// Case study / AI platform analytics
Product analytics for an enterprise GenAI platform.
Adding the front-end measurement layer to Fuel iX, TELUS Digital’s enterprise GenAI platform. The product had backend metering for billing, but not behavioural telemetry for product and engineering questions.
What happened
An enterprise GenAI platform gained a behavioural analytics pipeline.
Fuel iX had scaled from internal tool to commercial GenAI platform, but still had only backend metering for billing. The front-end behavioural signal layer — GA4 event collection and BigQuery export covering platform-wide and organisation-specific events — was designed and delivered to fill that gap.
The pipeline went live and into active engineering use within a week — the team had been waiting for this data and moved immediately. An internal BigQuery SME role continued after delivery, including a GA4-querying primer for engineers beginning to work with the data.
Context
The product had usage metering, not product behaviour.
Backend billing metering could support commercial operations, but it could not answer product questions about which agents or copilots were being used, how often, by whom, or how behaviour differed across organisations.
Task
Add a front-end behavioural layer without making engineering dependent on analytics.
The work needed to produce usable event data, export it to BigQuery, and equip the engineering team to query it directly rather than waiting on an analyst for every answer.
Event layerGA4 collection
Platform-wide and organisation-specific behavioural events captured from the GenAI product.
BigQuery exportQueryable stream
GA4 export made the event stream available to engineering and product analysis.
KPI queriesQuery primer
Schema, event-parameter unnesting, and example queries including copilot count and conversations per user.
AdvisoryPlatform input
Measurement-layer advisory input to a related platform tool.
Designed GA4 event collection for platform-wide and organisation-specific behavioural events.
Delivered the export pipeline that made front-end events available for deeper analysis.
Prepared examples including total copilot count, conversations per user, and Custom Copilot access-type distribution.
Provided advisory input to a related platform tool on its own measurement layer.
Action
The data layer was designed for engineers to use.
GA4 event collection and BigQuery export pipeline were the delivery; engineering adoption was the harder part. A querying primer — event-parameter unnesting, schema model, KPI queries — made the data available to the team building the platform, not just accessible in principle.
Outcome
Engineering could query how the platform was actually being used.
- GA4 event layer delivered for platform-wide and organisation-specific behavioural events.
- BigQuery export pipeline confirmed live and in use by engineering within a week.
- GA4 querying primer created for the engineering team.
- Internal BigQuery SME support continued after delivery.
- Measurement advisory extended to a related platform tool.
Design lesson
AI product analytics still starts with ordinary behavioural questions.
Even on an enterprise GenAI platform, the core questions are familiar: who used what, how often, in which context, and what changed after release. The platform may be novel; the need for trustworthy behavioural telemetry is not.