ARCHA: what a mountain-tour operator taught us about governed data

The tourism industry of the Kyrgyz Republic runs on knowledge that rarely makes it into a database. Which camp is reachable from which pass in early July. Which provider actually answers the phone in Osh. What a horse day costs this season on the south shore of Issyk-Kul. When we started working with tour operators, that knowledge lived in three places: Excel files, WhatsApp threads and the heads of two or three coordinators.

Costing a single custom tour meant reconciling all three. It took days, and the result depended on who happened to be in the office. That is the exact shape of the problem we care about: the data existed — it just could not be heard.

From spreadsheets to a graph

We ingested the operators' spreadsheets, price lists and route notes and organized them into a graph: 73+ nodes — cities, camps, passes — connected by edges that carry real costs and travel times. The graph is undirected: a trail works in both directions, and one edge holds one truth about it.

Structure changes what questions can be asked. Once camps and passes are nodes with priced edges, "what does a 6-day route from Karakol cost" stops being a research project and becomes a query.

A calculator on top, an agent in progress

On top of the graph we built a tour calculator: pick the points, get the route and the full cost estimate. What used to be assembled by hand over days is now computed automatically — consistently, from one governed source. An AI agent that plans tours conversationally over the same graph is in development; it will work with the same data, under the same rules.

Why this matters beyond tourism

ARCHA is small compared to an enterprise document estate, and that is exactly why it is a clean illustration: this is the ENTUM approach — ingest messy reality, organize it into a governed structure, let the data speak. The industry changes; the method does not.