Rob McDougall is the driving force behind Upstream Works Software. He is passionate about using the best technology to make the agent experience better. With so much technology change, this is hard to do, but the stakes have never been higher for contact centers.

Artificial Intelligence (AI) application integrations hold great promise, but remain poorly understood by contact center leaders and is overhyped by vendors. I recently interviewed Rob to clarify this – see the highlights below.

Reality Check: The Basics of AI

There is certainly a lot of interest in AI application integrations from contact center leaders, and for Rob it’s “high on the hype cycle.” He likened this to Skynet from the Terminator film series, where AI is being touted as an all-knowing solution that can address all of the contact center’s challenges, but “it’s not the total package yet.”

Rather than think about AI as an engine to drive automation in order to replace agents, Rob sees it as a way to augment their jobs. As such, AI should be viewed as being about developing specific applications for specific use cases or to address specific problems. Coming back to the Skynet reference, this is about artificial general intelligence (aka strong AI), where AI can handle everything, at which point the need for agents is greatly diminished. Rob feels we’re “20 years away from that, and what we have today is instead weak AI, or narrow intelligence.” This form of AI is task-specific, where a high degree of training is required even to do fairly basic things like simple forms of self-service.

Today’s level of AI can handle this fairly well, but “it’s not ready for long tail applications, such as what to do when an agent goes off-script,” Rob says. Conversational AI (CAI) can do things like semantic matching and inference, but current expectations shouldn’t be set much higher at the moment. Even this, however, still makes AI worth pursuing, as Rob notes, “the agent’s job is harder now.” As we move into the world of digital CX, there is more data being generated and more channels to manage. In response, agents need better tools.

Core Use Cases for AI

For Rob, the idea of “narrow intelligence” is about AI use cases that make the agent experience better. Rather than reduce the number of agents, AI should be used to drive efficiencies in their work and raise service levels that will ultimately lead to improved CX. Good examples would be reducing Average Handle Time (AHT), having fewer repeated calls, and lower call volumes overall.

Each of these help agent efficiency, but each also needs a specific AI-driven application. To illustrate, Rob cited using conversational AI to pre-collect information about the customer, so the agent only spends time dealing with the problem at hand. This allows agents to make more efficient use of their time and leads to faster service for customers.

While this helps build the case for using AI application integrations in the contact center, Rob also cautioned about setting realistic expectations. Clear parameters must be used in terms of how far you can go with conversational AI to automate tasks like pre-collecting customer data, or helping customers navigate forms. If there’s too much ambiguity in the data, conversational AI may end up providing inaccurate information to agents or get stuck in a loop asking for the same information over and over. In these cases, the benefits of AI would be undermined, and CX would no doubt suffer.

Decision-Makers’ Resistance to AI

Not surprisingly, the dissonance between AI’s promise and what it can actually deliver creates some resistance and holdbacks among contact center decision-makers. Aside from the natural desire to adopt the latest technology, many see AI as a bit of a silver bullet to address the challenges of meeting today’s CX expectations with a core set of legacy technologies. In Rob’s view, customers are still “star struck by AI, and rightly so.”

Not only are there a lot of new AI offerings from unfamiliar vendors, but internally, contact centers are still very early on their AI journey. Aside from all of this being new, use cases need to be determined, and from there, business cases still need to be developed. To all of this, Rob’s advice is to “just pick something.” The important thing is getting started on this journey, especially since AI is a family of iterative technologies – such as machine learning and natural language understanding – that learn over time, and continuously improve. Better to start with some basic applications that AI can do well right away, than having to keep training agents to learn as they go.

Rob’s view is, “don’t think too big, look for simple interactions that you get a lot of.” This approach to AI is more about hitting singles than home runs, as there will definitely be a learning curve. Also, it’s easy for IT to underestimate how things can get complicated quickly with AI. If IT takes on too much – like trying to build a complex chatbot – deploying it properly could prove problematic.

How Contact Center Leaders Should Think About AI

This is about setting realistic expectations and builds on the above theme. “Deploy it narrowly, not organization-wide,” Rob suggests, where the focus should be on contact center needs, and applications that AI can effectively address. Perhaps more importantly, AI should not be viewed as a horizontal solution where one application can address all needs.

Instead, Rob says contact center leaders need to “recognize that each problem set needs its own AI application – utterances, conversations, translations, etc.”

As AI use cases grow in the contact center, so will the number of applications, and with that will come new integration challenges. To compound matters, these various applications will likely come from multiple vendors, so IT must be prepared for all this.

At this point in time, AI is far from being plug-and-play, so choosing the right technology partners will be critical. Rob’s caution here is to make sure AI vendors “have specific expertise, not just in contact center, but for the specific problem set you’re trying to address, and possibly for the vertical market this applies to.”

Rob’s main takeaway here is that vendors who can provide an integrated desktop have a key role to play when deploying AI to improve the agent experience. If AI applications are added in a standalone fashion, they will create more silos, making the task of integrating all of them into a seamless agent experience much more difficult.

The key to success is to mitigate the creation of these silos, especially since AI is constantly evolving. All of this should be transparent to agents, so this is more than a matter of adding an AI application without consideration for the bigger picture.

As an AI evangelist, Rob and Upstream Works are Helping AI Help Agents with AI application integration capabilities to help contact center leaders realize more value from AI.