AI Presence - Core Design

Goal

Our motivation was reflecting to clients how LLMs “perceive” their brand identity.

Heuristics

We prompt the LLM with a general question like:

“What are the leading companies in the running shoes industry? Which ones will you recommend buying? Which brands fall behind and I should avoid them?”

We then process the response through non-bias sentiment analysis and sum the impressions by tone.

Method

We economize with LLM batch APIs. We use a thin ‘Interaction layer’ for communication, and scale compute using Lambdas.

F**K Ups

Rate limits on batch APIs cause requests to throttle at scale.

Solution

Mirror the API queue and manage it within the system.