What does "Tested by AI Savings Lab" mean?

We built or ran the setup ourselves using test accounts and synthetic data, followed the written test plan, and recorded dates, time spent, costs and errors. It does not mean a certified or third-party lab test.

Where do prices come from?

Vendor pricing pages (or the vendor's own announcement when the page cannot be read), with the link and the date we checked. Case studies use what the owner actually paid.

Testing methodology

This page explains exactly what our labels and numbers mean. We would rather leave a field blank than guess.

Verification statuses

Not yet tested: Researched only. Prices are sourced, but no setup has been run. Net savings, setup time and success rates are blank.
Researched (vendor claims): We have recorded what the vendor says about the product (features, pricing), with links. Nothing has been verified by us.
Tested by AI Savings Lab: We ran it ourselves with test data according to the written test plan. Results show dates, number of runs, time, costs and errors.
Case study (owner-documented): A real owner documented their own setup with us: what was replaced, costs, time and oversight. Their numbers are reported as theirs, and we say which ones we could check.

How we test a replacement

Write the test plan first: the task, what "done" means, the test data, and the number of runs.
Use test accounts and synthetic data. No real bank, client, tax or password data is uploaded anywhere.
Keep humans in the loop. Anything that sends messages, moves money, cancels a service or files something needs a human's approval.
Record everything: dates, setup time, time per run, interventions, errors, costs.
Score it on our standard scorecard.
Re-check when prices or products change; every page shows its last-reviewed date.

Savings math

Gross savings = what you stop paying. Net savings = gross − AI cost − value of maintenance time; the first year also subtracts setup time and one-time costs. The formulas are in the calculator and this explainer. Where we show "potential gross" in the directory, it is the vendor list price × 12, clearly labeled as an estimate.

Human oversight levels

Each directory entry shows the oversight we require by policy: High for anything touching money, clients or credentials; Medium for content that goes public; Low for private, read-only summaries. This is a safety rule, not a test result.

Planned level

"Beginner / Intermediate / Advanced" is our pre-test judgment of the build. After testing we replace it with the measured setup time.

Corrections

Found an error? We fix it, note the change and date on the page, and say what changed. See corrections policy.

Questions

What does "Tested by AI Savings Lab" mean?: We built or ran the setup ourselves using test accounts and synthetic data, followed the written test plan, and recorded dates, time spent, costs and errors. It does not mean a certified or third-party lab test.
Where do prices come from?: Vendor pricing pages (or the vendor's own announcement when the page cannot be read), with the link and the date we checked. Case studies use what the owner actually paid.