RevGen’s 5 AI Themes to Watch from the Gartner IT Symposium and Xpo

At the 2024 Gartner IT Symposium and Xpo held this October, the focus was on AI—where it’s going, how fast it will get there, and what its limits are. We had the pleasure of attending and absorbing the nearly overwhelming amount of thought leadership. Despite the variety of topics, there were five key AI themes that resonated most with what we’re seeing in the market and that we’ll be following closely in 2025.

The 5 AI Themes to Watch

1. The Acceleration of Generative AI

Unsurprisingly, the most discussed AI technique by far was Generative AI (GenAI), which according to Gartner’s research is already used by 29% of leading enterprises in the US, UK, and Germany. It narrowly beats out Machine Learning (28%) and Natural Language Processing (27%), suggesting that businesses are betting on GenAI to increase efficiencies, whether that is in text, image, or video generation—and have accelerated their usage of GenAI within a very short window of time, generally less than two years.

As with many new technological advancements, GenAI is seeing more resources and investment because of these bets, yet most models aren’t making it out of the Proof-of-Concept phase and into full-blown production. Sound familiar? Think of the early days of machine learning-driven data science models.

This has had a negative impact on the overall ROI of GenAI so far, however, experts predict this will shift in 2025 as companies move to productionalize these models, leveraging further acceleration in Large Language Models (LLMs), data-readiness techniques, and other tools and innovations.

While some organizations will move quickly and others more steadily, we’ve found that both approaches can bear fruit, as long as strategy, investment, and level of AI governance follow accordingly.

2. The Rise of Agentic AI

Chatbots have long been a staple of the customer service world, however the new capabilities of AI have led to the development of virtual assistants in nearly every industry and function. These can range from personalized customer experience bots with more natural language capabilities and interactions to digital co-workers that can operate more autonomously to solve problems and complete complex business and IT tasks.

Some of the tools to create these AI agents are already generally available, such as Salesforce’s AgentForce and Microsoft CoPilot’s Agent Creation tool in CoPilot Studio. Others are in product development and are coming to market quickly, such as UiPath’s Agent Builder. This is expected to increase over the next year, with even more major software vendors providing these embedded agentic capabilities to enhance their platforms.

The key difference between these new AI agents and traditional bots is the “autonomous” part, where they are capable of making decisions based on memory, planning, analysis, and the technologies they are tied to—for more ambiguous tasks than what we’ve seen to date.

Of course, autonomy comes with a price. How do you control an AI employee? What kind of guardrails and security need to be in place?

All these questions will be on CIOs’ minds as the digital, AI-driven workforce ramps up in 2025 (and directly inform Theme #5).

3. Multi-Modal Large Language Models

One of the AI trends that’s less likely to make headlines while still making a big impact is the maturation of Large Language Models. Specifically, the shift from text as a primary (or often sole) source of data to incorporating sound, image, video, and other unstructured data.

Think about it like this: When an organization trains employees on a new process, they don’t just pass out documentation and call it done. There are lunch and learn sessions, slide decks, verbal Q&As, videos to watch, and yes, documentation, too.

Also, when an organization stores data, oftentimes that documentation is in formats such as PDFs, which are difficult to mine for value. In fact, Gartner shares that only 10%-30% of organizational data is structured. In other words, very little data is quickly accessible and understandable for ease of use in organizational operations, such as customer renewals, sales outreach, accounting, etc.

It makes sense, then, that LLMs would improve their outputs with a wider variety of inputs. To continue mining the value of GenAI with organizational “dark data”, it will be important to evolve the versatility of the models that power it in 2025. In good news, we’ve already seen tremendous strides just over the past year.

4. The Data Bottleneck

As we go further down the path of AI maturity, there is one thing that is increasingly clear—most data is not currently ready for AI usage. As referenced in the multi-modal AI theme, data can reside in all formats and corners of an organization, with only an estimated 10%-30% being structured. The real magic will be in making that data accessible in a translated and more understandable manner for use by AI. By the way, AI can help with this task too.

While data management has been an IT priority for what feels like decades now, the goal is shifting from centralization and integration to intelligent retrieval and contextual relationships for AI readiness. Without more related and easily accessible and understandable data to augment and fine tune AI models, the models simply cannot advance to the level of tailoring organizations need to gain a competitive advantage and sufficient ROI.

It is important to note that this does not mean physically centralizing and integrating all data akin to our data management techniques of the past, but rather leveraging more advanced techniques to attain, translate, and relate all organizational data.

To make data AI ready, we expect to see need-driven innovation in areas such as advanced knowledge graphs, semantics, intelligent document processing, and data retrieval technologies, all of which will fuel the AI surge.

5. AI Governance and Security

It’s been mentioned in this article already, but the topic of AI governance, security, and risk came up enough to warrant its own callout.

Now that Pandora’s AI box is open it can’t be closed, so how do you mitigate the risks?

The unprecedented size and speed of the AI wave is generating a need for new tools and frameworks to ensure that it doesn’t expose a company to security risks, privacy issues, and even unintended consequences that can cause bigger problems for employee productivity (think productivity leakage). Gartner shared their category of these newly advancing tools and techniques under the acronym of TRiSM, or Trust, Risk, and Security Management.

And these are real risks, not hypotheticals. In fact, Gartner shares that almost 30% of enterprises having deployed AI have experienced an AI related security breach.

As for potential customer-facing risks, think of all the times when someone has asked ChatGPT questions that returned incorrect answers, or hallucinations. These risks will only be exacerbated when these tools are exposed by organizations to their customers directly, and without human oversight. For example, an AI chat bot inaccurately offering customers a deal for a product or service that isn’t an approved promotion for that specific product or service at all. How will companies in that situation adhere to something committed on their behalf by an AI employee?

Every company that is pursuing AI – and this is almost every company – needs to have a strategy for governance and security so that this AI wave doesn’t become a tsunami that can wipe a business out. Different levels of risk and responsibility will be applicable to every organization depending on their AI strategy and implementation pace set. There’s a big difference between an embedded AI strategy and a custom build your own AI strategy – is your vendor responsible for implementing guardrails or are you responsible?

As we head into 2025, RevGen will be keeping a close watch on how these themes evolve and how they impact our clients. Will they spur new technologies? Will one LLM rise above the rest? Will there be new security risks we haven’t even anticipated?

One thing is guaranteed to be true: it’s going to be an exciting and eventful year.

To learn more about our approach to Artificial Intelligence and the services we offer, visit our site. We also offer AI Accelerator Workshops, to fast track you to ROI from your AI initiatives.

Pero Dalkovski is RevGen’s Vice President of Data and Technology. He has spent his career helping clients strategize and implement innovative data, analytics, and technology solutions that deliver business value.