On the heels of the recent solar eclipse, I’ve been thinking about how Generative AI (GenAI) has totally eclipsed everything in the technology space over the previous few years. Almost every session at Enterprise Connect last month had a GenAI element, if not a dominant GenAI focus.
But we are moving out of the zone of totality as we progress through 2024.
In 2023, Gartner’s Hype Cycle plotted GenAI in the ascendant “Peak of Inflated Expectations” zone, but does that mean it is necessarily headed towards the dreaded “Trough of Disillusionment?”
There have been several high-profile AI failures that folks point to as if to say, “See, it’s not ready!” Even on this very site, leading analysts have been tamping down expectations.
So what’s really going on? Why is GenAI taking some lumps in the real world?
Many of the problems we are experiencing in the early stages of GenAI have to do with inaccurate or out-of-date data.
Last month at Enterprise Connect 2024, I had the opportunity to interview NICE’s Aaron Rice, GM of CXExpert. Aaron asked me an unexpected question: What’s in your fridge? Is anything spoiled or expired? Did other people leave things in there? His point was that anything you wanted to cook would only be as good as the quality of the ingredients in your fridge – not how well those ingredients are organized on the shelves, but how fresh they are. The same goes for knowledge management. It doesn’t matter how good the AI tool is if the organizational knowledge is expired and spoiled.
This is both a huge problem and a huge opportunity. The good news is that the right AI-powered tools can help clean up your knowledge management mess and help maintain it so that data remains unspoiled and fresh. Aaron went on in our interview to talk about how AI can help identify areas of concern and how using Gen AI can add to or replace out-of-date-knowledge.
One of GenAI’s advantages is that it makes AI accessible to almost anyone compared to previous AI implementations or tools. The unintended consequence is that both end-user customers and some vendors have been implementing their own customer-facing AI solutions based largely on publicly available models.
As my colleague Martha Buyer pointed out in her recent No Jitter post, this can be problematic, as many organizations don’t have the experience, knowledge, or guidelines to properly deploy GenAI on their own. There are legal, financial, and ethical risks that all go into AI deployments that have to be fully understood before acting.
This is why most organizations should partner for their AI solutions. The DIY approach just carries too much risk. This focus on partnership allows organizations to focus on their business needs and let their partners figure out all of the details and potential pitfalls. I’ve sat through dozens of briefings on AI, and I’m always pleasantly surprised at the care and critical thinking that has gone into purpose-built AI solutions, where a lot of thought has gone into addressing potential areas of concern. By partnering, you can learn through others’ mistakes rather than making mistakes on your own.
While I agree we can’t let AI loose on all self-service customer interactions, it absolutely has a place today in many areas. The first temptation is to go all-in on GenAI and fly full-speed ahead into potential disaster. The second temptation is to do nothing for fear of making mistakes. Both approaches carry risk: the risk of making critical mistakes vs. the risk of falling behind competitors.
The key is to target the use of AI, starting with use cases with the lowest risk and highest gain. Many of these will initially be something other than customer-facing interactions. However, as adoption accelerates, some AI solutions can recommend and even help build AI-driven interactions while mitigating risk factors.
One of the reasons I believe AI will live up to the hype and may very well avoid the bottom of the “Trough of Disillusionment” is its ability to improve itself. AI can make itself better over time. For example, if an AI chatbot failed at an interaction and an agent had to take over the chat, with the right solution, the software could learn what the chatbot did wrong and what the agent did right to properly handle the next similar interaction.
In summary, successful AI depends on a clean and well-stocked fridge, choosing the right partner, and a measured approach to dialing the right amount in the right areas.
The next eclipse in the US isn’t until 2033; for the lower 48, it will be in 2044.
When will the next great technology eclipse be, and when will it happen?
This post is written on behalf of BCStrategies, an industry resource for enterprises, vendors, system integrators, and anyone interested in the growing business communications arena. A supplier of objective information on business communications, BCStrategies is supported by an alliance of leading communication industry advisors, analysts, and consultants who have worked in the various segments of the dynamic business communications market.
