct description of a large part of the market, applied to a product designed differently.
What does a Digital FTE have that a chatbot does not?
A remit: a written function it owns, with a scope, a never-do list, and targets it is accountable for. A schedule: it works every day whether or not anyone opens a window, and reports in the first person in a morning brief. Tools: dedicated, scoped, encrypted connections to the systems the function runs in, a ledger, a helpdesk, a CRM, a site, through which it reads evidence and, on approval, acts. A gate: every proposal passes through the approval gate and the autonomy ladder, with the audit trail recording the decision. A memory: the Brand Brain, with provenance on every item, confirmed by the manager. And a manager: the Performance tab, the monthly review, the send-back reasons, the targets.
A chatbot has none of these. It has a window.
- Remit, schedule, tools, gate, memory, manager
- Reports unprompted; is measured; is reviewed
- Shares only the model
Why does ownership change the economics?
Because a chatbot saves minutes on tasks a person still owns, and the person still has to remember to do the task, open the window, and carry the answer somewhere. A Digital FTE owns the function, so the task happens whether or not anyone remembers, the evidence is gathered from the tools, and the person's involvement shrinks to the approval and the review. The cost per outcome is computed against a baseline because there is an outcome to count; a chatbot's value is diffuse and usually unmeasured, which is why chatbot pilots rarely survive a budget review.
The whitepaper on the economics of digital labor works this through with a labelled example.
How can you tell them apart in a demo?
Ask what it did yesterday without being asked; a Digital FTE opens a morning brief with items, a chatbot has nothing to say. Ask to see the remit and the targets. Ask which tools it is connected to and with what scope, and ask to disconnect one. Ask to see a proposal with its evidence and the approval it is waiting for. Ask to see the memory and the provenance of one item in it. Ask what it will do tomorrow if nobody logs in. A product that can answer all six with something on the screen is an employee in the sense that matters; one that answers with a chat window is a chatbot with a new name.
The comparison of Digital FTE, AI agent, and chatbot in the library gives the three-column version.