AI in enterprise software keeps improving all the time – a new version, a few extra features, minor refinements. Every so often, though, a step comes along that is different. At OpenText Content Management, that is exactly what is happening: a system that until recently only answered questions now carries out tasks on its own – creating a workspace, updating metadata, generating a document.
Where AI in Document Management Stands Today: From Assistant to Autonomous Action
Until recently, it was simple: AI in document management helped write a summary, translate text, or assist with drafting an email. It was useful, but passive – it waited for someone to assign a task, then returned text. Done, end of interaction.
This is changing. OpenText is moving its AI layer, Content Aviator, toward what is known as agentic AI – a system that carries out tasks directly instead of merely responding to them. The difference is best illustrated with an example: previously, you would ask AI to “write a summary of this contract” and receive text back. Today, it is enough to state what you want to do – for example, create a new workspace or update metadata – and the system identifies the intent, selects the right tool, and completes the task.
For this to work, AI needs to know more than just what is in the documents. It now also recognizes users, their roles, and the relationships between individual workspaces. This is why OpenText is building what is known as a knowledge graph – essentially a map of the connections between enterprise documents and data, enabling AI to identify relationships across the entire organization, not just within the single folder or project currently in view.
As a result, document management is no longer a place AI occasionally visits to offer assistance. Instead, it becomes the entry point through which AI gains the context it needs to act independently.
OpenText is far from alone in this. Microsoft, Google, and ServiceNow are all pushing a similar shift from search tools toward autonomous agents across their platforms – this reflects a direction for the entire enterprise content management market, not an isolated experiment by a single vendor.
What This Means in Practice – Scenarios Already Working Today
This is best understood through what OpenText itself demonstrates as a working scenario, not a promise for the future.
Take workspace creation (so-called Business Workspaces), for example. The user simply states that they want to create a new workspace, providing a name and description – and the system identifies what needs to be done. If information is missing from the request, the AI does not guess; instead, it asks for clarification before proceeding. Only once it has everything it needs does it create the workspace and switch the user directly into it, allowing them to continue working where they actually need to.
Metadata updates work in a similarly simple way. Instead of clicking through forms, the user simply states that, for example, a customer’s location needs to change – and it is done, with the change reflected immediately.
Another example comes from HR: an employee needs an employment confirmation letter. The user simply requests it within that employee’s workspace. The system locates the corresponding template on its own, asks the user to confirm it is the correct one, then generates and saves the document – within seconds, rather than after days of waiting on the HR department.
Then there is the knowledge graph in action: it enables the system to answer a question whose answer does not exist within the currently open document at all. A typical example is locating a contract linked to a specific piece of material – even if it is stored in a completely different, connected workspace, the AI locates it and presents it directly.
What do all these scenarios have in common? The user states what they want, the system determines how to carry it out, and for more sensitive steps, it confirms it has correctly understood the request before proceeding.
Even so, it is worth taking a realistic view. Many vendors on the market are simply rebranding older chatbots and automation tools under the “agentic AI” label, without genuine autonomous decision-making behind them. And even where an agent genuinely works, the most common cause of failure in production is poor underlying data quality and unclear process ownership – not limitations of the model itself. Before moving forward with deployment, it is worth verifying whether your document repository has sufficiently clean metadata and clearly assigned process owners. Without that, even the best agent will not help.
What’s Coming in the Months Ahead
This is not a one-time update that OpenText will release and revisit a year later. It is a gradually rolling roadmap, and the nearest milestone is integration with Microsoft Copilot. Content Aviator will be deployable as a dedicated AI agent directly within Copilot – users with a Microsoft 365 license will be able to query enterprise documents without leaving Word or Excel, the environment where they normally work.
This is enabled by standardized protocols that let different AI tools “understand” one another and hand off tasks – this is the first step, with further partner integrations expected to follow.
The agentic capabilities themselves also continue to expand: automatic document comparison, searching for and creating workspaces without human involvement, anonymizing personal data in documents, or triggering workflows on a specific request. The direction is clear – an increasing share of routine tasks is shifting to the system, while people remain responsible for decision-making.
What to Take Away for Decision-Making Right Now
The most important point, however, is not found in the feature list. It lies in where this AI layer can actually run – and that already belongs on the C-level agenda today, not just in the IT department.
OpenText currently offers four deployment models, each representing a different level of control over data:
- Private Cloud – Content Aviator runs in the OpenText cloud and connects to third-party language models.
- Hybrid deployment – Content Aviator runs in the OpenText cloud, while document management itself remains on your own servers.
- Fully on-premises – both the AI layer and the language model run behind your own firewall, with no data leaving your data center. Intended primarily for organizations with high data protection requirements.
- Customer Cloud (BYOK) – you run in your own cloud with your own contract with the language model provider, without the need to host your own infrastructure.
In simplified terms: the more control over data you want, the more responsibility (and cost for your own infrastructure) you take on. Cloud variants deploy faster and receive updates without involving your IT department. The on-premises variant requires you to provide and maintain your own infrastructure, but data never leaves your environment, not even for a moment. BYOK is a middle ground – you control the contract with the model provider without having to build and manage your own servers.
This choice is not a detail to be settled later with the implementation team. It is a decision about where the company will send its most sensitive data and who will have access to it – and it should be made before the company commits to anything.
“The choice of deployment model for the AI layer is decided today at the C-level, not by the IT department – because it is no longer a question of technical configuration, but of where the company sends its most sensitive data and who has access to it.” Jiří Jakeš, IXTENT
The key questions to ask before deciding are: where will your data actually end up, who will have access to it even within the AI layer, and is your document repository clean enough for an agent to extract anything usable from it. Without answers to these questions, it hardly matters which of the four models you choose. If you would like to work through this for your specific environment, get in touch.
Author
Lukáš Hronek, Head of OpenText Team
Lukáš has worked in document management since 2018. He specializes in the OpenText Content Management and OpenText Intelligent Capture platforms, which he has deployed for large organizations in the energy, banking, and other sectors – including Západoslovenská distribučná, E.ON, Komerční banka, and GECO. He has experience in analysis, consulting, and technical implementation of DMS solutions. At IXTENT, he leads the OpenText team.