Amazon: measure AI by results, not by usage volume
In brief
According to Next, Amazon ended an internal ranking of AI use. A September memo by engineering vice-president Terk Terkowitz, made public on 27 October, tells software engineers to prefer concrete results. Token consumption and similar metrics remain, but he says they are imperfect and, if treated as goals, can raise costs and complexity.
Source : Next - « IA : Amazon veut des résultats, pas de tokenmaxxing »
What it changes for an SME
According to Next, pushing staff to use AI and tracking volume rather than outcomes is not unique to Amazon and can inflate API bills. If you mainly follow consumption, you may pay more without saving time. Amazon still collects usage figures but, according to the memo, low use is not a failure when the tools do not help, and high use only matters if deliveries improve. For an SME, the useful test is time, cost and customer impact.
What you can do
- Check whether your AI indicators track a business result or only usage volume.
- Ask teams to record cases where current AI tools do not help, instead of setting a usage quota.
- Compare recent API costs with the deliveries or time actually saved.
Let's talk about the time you could save.
In 20 minutes, we find where AI could save you time or money. Free, no commitment.