AI-Threshold FAQ — how the model thinks

The short version of the model, and why it is willing to tell you not to bother.

How do I decide whether to automate a task?

Compare what the task costs you by hand (frequency × time × loaded hourly rate) against the all-in cost of automating it: the one-time build or setup, the ongoing tool subscription and upkeep, and the part a human still has to do. If the saving recovers the upfront cost within your decision window, it is worth doing. AI-Threshold also discounts the freed time by how much of it you can actually convert into money.

Why would a calculator ever say don't automate?

Because automation often is not worth it: the task may be too infrequent to recoup the build cost, the ongoing cost may exceed the saving, or an off-the-shelf AI feature may replace the task before a custom build pays for itself. Telling you not to bother is the honest, trustworthy answer when the numbers say so.

What is residual human work (the 80/20)?

Automation is rarely 100%. Typically you automate the easy 80% and a human still handles the remaining 20% — exceptions, judgement, review. That residual is often the hardest, most expensive slice, so AI-Threshold treats it as a first-class input rather than assuming the task disappears entirely.

Why discount the time I save?

Saved minutes only have value if you redeploy them to something worthwhile. Five minutes scattered through the day rarely convert to money; a freed four-hour block can. AI-Threshold applies an opportunity-cost factor based on how the freed time is used and whether it arrives in usable blocks.

What does the confidence percentage mean?

It is the share of thousands of Monte-Carlo simulations — run across plausible ranges for your uncertain inputs, especially setup time — in which the automation pays back within your horizon. It replaces a single false-precision number with an honest probability.

Should I build, buy, or delegate?

AI-Threshold scores all four options on the same basis — keep doing it by hand, build a custom automation, buy an off-the-shelf tool, or delegate to someone at a lower rate — and recommends the one with the best net value over your horizon.

How accurate are the numbers?

They are deliberately rough — directional estimates, not accounting. The point is to make the trade-off visible and show how sensitive it is to the inputs you are least sure about. Tune the inputs to match your reality, and watch the verdict and the confidence move.

A worked example: chasing invoices

Chasing overdue invoices 8 times a week, ~10 minutes each, at a $45/hr loaded rate costs about $260/month by hand. Automating 80% with a $30/month tool leaves ~$82/month of cost, a paper saving of ~$178/month. But the freed time is scattered, so only ~25% converts — a realistic saving near $44/month. Against a $720 build, that is roughly a 16-month payback — and invoice-chasing is generic enough that an off-the-shelf tool may do it within ~12 months. Verdict: probably buy, do not build.