Open Text and Cohere team up to bring agentic AI to regulated sectors: whoever already owns the data wins

By Zendoric · September 17, 2026. Open Text (Nasdaq: OTEX), the Canadian enterprise data management software provider, announced on September 16 a partnership with Cohere to bring agentic AI — systems capable of carrying out multi-step tasks on their own, not just answering questions — to heavily regulated sectors: government, healthcare and financial services, as reported by Simply Wall St.

The deal combines North, Cohere's agent platform, with Open Text's tools for managing unstructured, operational and transactional information inside organizations. The source does not spell out how it will fit into the product catalogue: it points to Aviator, the AI agent brand Open Text already offers in its enterprise cloud, as the place to watch for future launches that explicitly cite Cohere, and lists among the open questions the deployment options in sovereign cloud (infrastructure hosted under a country's own data residency laws) or on-premises, that is, on the customer's own servers without going through the public cloud.

Beyond that, the announcement is thin on verifiable facts: there are no contract figures, no launch timeline and no technical detail on North. Simply Wall St itself notes that the real test will come if Open Text reports concrete contract wins in government or regulated sectors, and if 'AI bookings' show up within its enterprise cloud metrics. Until then, it is a statement of intent.

As sector context, Cohere has built much of its business in the enterprise and government arena — with prior deals with cloud providers and public administrations that demand control over where and how their models run — a different positioning from OpenAI, Anthropic or Google, which are more focused on the end user. For a lab that does not compete on mass adoption, partnering with whoever already holds the commercial relationship and the compliance infrastructure inside those organizations is, quite simply, its route to distribution.

Our reading is that the real product being sold here is not model capability, but trust. Open Text has been inside the data flows of governments, hospitals and banks for years; that access and that regulatory-compliance credibility are the scarce asset, not the agent itself. It is the same pattern we have been observing in the consolidation of the enterprise AI stack: frontier labs supply the model, but whoever controls the integration, the data governance and the relationship with the regulated customer is the one who decides whether the deployment actually happens. This kind of announcement does deserve the usual caution: enterprise AI partnerships are cheap to announce and expensive to execute, and many never move beyond the press release into measurable revenue.

There is also a direct connection to something we have been documenting for months: healthcare, banking and public administration are precisely the sectors where administrative and back-office work — processing, document verification, first-line support — is most exposed to automation, while clinical judgment, expert risk assessment or direct care hold up better. This kind of infrastructure is, literally, what makes that shift possible: agents that digest operational and transactional data inside systems that today depend on people for repetitive tasks.

In the short term, that means real friction in the administrative workforces of public agencies, insurers and health systems, with the added difficulty that these sectors handle sensitive data and tolerate no compliance errors. In the long term, if the promise of 'trustworthy agentic AI' holds, it is exactly the kind of back-office automation that frees up resources for what does require a human presence: clinical care, complex financial advice, public service with judgment. That is the ground on which this announcement, modest in facts today, will either make sense or turn out to be one more on the long list of AI alliances that never led to a signed contract.