How Fable Food Turned Six Time Zones of Scattered Data into One AI-Ready Source of Truth

Fable's first move with Coupler.io was to pull data out of separate inventory, finance, and CRM systems and consolidate it into a central BigQuery warehouse, powering company-wide Google Data Studio reporting on basics like stock location, shipments, and value.

Once that foundation existed, Fable extended it to power an AI agent. Working domain by domain, inventory first, then finance, then CRM, the Coupler.io team restructured BigQuery tables and built data dictionaries explaining what each field meant, where it came from, and whether it was a true source of truth. That structure is what lets Fable's AI agent, nicknamed Mycelium, search and retrieve information quickly and reliably.

Mycelium runs on Anthropic's AI models and is deliberately locked to 100% read-only access. No control of devices, browser use, write, edit, or delete capability on source data. Employees ask Mycelium questions directly in Slack; it pulls context from the Slack thread plus the BigQuery warehouse, then replies in the same channel, visible to the team, so answers can be corrected if needed.

Fable's team is spread across roughly six countries, and before Mycelium, getting an answer to a simple question meant tracking down whoever owned that particular system, then waiting, sometimes a full day, if the first question wasn't specific enough.

Now, employees ask Mycelium directly in Slack and get an answer immediately, without waiting on a single person's time zone. The exchange stays visible in the channel, so if Mycelium gets something wrong, anyone can step in and correct it.

Google Data Studio dashboards worked well for operations and finance teams who used them daily, but for a salesperson with an occasional question, learning what each field meant and how to filter it wasn't worth the effort. Most defaulted to just asking a colleague instead.

Coupler.io is what makes the shortcut possible: it keeps sales and inventory data syncing into BigQuery, structured and current, so any question asked against it gets a reliable answer. That's the data Mycelium reads from when a sales rep asks, in plain language, for sales history by product and month or current inventory levels, and gets back a direct answer or a generated chart in Slack. No Google Data Studio or BigQuery training required.

Beyond simple lookups, Chris runs deeper analysis on the same warehouse: investigating why supply isn't meeting demand, or where safety stock levels are falling short.

Coupler.io keeps that BigQuery data structured and current, which is what makes it usable for this kind of analysis in the first place. Chris uses Claude to generate queries, to pull central data into a financial operating model from Riverlogic. Once a useful approach is found, Fable turns it into a repeatable playbook, so the same analysis can be re-run on demand instead of rebuilt from scratch each time.