Work/In production
AIICO INTELLECT
An internal AI platform for enterprise knowledge and decision support.
- Loop it closes
- The distance between a question a staff member has and the document that already answers it.
- Role
- Built and deployed in-house
- Context
- AIICO Insurance
The situation
A large insurance business, decades of policies, procedures and internal documents, and no AI team. The knowledge existed. It was simply unreachable at the moment anyone needed it — so people asked a colleague, or guessed, or did the thing the way it was done last time.
What it does
INTELLECT lets staff query policies, procedures and internal documents in natural language and get grounded answers with source references — not a confident paragraph with no provenance. The references are not a nicety; in a regulated business they are the entire reason anyone is allowed to trust the output.
It is used for onboarding and day-to-day operations: the two places where “where do I find this?” costs the most time.
Building AI where there is no AI team
This is the part that transfers to any enterprise. The engineering was the small half.
Trust is earned before it is granted. You do not get to replace a working process. You get to sit beside it, be right repeatedly, and get adopted.
Grounding is non-negotiable. Every answer points at its source. An ungrounded system in this environment is not a weaker product — it is an unusable one.
Imperfect data is the actual terrain. Waiting for clean documents means shipping nothing. The system has to be built for the corpus that exists.
Where it is going
It is evolving into an analytics engine — the same natural-language surface, pointed at business questions instead of documents:
“How many policies were sold last quarter?”
Same principle, harder grounding problem: the answer has to be traceable to the numbers it came from.