Hyland used its CommunityLIVE 2026 keynote to push two products into general availability, the Enterprise Context Engine and the Enterprise Agent Mesh, while also introducing Agent Lifecycle Management and a Headless mode for the Content Innovation Cloud that points directly at Snowflake and Databricks pipelines, according to coverage from The New Stack.
CEO Jitesh Ghai, who took the job in May 2024 after running product at Informatica, used the stage to pick a fight with the prevailing agent-platform orthodoxy. 'There are many folks who are saying you need to completely revisit all your business processes to agent-enable your enterprise. This is what I call blowing things up, and I don't think any of it is necessary. In fact, I think it's improper,' Ghai said. That framing matters because Hyland sits on roughly 15,000 customers and more than $1 billion in revenue concentrated in regulated industries, exactly the buyers least willing to rebuild around an unproven agent runtime.
The pitch against rebuilding the enterprise around agents
Ghai's argument leans on two numbers Hyland keeps repeating. Knowledge workers, he says, spend 20 to 40 percent of their time on what he calls 'human ETL', the manual shuffling of context between systems. And 70 to 90 percent of enterprise data is unstructured, the documents, claims files, case notes, and contracts that Hyland has spent two decades indexing. 'If you want context, you have to meet an organization where it is, not reinvent yourself as a new organization,' Ghai said. The Enterprise Context Engine is the productized version of that thesis. It sits on top of the Content Innovation Cloud's federation layer and applies industry ontologies for healthcare, insurance, banking, government, and education, so agents query a domain model rather than raw blobs.
Why the Headless mode points straight at Snowflake and Databricks
The most strategically interesting release is the Headless mode for the Content Innovation Cloud. Instead of forcing customers through Hyland's own UI and workflow, Headless exposes the AI-native fabric, the federated content plus the context engine, as consumable APIs. The named targets are the Snowflake and Databricks ecosystems, where most large enterprises have already made their data platform decisions and where unstructured content has historically been a second-class citizen. Snowflake and Databricks both ship their own document AI features, but neither has Hyland's two decades of vertical metadata or its installed footprint inside the back offices that own the source documents. Hyland is effectively trying to become the unstructured-context supplier to the two warehouses competing to own enterprise AI workloads, rather than a parallel destination customers have to migrate toward.
Agent Passport and the bet on regulated-industry governance
Underneath the agent layer sits a governance stack that reads more like an identity product than a content product. Agent Lifecycle Management ships with an Agent Library for reusable templates and an Agent Passport that carries identity, capabilities, guardrails, and compliance attributes with each agent as it moves between systems. A Control Tower for the Agent Mesh is on the roadmap, giving operations teams a single plane to observe, throttle, and revoke agent behavior across the federated estate. For Hyland's installed base in claims processing, loan origination, patient records, and citizen services, the Passport is the more defensible piece. Auditors and regulators are already starting to ask what an agent is allowed to touch and who signed off on its scope, and a per-agent credential with attached policy is a cleaner answer than a spreadsheet of API keys and a Confluence page.
The neutrality claim and the OpenText, Box squeeze
Ghai was careful to frame Hyland as a Switzerland in a market that is about to splinter. 'There's going to be fragmentation. We recognize that there are other vendors that could equally benefit. And that is why we believe we have to be independent and neutral, open and modular,' he said. The unspoken targets are OpenText Content Cloud and Box, both of which are pitching their own AI fabrics on top of long-standing content repositories. Hyland's wager is that a federation-first, ontology-driven, agent-governed posture beats a single-repository AI bolt-on, especially for customers who already run three or four content systems and have no appetite to consolidate them before turning on agents.
Where we would place the bet for context-layer pilots in 2026
For the CTOs we work with, the practical question is whether to wait for Snowflake Cortex and Databricks Agent Bricks to grow native unstructured-context handling, or to bring in a specialist like Hyland now. Our read is that if your unstructured estate is concentrated in regulated workflows (claims, KYC files, clinical notes, case management) and you already pay Hyland or OpenText, the Headless mode plus Context Engine is worth a 90-day paid pilot in Q1 2026, scoped to one Snowflake-resident workload and one agent use case with a measurable cycle-time target tied to the 20 to 40 percent human-ETL number Ghai keeps citing. The Agent Passport is the line item to push on in the contract, because portability of governance metadata is what protects you if you swap runtimes in 18 months. If your content is mostly in SharePoint and Google Drive and your regulated surface area is small, buying a dedicated context layer in 2026 is premature, and the better move is to let Databricks and Snowflake native tooling mature through their summer releases. The build-versus-buy line we are drawing for clients: buy the governance plane, rent the runtime, and refuse to pay for another UI.
The concrete test comes at Snowflake Summit in June 2026. If Hyland appears on the partner keynote with a named Headless reference customer running in production on a regulated workload, the neutrality pitch is real and the Context Engine earns a default slot in regulated-industry RFPs for the back half of the year. If it does not, expect Snowflake and Databricks to ship competing unstructured-context primitives by Q4 2026, and Hyland's window narrows to its existing 15,000-account base before the warehouses absorb the category outright.
