The Big Challenge of Enterprise: Generative AI and Knowledge Management

Generative artificial intelligence, the key driver of large language models (LLMs) like ChatGPT, has gained a reputation primarily as a way to automate content creation.

And while it is highly adept at writing blogs and white papers and generating spoken words, computer code, and even works of art, its broader contribution to the business model is likely to come from its capacity for knowledge management.

KM is essentially how organizations optimize available data so it can be leveraged to its fullest extent. The larger the organization, the more data it accumulates, so there must be an efficient way to store, categorize, retrieve, process, and disseminate data at high speeds and with optimal performance.

A Heavy Lift

However, this can be a herculean task, so it is no wonder that many organizations are eager to unleash the power of AI on their KM platforms. But what about GAI, in particular, has made it such a promising development for this particular application?

It turns out that GAI is so good at creating new content because it can locate, combine, and otherwise transform existing content to suit new objectives. If, say, someone needs a white paper explaining how different products can be used together to tap new markets, GIA can seek out all the relevant product data, market analyses, and other sources of knowledge to do it.

David Pickering, senior technical marketing engineer at Australian automation developer Ivanti, says this allows it to not only generate more knowledge from existing knowledge, but the new knowledge will be more accurate, engaging, and contextual for any given project. In turn, this gives GAI the propensity to generate new insights and ideas and provide faster problem resolution.

Searching with Artificial Intelligence

GAI’s enterprise search functions alone fuel?an entirely new level of productivity in the enterprise, says Eddie Zhou, founding engineer of search developer Glean. Not only does it allow knowledge workers to interact more naturally with digital work assistants, elevating their ability to quickly find the information they need, but with the proper training, it can even translate between multiple languages to provide access to knowledge that would otherwise remain locked away.

Another key advantage is the ability to automate security, privacy, and access across the knowledge base. Before any content is released, GAI can automatically compare the permissions embedded in the data to those of the requesting employee. If they don’t match, the data cannot be touched. In this way, organizations not only simplify the dissemination of data but enhance its protection as well. (Full Disclosure: I participated in the webinar this article was drawn from.)

Boosting Knowledge-Dependent Applications

GAI is already making its way into applications that populate critical enterprise functions, such as human resources, customer service, and application modernization. IBM’s Institute of Business Value estimates it can improve customer experience metrics by as much as 70% using tools like augmented generation, summarization, and classification. This allows applications and platforms to tailor their output based on customer history and sentiment.

Meanwhile, GAI can refine job classifications based on required skills and responsibilities in HR, then match those points to a pool of applicants or even search for appropriate candidates elsewhere. And across the enterprise, GAI can continuously update applications as processes, organization structures, and even business models evolve to meet changing demand.

The common thread in all these contributions is that the human workforce spends less time doing tedious, repetitive work and more time pursuing higher-level strategic objectives.

GAI, Train Thyself

Coming full circle, GAI can even enhance knowledge management to train new AI models, including those built on GAI itself. In a recent article in Harvard Business Review, Thomas H. Davenport, distinguished professor of IT and management at Babson College, and Maryam Alavi, professor of IT Management at the Georgia Institute of Technology’s Scheller College of Business, explain how GAI-based knowledge management can be used to fine-tune existing models or build them from scratch.

In some cases, GAI can customize content for a particular model using domain-specific knowledge. In this way, a single model trained on basic knowledge and language-based interaction can address a wide range of medical, legal, or financial queries based on the user’s needs. The challenge is to expose the model to sufficiently large data sets without broadening the scope to the point where it is trying to satisfy too many disparate attributes.

Another key to successfully implementing these processes is data quality, which can be automated to some extent but ultimately relies on the human workforce for proper curation. Without this, GAI has a tendency to “hallucinate“, that is, assert claims as fact based on data that is incorrect or nonexistent. No matter how a model is trained, constant evaluation is essential once it enters production environments.

The Bottom Line

Sir Francis Bacon is said to have coined the phrase “knowledge is power” (scientia potential est), although similar expressions date back to some of the earliest known writings. In today’s digital age, knowledge-driven power has increased by incalculable factors with the adoption of massive data processing and widespread dissemination. Artificial intelligence is poised to push this to an entirely new level, with an outcome that is still hard to fathom.

Using GAI as a means to manage knowledge will be crucial to navigating our increasingly connected world going forward, but it can also lead to false conclusions and poor insight if not developed and managed properly. Getting this right is probably the most important challenge facing the enterprise right now.