A Workplace Field Guide

The verification-tax calculator

The relevant comparison is never "AI draft vs. nothing." It's "AI draft plus verification time vs. doing it right the first time." Put in your numbers and get a weekly figure you can put in a memo.

Runs in your browser Nothing is stored or sent Inputs live in the URL — bookmark a scenario

Enter your numbers

Use real observations from your own work, even rough ones; the toolkit's incident log is built for gathering exactly these. If you have to estimate, estimate conservatively in the tool's favor — a net-negative result from generous assumptions is far harder to argue with.

How many times per week this AI-assisted task happens.
Drafting time the tool removes, before any checking.
Verifying facts and figures, fixing errors, re-prompting, reformatting.
How often a mistake survives review and reaches a client, court, regulator, or system.
Rework, make-goods, refunds, billing write-offs, or a fraction of a worst-case incident.
Salary plus benefits and overhead for whoever does this work. A common shortcut: annual salary ÷ 1,000.
License and seat fees attributable to this workflow. A monthly per-seat price × seats ÷ 4.3 gets you close.
The real weekly ledger
Net hours / week
Labor value / week
Expected error cost / week
Net impact / week
Net impact / year (50 wks)
Verification tax (% of gross savings)
Enter your numbers above.

Reading the result honestly

The model counts four things: the drafting time the tool removes, the review time it adds, the expected cost of errors that slip through, and the subscription itself — the visible cost that, as the guide's fourth risk category notes, is only the start of the ledger. A net-positive number doesn't end the conversation; it means the tool survives the economics and the discussion can move to the other risk categories — confidentiality, liability, client expectations. A net-negative number is your headline: lead the memo with it, show your inputs, and invite the manager to challenge the assumptions. Inputs they adjust themselves become numbers they believe.

Two cautions. First, the slipped-error rate is the input people most want to set to zero; the incidents library exists to show why that's not realistic even with diligent professionals reviewing. Second, this calculator measures one workflow, not the tool in general. A tool can be net-positive for internal brainstorming and net-negative for client deliverables at the same desk, which is exactly the argument for scoped use rather than blanket adoption.