Systems that answer
The material the system answers from. It is the product; the model is a commodity you rent by the token.
Also called Retrieval corpus · База знаний
Whether a deployment works is decided almost entirely here, and almost never in the choice of model. Your catalogue, prices, prerequisites, exclusions, delivery terms and the eleven edge cases everybody knows verbally — either they exist in a form something can retrieve, or the system has nothing to be right about. In most companies they exist in a senior person's head, a spreadsheet nobody else opens, and four years of chat history.
Which makes the first month of one of these projects an extraction problem rather than an engineering one, and it is worth pricing it as such. The uncomfortable part is that writing down the exclusions forces decisions the business has been avoiding: what the price actually is, who is not a customer, what happens when the thing arrives late.
It also has to stay current, and currency is a process rather than a delivery. A knowledge base with last quarter's prices is worse than no system at all, because it produces confident wrong commitments at scale rather than an absence somebody would have escalated.
A crawl of your own marketing pages produces a system fluent in your positioning and ignorant of your business. It will answer questions about values and improvise about lead times, which is the exact inversion of what a buyer needs. The material worth retrieving is the operational material, and most of it has never been written down.
Knowing the definition is not the same as being able to check the figure. These are the procedures that do the second thing.
- Accepting an automated agent
- “The agent is trained on your data and ready to go live.” · 45 minutes, 6 questions.
The definitions are the easy part. Whether the figure on your dashboard was computed this way is a different question, and usually the more expensive one.