Chapter 2. Aligning Business Metrics with Technical Reality
Imagine the board mandate from Chapter 1 has moved from planning to early adoption. The organization still wants to reduce unmanaged external AI spending, keep sensitive support and code work inside its own environment when that makes sense, and make the economics of inference visible. The first product is still the unified internal assistant. The difference is that more teams are now using it.
For continuity, assume this is still the 300B-class model candidate from Chapter 1. The question has moved from memory fit to workload promises: can one expensive shared model carry different service behavior targets, or Service Level Objectives (SLOs), while different teams own the cost?
On a dashboard, it can look like one service. One model. One serving platform. One monthly token line. The diagram on the dashboard is calm again, and again, that is the warning sign. The calm view hides the real story: different users are asking the same assistant to satisfy different promises behind the scenes.
A support ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access