Air Canada: How the country's largest airline connected its AI to fly higher with Dataiku

Air Canada uses Dataiku, the Platform for AI Success, to orchestrate analytics, AI, and agent workflows so different teams get the answers when they need them.

Last year, Air Canada, the country's largest airline, carried more than 45 million passengers across six continents on more than 190 routes. Running an airline that size means making hundreds of decisions a day that depend on each other: how to optimize for cargo space, what offer reaches which loyalty member, how much demand a route will see next winter.

Air Canada was never short on ideas to improve business performance, but 18 teams each had their own data, their own tools, and their own starting point spread across systems never built to talk to each other. So the analytics organization, led by Milena Kumurdjieva, Managing Director, Data and AI, decided to connect the infrastructure instead of replacing it. Air Canada leveraged Dataiku as its orchestration layer, on top of Snowflake and alongside the booking systems, marketing platforms, and data pipelines its teams already used.

Eighteen teams, one place to build

Air Canada now runs analytics, machine learning, and AI agents all in one place, and the teams closest to each problem build for themselves in this shared workspace. In marketing, that looks like NAVI, short for Navigator: a system of specialized agents that Luc Gagnon, Director of Travel Marketing Analytics, and his team built to guide marketers through campaign development. NAVI orchestrates capabilities such as audience selection, route and market analysis, and standardized campaign outputs, helping marketers move faster from campaign intent to actionable, data-informed decisions.

Adobe's Workfront, Experience Platform, Journey Optimizer and Campaign then carry the approved work into execution. In cargo, the team went after a different problem: booking data nobody could see without checking flight by flight. Dataiku brings those sources together in one repeatable workflow. Marketing, cargo, sustainability, tax, and forecasting all build in the same environment, so what one team creates can travel to the next.

Governance enabled the agents to leave the pilot

Dataiku sat on top of the infrastructure Air Canada already had rather than replacing it. Snowflake holds governed data at scale, and Dataiku orchestrates the analytics, models, and agent workflows that run atop it. It opened the work to people who know the business but don't write production code. Air Canada started with 15 licenses and now has more than 100 Dataiku users company-wide. Without governance built in, each workflow would need to be reviewed individually, slowing growth. Instead, Air Canada leverages governance as an accelerator, by building within established guardrails to ensure their agents and other workflows meet the corporate and regulatory compliance requirements. 

What faster insights are worth

Once a flight leaves the gate, any cargo capacity that went unsold is revenue Air Canada never gets back. Preventing that falls to the Cargo capacity and revenue management team, who for years worked without a real-time view of consolidated cargo bookings across the carrier–insights that previously required tediously checking flights one at a time. Now that Cargo booking data flows through Dataiku, the team sees shifting capacity and pricing while there is still time to fill the aircraft. Air Canada credits the change with contributing to a stronger revenue performance.

Travel Marketing tells a similar story about speed. With NAVI, analytics activities that could previously take weeks or even months can, for supported use cases, now be completed 83% faster, helping the right message reach the right customer at the right time. For a recent marketing campaign, NAVI helped reduce production time for comparable digital banner assets from approximately 3 days to 4 hours by reusing Workfront campaign data, enriching the required context, and automating downstream content, HTML generation and QA, improving both campaign relevance and performance.

Building an airline that predicts instead of explains

Air Canada defines AI success by business value first, trust second. The team is building toward a reusable enterprise advantage for data, where loyalty or forecasting can pick up a dataset, model, or agent built for cargo or another area of the business instead of rebuilding for each new use case. More agentic workflows follow from there: systems that plan and draft the work, with people who approve it. What that foundation already delivers is concrete. Eighteen departments building on one platform. More than 100 users from what started as 15. Campaigns in hours instead of months. An airline that went from explaining what happened to predicting what will. 

Forty-five million people a year trust Air Canada with a journey that matters to them. This foundation is what lets the airline meet each of those journeys as their gateway to the world.