Generative AI (GenAI) is being brought into organisations to provide a wide range of capabilities, as outlined in our AI framework for the digital workplace. In many of these use cases, the ‘superpower’ of the AI is to draw on internal documents and resources (supplemented by its existing ‘knowledge’) to give clear answers or to write organisationally-specific content. It’s remarkable to see this at work, whereby typing in a carefully-considered prompt can generate a clear response, even when ‘making sense’ of huge and complex source material.

But if this is GenAI’s superpower, then its ‘kryptonite’ is its dependence on good information to draw on. Out-of-date information will produce incorrect answers, and missing details can lead GenAI to ‘hallucinate’ completely made-up responses. (As they say: garbage in, garbage out.)

There are a wide range of sources across the digital workplace that could potentially be utilised by GenAI. To be valuable, these sources need to contain information that’s useful. They also need to have robust and effective governance that’s sustained over time. So, which of the major information repositories will be of use to GenAI?

Let’s look at each of these information sources in turn:

Intranets

Intranets provide a wealth of useful information, from corporate policies and news, through to operational updates and audience-specific resources. Following the ‘shop window’ metaphor, intranets contain information that’s intended to be shared, making it highly valuable for GenAI solutions.

In general, intranets are governed, either with a formal set of governance documents and guidelines, or with a set of informal practices that are broadly followed. The fundamentals of effective lifecycle management are also often in place, with usage reports and content review workflows. That’s not to say intranet governance is perfect–far from it!–but at least the basics are generally covered off, with the intent to do more.

Knowledgebases

Knowledge repositories vary greatly in size and purpose. They can be intended to provide comprehensive coverage of a topic, such as engineering guidelines or support resources for a call centre; or they can address specific needs, such as help articles in ServiceNow. Almost all provide valuable source material for GenAI, depending on the depth and breadth of the knowledgebase.

Governance is much more problematic when it comes to knowledgebases. While ISO-compliant engineering guidelines may be tightly controlled, this sadly tends to be the exception rather than the rule. Frontline knowledge resources are often under-appreciated and poorly managed (despite being business critical), and self-service content in ServiceNow (and similar solutions) is almost always totally ungoverned (and unowned). Without good governance that ensures that information is current and correct, knowledgebases are an uncertain resource for GenAI.

File servers

There are tens of thousands of useful documents on corporate file servers, scattered in among millions of other files. There will be multiple versions of many documents, along with drafts, conflicting information and out-of-date resources. With a ‘drag-and-drop migration’ from file shares into SharePoint document libraries, the same problems remain. It’s wildly optimistic to think that GenAI can ‘make sense’ of all this, giving clear and accurate answers where the source material is so muddled.

File servers are the ultimate expression of the ‘Wild West’, with thousands of people creating, updating (and occasionally deleting) documents without any overall guidelines or structure. While GenAI won’t be baffled by idiosyncratic file naming, it will find the ad-hoc access rights to be a major stumbling block. It’s hard to see how sufficient order can be brought to file servers to make them a valuable AI resource.

Collaboration spaces

Work is increasingly being done in tools such as Microsoft Teams, which now provide powerful capabilities for teams to manage their conversations (and the documents that go with them). Collaboration spaces are the natural home for project documents, which can grow to become a substantial resource. The challenge, however, is to separate out the valuable information from the general discussions (and debates!) and the myriad in-progress and draft documents. There is a very high risk that GenAI will return incorrect information, when trying to sort through the sheer volume of collaboration activity.

Collaboration spaces, aren’t tightly managed with formalised governance. There may be some ‘rules of the road’ in terms of appropriate usage, but otherwise it’s largely ‘anything goes’. At best, there will be governance around the creation of new collaboration spaces, and a mechanism for retiring them when they are no longer needed or used.

Records management

Records management systems are distinctly different from other information platforms in that they are, by definition, tightly governed. Records management policies define (in detail), where records should be stored, how they should be structured and how long they should be kept for. Government agencies often have the tightest set of records management policies, to ensure that key decisions are captured and retained.

It’s unlikely, however, that records management systems will be of value to GenAI solutions. Records, by their very nature, are snapshots of the past, potentially going back decades. The intent of records policies is also to capture everything from across the organisation, from the smallest decision in a local business area through to major policy updates. Records management systems are designed for retention not retrieval, and even GenAI solutions will be overwhelmed by the sheer volume of documents in these systems.

Enable GenAI superpowers

Of the platforms listed above, intranets are the only clear ‘winner’ as a resource for GenAI. To enable these true AI superpowers, intranet governance must be greatly improved, utilising a formal and documented structure such as the Intranet Operating Model. The value of intranets as a foundation that underpins AI needs to be better understood and articulated, to ensure that they are sufficiently resourced and supported.

Knowledgebases, by their very nature, can be greatly enhanced by the addition of GenAI. Unfortunately, establishing sufficient governance often means starting from a blank sheet of paper, and a lot more needs to be done to recognise the importance of knowledge management for frontline staff and other similar cohorts.

More broadly, effective information management is needed to bring more order to the wider digital workplace. This is no easy task! Organisations gave up trying to put structure around file servers long ago, and to date there hasn’t been much of an effort applied to document management in platforms such as SharePoint.

Generative AI will undoubtedly change the way we work, so let’s put in the hard work now to shape and manage the information that these solutions will require–so we can benefit from its superpowers!