What does the myth say?
That an AI employee, unlike a person, needs no onboarding: connect it, and it performs at full quality from the first proposal. Under this belief, a first-week send-back is a defect, a first-month approval rate below the eventual one is a failure, and a manager who has to correct anything has been sold something unfinished. Pilots run under this belief end in week two, which is the week before they would have worked.
The examination is about what the first month actually is, and how to read it.
Where does the belief come from?
From demos, which show a calibrated system on curated inputs. From the reasonable expectation that software either works or does not. From the word employee without the corresponding expectation of a first month. And from vendors who let the belief stand because it closes a sale. It also comes from a real contrast with human hires, who take months to be useful; an AI employee that is useful on day one invites the inference that it is finished on day one.
Useful and calibrated are different, and the first month is the gap between them.
What is true in it?
That a Digital FTE is useful on day one in a way a human hire is not: the first-run wizard pre-fills the Brand Brain from your site and from what other FTEs already know, the first proposals arrive with evidence attached, and the first reconciliation, shortlist, or cited reply is real work. That there is no training period in the human sense, no shadowing, no ramp of months. That a role which is still producing substantive send-backs in week six has a remit or a memory problem the manager should fix rather than wait out. Day one is not nothing.
What is not true is that day one is the finished state.
What does the first month actually look like?
Week one: setup, the remit, the connections, and close reading of every proposal; send-backs are about facts the memory lacks and rules the remit did not state. Week two: the memory has been corrected, the remit tuned, and send-backs shift toward edge cases. Week three: send-backs are mostly about tone, which is the sign the substance has transferred. Week four: the first monthly review, reading the FTE's own first-person report, deciding targets and remit, and promoting nothing or one routine class within bounds. The approval rate rises through the month; the send-back reasons are the better signal because they say why.
The first cycle includes all of this and is labelled first-cycle in any comparison. The second review is the honest one.
- Week 1: facts and rules
- Week 2: edge cases
- Week 3: tone
- Week 4: review, targets, maybe one promotion
Why does the belief end pilots early?
Because a manager expecting perfection reads a week-one send-back as a verdict, or promotes a class in week two to prove the pilot is working and then experiences the step-back, or compares first-cycle cost per outcome against the baseline and concludes it did not pay. Each of these is the calibration period misread as the result. The playbooks in the library are time-boxed to thirty, sixty, or ninety days for this reason, and each ends with what done looks like, so a manager knows what to expect at each point.
A pilot that ends in week two has measured the setup, not the function.
How can you check for yourself?
Read the send-back reasons weekly and sort them: facts, rules, edge cases, tone. If the mix is moving down that list, calibration is happening on schedule. If facts are still dominant in week four, the memory has a gap to fill or the remit a rule to add, and the fix is specific. Read the FTE's own report at the first review and check its claims against the record. Then wait for the second review before comparing cost per outcome to the baseline, and label the first cycle as such in any note to a board.
The template for the monthly review note in the library gives the fields.