Skip to content
Theme
OnDevice Tools

GadgetsLaptops & PCs

Enterprise AI ROI may hinge on document management

A TechRadar analysis says businesses may weaken returns from enterprise AI if they ignore the documents and context that AI systems rely on.

Illustration of enterprise AI software organising business documents and metadata

Key takeaways

  • TechRadar says enterprise AI returns may depend on document management and context, not only models.
  • The article warns that AI systems can repeat errors if they use outdated or unapproved documents.
  • It recommends starting with high-value content such as contracts, procedures, customer records and financial data.
  • It says metadata and governance can help control access, versions and workflows.
In this article
  1. The overlooked layer
  2. Why AI needs more than search
  3. The cost of poor context
  4. Start with high-value content
  5. What it means for AI investment

The overlooked layer

In an article published on 18 November 2026, TechRadar says enterprises are investing in enterprise AI as systems move beyond simple answers to decisions and actions. The article says document management may strongly influence whether those investments pay off. It describes document management as the discipline of organising files, versions, access rights and context.

The piece names Ryan Barry, a senior vice president at M-Files, and Ian Story, a principal architect working on OneDrive and SharePoint at Microsoft. It says the issue is not a lack of information. Many companies have too much content spread across shared drives, business apps, collaboration platforms and personal workspaces. Collaboration tools have increased the volume of shared material, yet they have also made trust harder to judge.

Illustration: AI, search, business, law and the labels AI Roi, Document Management, Enterprise AI

TechRadar says a common error is to assume that making content searchable is enough for AI readiness. Search can locate a file, but it may not show whether the file is current, approved or relevant to a task. The article says locating a relevant file is just the start. AI may also need to know why a document exists, whether it is approved, where it belongs in a larger process, and what other information gives it meaning.

The article warns that errors can grow when AI agents act inside workflows. If a person opens the wrong file, there may be time to notice. If an automated system uses old guidance, the mistake may spread before anyone checks. Humans often rely on unwritten knowledge. They know who owns a process, which version is current and when an exception applies. AI does not automatically have that understanding, so the article says it must be provided carefully.

The cost of poor context

The article cites a forecast that AI spending is projected to hit $2.5 trillion in 2026. It says organisations that focus only on models, infrastructure and applications may weaken returns if the information layer is inaccurate or poorly controlled. The concern is not that AI tools are weak, but that they may receive weak inputs. The article says the question is no longer only what AI can do. It is whether AI has suitable information to do the task well.

Examples in the article include several copies of the same contract, an old policy stored beside a current one, and exception guidance known only to a few employees. It says most enterprises do not lack information. They have large volumes created over years of collaboration. The problem is selecting and governing the material that matters. People often use experience to judge which version is correct. AI systems do not have that experience unless the organisation supplies it in a structured way.

Start with high-value content

TechRadar does not suggest making every document perfect. It recommends identifying strategic business content first. This is information that is critical to operations, where mistakes are expensive, and where AI needs extra context. Contracts, standard operating procedures, customer records and financial data are given as examples. The article says not all content deserves the same degree of attention. Making every file perfect would be a large task and may not be necessary.

The article offers a practical checklist. Leaders can ask whether a clear authoritative version exists, who owns it, who may access it, what related material gives context, and whether AI can reliably decide when to use it. These questions resemble traditional information governance, but the article says they become central to AI readiness. It also says this approach can link information work to specific AI use cases. Instead of a broad transformation programme, teams can focus on content tied to the tasks they want AI to perform.

What it means for AI investment

The article says added context can strengthen control. Metadata, which means data that describes other data, can describe a document and link related records. This can help organisations apply permissions, identify approved versions, place content in workflows and set points where a human must step in. The article says this becomes more important as AI starts to act for the business.

For businesses, the takeaway is that enterprise AI may need more than new models or agents. It may also need trustworthy content, clear ownership, controlled access and rules for use. TechRadar presents document management as a less visible but important foundation for AI returns. It says big conversations about AI often focus on models, agents and fresh capabilities. Those matter, but the article argues that the information layer deserves equal attention.

Sources

  1. The overlooked technology that could determine the ROI of enterprise AITechRadar

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.