AI in the operation · Tooling

Size the tooling budget

Turn three one-week tool trials into the most each tool is worth a year in saved hours, then choose which one to fund first.

estimateeasyExercise · 3 min · 52 of 60

On this page

Three tools have been trialled on the team's own work, and nobody has asked a vendor for a price yet. The estimate is how much each one may cost before it stops paying.

The situation

A five-person marketing team is proposing next year's tooling budget: three requests, one from each of three capability classes, each trialled for a week on the team's own work. Use the course's $100 marketing hour and 52 weeks a year.

Generation, for paid-social ad variants. The team writes 120 a month at 20 minutes each. In the trial, review took 6 minutes a variant, and 30% of variants still needed a full rewrite at the old 20 minutes.
Enrichment, for the 300 leads a month that sales calls. Marketing operations researches each by hand in 6 minutes. The tool matched 70% of leads; a matched lead needs 1 minute of checking, and the rest are researched by hand as before.
Analysis, for the weekly performance report: 4 hours a week by hand, against 1 hour a week checking the tool's draft.

The question

Estimate the most the team should pay a year for each tool, and say which one you would fund first.

A strong answer · Rubric · Where answers go wrong

Write your answer first, then open

A strong answer

The most a tool is worth is the human time it saves, net of the time it creates.

generation   by hand    120 x 20 min                           = 40 h a month
             with tool  120 x 6 min + 36 rewrites x 20 min     = 24 h
             saving     16 h x $100 x 12                       = $19,200 a year
enrichment   by hand    300 x 6 min                            = 30 h a month
             with tool  210 x 1 min + 90 x 6 min               = 12.5 h
             saving     17.5 h x $100 x 12                     = $21,000 a year
analysis     by hand    4 h a week x 52 / 12                   = 17.3 h a month
             with tool  1 h a week x 52 / 12                   = 4.3 h
             saving     13 h x $100 x 12                       = $15,600 a year
total                   about 46 h a month                     = $55,800 a year

These are ceilings: a tool priced at its ceiling leaves the team exactly where it started. A one-week trial on one batch is the optimistic case, so budget at most half of each ceiling, about $28,000 in all. The hours are only worth $100 if they go somewhere: 46 hours a month is about a quarter of one person, and it saves cash only if it replaces a hire or moves to work that earns.

Fund enrichment first. It saves the most, and its failure is cheap and visible: an unmatched lead falls back to the six-minute manual path. Generation's failure is an unsubstantiated claim in a live ad, which costs more than time and keeps the 6-minute review in place however good the tool gets. The report saves the least.

What would change the order: enrichment's match rate. At 40% its saving falls to 300 × 40% × 5 minutes = 10 hours a month, $12,000 a year, and generation moves first. Generation's saving vanishes if 70% of variants need a rewrite.

Rubric

Converts each trial into hours a month by hand and with the tool, the arithmetic shown
Counts the work the tool creates (review, rewrites, unmatched leads, checking), not only the time it removes
States the three ceilings at $100 an hour: about $19,200, $21,000 and $15,600 a year
Distinguishes a ceiling from a budget and gives a reason to pay below it
Chooses one tool to fund first, with a reason that includes its failure mode
Names the measured rate that would reverse the choice

Where answers go wrong

"Generation saves 40 hours a month, because the tool writes the variants." It counts the gross time and forgets 24 hours of review and rewrites, overstating the saving two and a half times. The week-long trial in tooling literacy without vendor worship exists to catch exactly that.

"Spend the full $55,800; the tools pay for themselves." At the ceiling they pay for themselves and nothing more, and hours saved but not redeployed return no cash at all.

"Fund the report first, because the founder reads it every week." Visibility is not a number. The report has the smallest saving of the three, and the choice should rest on hours and on what each tool's failures cost.