Global – October 27, 2026 -- Snowflake's sales finance team has cut its monthly forecast and variance analysis from more than a week to a few hours. It did so by giving its CoCo AI standing context about company data, terminology and workflows.
Profile cuts steps per answer from 3.0 to 1.3
Across 44 finance users, those with a 'GTM Finance' profile reached an answer in about 1.3 steps, against about 3 steps for those without one (Snowflake internal CoCo usage, February to May 2026). Without the profile, the AI spent roughly three steps per session finding the right data source and resolving terminology.
Each extra step adds about 0.05 credits, roughly 10 cents at US$2 per credit. A five-step session therefore costs about three and a half times a one-step session. At every step count in the data, sessions with the profile cost less than those without.
Faster forecast enables midmonth updates
The sales finance team runs more than 30 recurring analyses each month, covering topline forecast, pipeline, commissions and business performance reporting. The forecast previously required several analysts to review pipeline deals, read CRM account notes and compare historical trends. Because it took more than a week, it could only run monthly.
The topline forecast skill now completes the same analysis in a few hours. Snowflake publishes a midmonth update, and any analyst can rerun the skill to test a scenario.
Gayatri Kannan said the profile removed the inconsistent and inaccurate results seen at launch, when the team had to keep reminding the AI of terminology and data sources.
Analysts build 36 skills in two months
The skill library grew from three starter skills to 36 in two months, and several Excel workbooks became shared skills. Three months after launch, 87% of first adopters still use CoCo daily.
The Bookings BvA skill produces month-end variance across deal categories in minutes, including which opportunities slipped and why. The sales travel expense report skill covers budget variance, out-of-policy spend and top spenders per leader, and drafts the email to each. Work that took hours across two teams now takes about 15 minutes.
Central certification keeps analyst-built logic in bounds
The analytics team certifies data sources, and a technical source guide defines which to use and why. The team also owns the terminology layer, while analysts build skills on top. Skills are shared and visible to the teams that depend on them.
Snowflake recommends a three-step sequence for other teams: certify sources first, encode terminology and output formats second, then open skill-building to the functional team. Its corporate FP&A team has applied the same approach to long-range planning.