Five points to check before you commit

Automating the wrong process is not a pricing mistake

These are the places an AI business case is most often overstated. Each of the five is a question you can put to whoever presents the case, in the meeting, without the document in front of you. None of them requires us.

Counterweight advises on AI investment decisions and builds nothing. No vendor pays us anything. Every engagement is run by one named adviser.
  1. 01

    Is it sized to the problem, or to the budget that was free?

    A 1× investment on a 100× problem gives a 1× result. Ask also whether the same money would return more somewhere else, and whether you are buying similar pilots over and over again.

  2. 02

    Three year TCO, or the first year price?

    Most of it is no longer a licence. It is consumption, priced per token or per call, so the bill grows with use and the pilot was the cheapest this will ever be.

  3. 03

    Where does the saved time go?

    Hours do not leave the company because a case says they will. If they reappear as review, exceptions and escalations, nothing was saved, and the only way to find out is to name the queue that absorbs them before you commit.

  4. 04

    Does the benefit apply to the whole volume, or only to the part that is actually automated?

    A case that counts fewer errors across the whole volume, when only a third of it will ever be handled automatically, is a miscalculation. The benefit is a third of what the case claims. It is easy to pinpoint once the model separates what is automated from what is not.

  5. 05

    Does the case assume a change to roles that you are able to make?

    A 10× or 100× change needs somebody’s work to change, and that is where the real timetable sits.

If the five checks leave you unsure, that is what the Decision Package is for. If they leave you certain, you have saved yourself the fee.

Tell us which AI investment is on your table, and we will say in half an hour whether a Decision Package is worth it. If it is not, we will say so in the first call rather than the third.