Editorial
A truly impressive amount of energy is currently being devoted to AI in the workplace. But this enthusiasm will soon be derailed unless we move beyond a very lazy lumping of everything into one AI category.
There are many different AI products in the wild, and an even greater array of underlying capabilities and functionality. Yet we often see amazing but hugely specialist solutions like Google’s protein-folding AI being mentioned in the same breath as GenAI-based copilot solutions that offer broader but more mundane capabilities.
To progress as an industry, we need to put better shape around our AI thinking, settling on common language and terminology and recognizing that different use cases will be tackled in different ways.
More immediately, we need to make sure we focus our precious energy and attention on business needs and solutions that will generate clear benefits. Unfortunately, we’re seeing the exact opposite: businesses (and vendors) are currently focusing on the wrong AI use cases, while overlooking more fruitful targets.
It’s not too late to change tack, but the starting point is to be clear on what we’re all talking about.
Putting Names and Shape Around AI in the Workplace
Last month, Step Two released a model that puts shape around the many use cases that AI can address in the workplace.
The AI framework for the digital workplace identifies 19 general use cases, plus eight more specialist use cases.
The framework breaks capabilities into four main categories (plus an ‘x-factor’):
- Content Creation: Using AI to produce information ranging from code to the written word, images and videos.
- Information Retrieval: Improving the fundamentals of findability through to new ways of accessing and interacting with information.
- Knowledge Management: Using the full power of AI to answer questions and organize information, all the way up to driving innovation.
- Productivity and Process: Cutting back on day-to-day basic admin, through to AI-powered digital assistants.
- X-Factor: An acknowledgement that many new ideas and approaches have yet to be conceived.
Within these overall categories, the general use cases identify broadly applicable approaches across many organizations, business areas and job roles. The specialist use cases are the super-cool things that you’ve read about in the newspaper, but either target a very specific need or are out of reach for most businesses.
With this framework in hand there’s real power in being specific about which AI use case we’re talking about. The benefits vary wildly, from minutes saved by the broad use of copilot-style productivity tools through to narrow (but potentially very large) gains from more specialist taxonomy creation tools, for example.
The risks are also very different. An AI-generated meeting summary missing a few points or action items poses few practical consequences, but a customer-facing call center giving incorrect advice based on AI-powered search results could have potentially large impacts.
Having laid the groundwork, let’s look at why I think we’re focusing on the wrong use cases right now.
Related Article: Generative AI Is Shaking Up People Analytics. 4 Use Cases
‘Copilot’ Productivity Tools
There is a whole suite of productivity tools and admin aids, branded as "Copilot" by Microsoft, but also offered by many other vendors. These offer the delightful prospect of automating drudge work, such as writing meeting summaries, highlighting key emails in cluttered inboxes or sorting out to-do lists.
Almost every day, vendors and users are finding cunning new ways to use these capabilities, whether it’s tackling another task or improving the impact of existing AI usage.
It would be easy to think it’s all about Copilot, based on the deluge of comms and marketing that’s coming from Microsoft (and partners). Cynically, it could also be pointed out that with per-user-per-month licensing for Copilot, there’s a clear reason why it’s getting the bulk of the focus.
The business case for Copilot-like capabilities, however, is broad but shallow. A lot of people can potentially benefit from this AI use case — in many different ways — but the benefits delivered are incremental, not transformative. History has shown that enterprise business cases built on minutes saved by employees struggle to get cut-through. Not least because it’s never certain whether those time savings will lead to other useful work being done, or just a lower intensity of work.
While Copilot is talked up as the next big thing, this glosses over the potentially daunting change and adoption effort that will be required. To fully realize the benefits of Copilot, individuals need to change how they work, not just tweak their current practices. There’s also an art to creating good prompts that will require large-scale training.
Experience is increasingly showing that a high-touch approach needs to be taken to the adoption piece, and it’s not clear which team will spearhead this, and where the resources will be found.
So in short: lots of people get some benefit, but not without a lot of effort put into change and adoption. So maybe this isn’t (for now) the killer use case?
Related Article: Change Management in the AI Age: How to Sidestep Common Mistakes
Finding the Needle in the Haystack
Another use case, which we’ve called "digital workplace information retrieval" (I know, that’s a clunky name — all suggestions welcome), involves using Copilot (or similar tools) to draw on everything in Microsoft 365 to provide answers or create content. The AI looks at everything that you have access to, including intranet sites, document collections and collaborative activity. This is being touted as a replacement for search, and the death of the intranet.
There are huge risks related to this, due to the terrible state of information management in every organization, regardless of industry or size. We’ve all seen entire network drives dragged-and-dropped into SharePoint without any cleanup. Permissions are also a mess, typically evolving organically over time without any overall policy or oversight.
As a result, incorrect (or sensitive) content will almost inevitably be surfaced. Think “tell me about the coming organizational restructure” or “provide a list of current customer complaints."
It's bordering on techno-utopian to think that AI can make sense of the mess of conflicting, duplicated, out-of-date and poorly written information that clogs up our digital workplaces.
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
The business case also rests on the troubled minutes saved argument, except in a few specific business scenarios. Do the risks outweigh the benefits? Is the effort needed to impose (much-needed) information management practices undermine the short-term opportunities? Perhaps.
Related Article: What the Ancient Greeks Can Teach Us About Enterprise Search
Querying Well-Managed Information
The final use case I’ll look at is what we call "querying information collections." This uses GenAI with retrieval-augmented generation (RAG) functionality to draw on a managed information collection, such as a policy and procedure library.
With it, employees can ask questions such as “what leave benefits am I entitled to?” or “what documents are required for claiming travel costs?” It can also summarize policies to provide succinct information for personal use or to share with others.
This use case is less frequently discussed, even though the tools involved are relatively straightforward, and with little effort, give reasonably good answers. More work will be required to give great, highly accurate results and this will almost certainly involve revising and improving the content that’s being used.
Because this still requires good prompts, this use case works best when supporting a highly trained workforce, such as call center workers. They are also well-positioned to interpret the answers provided, and to sense check whether they’re correct.
In general, the manageable risks and clear business benefits in a range of situations make this one of the use cases to focus more on.
Cutting Through the Idle Chat
Uses cases one and two are mentioned the most, despite their problems. Use case three is flying under the radar — and many of the other use cases in our AI framework are barely mentioned.
So: are we currently focusing on the right AI use cases in the digital workplace? Comments encouraged!
Learn how you can join our contributor community.
