There is a group of businesses that has been watching the AI agent wave from the sidelines, and not because they lack the budget or the ambition. Government agencies, hospitals, financial institutions, and other regulated organisations have a harder problem to solve: how do you deploy agentic AI when your data cannot leave a specific jurisdiction, when regulators audit your systems, and when a wrong answer carries legal consequences?

On September 16, 2026, at the ALL IN AI conference, OpenText and Cohere announced a strategic partnership aimed directly at that question.

What the Partnership Actually Does

OpenText has spent decades building enterprise content management. It sits on top of enormous volumes of unstructured, operational, and transactional data that most businesses have never been able to use effectively. Think contracts, invoices, case files, compliance records, and decades of scanned documents.

Cohere brings North, its enterprise AI platform designed to run in a fully private deployment. North can run on-premises, in a private cloud, on public cloud infrastructure, or inside a sovereign cloud, depending on what a particular government or regulated business requires. Cohere’s enterprise models are purpose-built for business use, not general consumer consumption.

The partnership combines both layers into a single integrated stack. OpenText provides the data context and management layer. Cohere provides the agent orchestration and reasoning layer. The distribution runs through SOLEX, a reseller relationship that allows clients to access the combined solution without having to negotiate two separate enterprise contracts.

The joint solution is expected to reach clients in early 2027.

Why Regulated Sectors Have Been Left Out

The mainstream AI agent story has largely been told through consumer and software-first businesses. Startups move fast. Tech companies have data infrastructure that is already cloud-native and relatively easy to instrument.

Regulated industries operate differently. A healthcare system cannot send patient records to a third-party API without triggering a cascade of compliance requirements. A government agency handling sensitive infrastructure data cannot rely on a model that phones home to a US data centre when the agency is bound by local data sovereignty laws. A financial services firm subject to GDPR, DORA, or equivalent national financial regulations needs full audit trails on every automated decision.

The gap between “AI is available” and “AI is available to us specifically” has been real and frustrating for these organisations.

OpenText and Cohere are not the first to try to address this. But the combination of a company with deep government and enterprise data relationships (OpenText) and a model provider that has made private deployment a core product commitment (Cohere) addresses both sides of the problem simultaneously.

The Data Foundation Problem is Still the Real Problem

One of the reasons AI agents underperform in enterprise settings is not the quality of the models. It is the quality and accessibility of the data the models are working with.

Most large organisations have data in dozens of disconnected systems. A customer service agent cannot give useful answers if it can only see one of the three databases that hold relevant information. A compliance automation tool cannot do its job if the relevant documents are locked in a legacy content management system with no API.

OpenText’s contribution to this partnership is arguably more important than it might seem from the announcement. Giving an AI agent clean, contextual access to decades of enterprise content is a hard engineering problem. It is the kind of work that normally takes years of integration effort before an AI layer can even be tested.

Packaging that capability with Cohere’s agent runtime means regulated organisations can deploy agents on top of their existing data estate without first rebuilding their data infrastructure from scratch.

What This Means for Business

If you operate in a regulated sector, this is one of the more practical AI deployment announcements of the year. The combination of private deployment options and established enterprise data integration addresses the two most common objections to agentic AI in your industry.

If you operate in an unregulated sector, the broader pattern is worth watching. The shift toward sovereign and private AI deployment is not just a compliance story. It reflects a growing enterprise expectation that AI runs inside your environment, with your data, under your control. That expectation will become standard across industries over the next few years.

On the data quality question: this partnership is a reminder that before AI agents can help your business, your data needs to be in reasonable shape. Unstructured content, disconnected systems, and inconsistent records are not just a technology problem. They are a strategic liability. Organisations that invest in data foundations now will see dramatically better results from every AI tool they deploy next year.

At Enterprise DNA, this is exactly the work we have been doing for years. Understanding your data, structuring it well, and building the analytical capability to act on it is what separates businesses that get real value from AI from those still waiting for the technology to be “ready.” The technology is already ready. The data readiness question is what actually decides who benefits.

If your organisation is looking to close that gap, the EDNA Learn platform has structured programs in data management, analytics, and AI foundations that give your team the skills to build on. And if you are ready to deploy AI agents in your operations, the Omni Advisory service can help you design an approach that matches your specific data environment and compliance requirements.