Novo Nordisk says it has nearly 8,000 AI agents across its business, supported by a Snowflake data platform holding more than 900TB of data. The interesting bit is not the agent count on its own: the company describes using the platform for sales support, translation and data-pipeline work, with controls intended to let staff explore data safely.
Watch Desk analysis
What happened
Novo’s data and AI vice-president, Loïc Giraud, told Diginomica that the company consolidated its Snowflake teams and integrated more datasets after he joined in 2024. Around 25,000 people across Novo affiliates use the platform directly or indirectly, he said. More than 60% of employees use AI weekly, and the company has nearly 8,000 agents running across the business, including in research and development.
In the US, agents help field sales staff prepare for visits to healthcare professionals. Novo also uses AI for translation and is exploring agents that build data pipelines. Staff can query Snowflake data in natural language, while the company says its systems can connect to different AI models and services.
Why it matters
This is a useful view of enterprise AI beyond the familiar chatbot demo: a large data platform, staff access, agents for specific jobs and the governance needed to connect them. Novo’s account gives readers concrete examples of where the work is happening, from preparing sales visits to translating company information across its affiliates.
The figures and benefits are company claims reported by Diginomica, not independently measured results. Still, the operational detail makes this more than an agent-counting exercise. The hard question is whether broad access and thousands of agents produce dependable work, rather than simply more activity to supervise.
Our read
The strongest lesson here is that the agents sit on top of years of data integration and organisational work. That is less dazzling than an army of bots, but probably more useful to anyone trying to build this in a real company. Keep the agent count in perspective; look for evidence of what the agents actually improve.
What to watch
- Whether Novo publishes measured results for sales, translation or pipeline-building agents.
- How staff access and governance work as the number of agents grows.
- Whether the company can switch models in practice, as Giraud says is important.
Discussion spark: Would you prioritise giving staff broad access to data and AI agents, as Novo describes, or require each use case to prove its value before scaling?
Sources and evidence
- Want to hear more about Agentic AI? (30 October 2026, 09:05 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.