Step 1 (day 1): decide what the FTE will own, so you know what to record
The baseline records the work the Digital FTE will take on, so start from the draft owns list of its remit: reconciliation and exception handling, first-line replies in one category, screening for one role, outreach and CRM hygiene, page production. List the tasks in that list as the rows of the record, and add a row for everything else, so the people logging do not have to decide what counts.
This takes an hour with the people who do the work and doubles as the first pass at the remit.
Step 2 (days 1–2): set up a sheet that takes ten minutes a day
Make one shared sheet: a row per task from step one, a column per working day, and a cell for hours. Add a second small table for outcomes: transactions reconciled, tickets resolved at first line, applications screened, outreach sent, pages published, counted per day in the units the FTE's Targets will use. Ask each person to fill their rows at the end of each day, in ten minutes, in round quarter-hours.
Precision to the minute is not the point; consistency for two weeks is.
Step 3 (days 1–14): capture the quality measures you already have
Alongside hours and outcomes, capture the quality measures that already exist, the same way they will be measured after the pilot: days to close and exceptions open on day one from the last two closes; first-line resolution and median first response from the helpdesk; time to shortlist from the last two hiring rounds; share of opportunities with a current next step from the CRM; console and sampled citation data for target queries. Where a measure does not exist, write none rather than inventing one.
The method matters more than the number: record how each measure was taken, so the after measurement can copy it.
- Hours by task, per person, per day
- Outcomes per day in the Targets' units
- Quality measures with the method used to take them
- The rate or cost per hour for each person, from finance
Step 4 (day 7): a mid-point check
At the end of the first week, look at the sheet with the team for two things: rows that are always empty, which are tasks the list overstated, and hours landing in the everything-else row that belong to a task, which is a row the list missed. Adjust and continue. A mid-point check keeps the second week honest and saves the whole record from a misunderstanding discovered at the end.
Resist the temptation to start the FTE early because the first week looked clear; a two-week baseline covers the week that was unusual.
Step 5 (day 14): total, price, and store
Total the hours by task and price them at each person's rate from finance; that is the baseline cost of the work per fortnight, which scales to a month. Total the outcomes; that is the baseline output. Attach the quality measures with their methods. Store the whole record with the date and the names of the people who logged it, alongside the remit it was built from.
Write one line of caveats: the week that was unusual, the task that was hard to separate, the measure that does not exist. Caveats recorded now are honesty later.
Step 6 (after two cycles): the comparison
After the Digital FTE's second full cycle, take the outcomes from Targets, the tokens from usage per role at the plan's rate, the logged human hours on approval and setup, and the same quality measures taken the same way, and set them against the stored baseline. Monthly value is the baseline cost of the outcomes now produced; monthly cost is the plan share plus human hours; the difference and the setup cost give the payback.
Because every input traces to the sheet, the workspace, or a logged hour, the comparison can be audited line by line by anyone who asks.
What does done look like?
Done is a dated record of hours by task, outcomes in the right units, quality measures with their methods, prices from finance, and a line of caveats, made in two weeks with ten minutes a day from the people who do the work. It is the least exciting document in the pilot and the one every honest number depends on. A finance leader who has it can sign a decision; one who does not can only agree with a feeling.
The template in the ftes.ai library gives the sheet's rows; the two weeks are yours.