Behind closed doors: Exclusive insights from our conversation with Kane Simms

“I’d rather be in boost.ai’s position of understanding what it takes to make enterprise conversations work, then layering in generative AI.”

That’s just one observation Kane Simms shared during our recent VIP dinner event the night before Boost Camp 2025. Set against the incredible backdrop of the MUNCH Museum and with rare sunlit views of the Oslofjord, the gathering offered attendees a rare opportunity to engage in thought-provoking conversations with Kane and fellow CX leaders. But it was the discussion between Kane and boost.ai CCO and Co-founder Henry Vaage Iversen that truly captivated. This wide-ranging conversation left attendees with a deeper understanding of AI’s potential and its challenges.

Here are key takeaways from the evening that set the stage for the rest of Boost Camp 2025.

Why generative AI isn’t the silver bullet organizations think it is

It’s the question on every business leader’s mind: How does generative AI compare to more traditional natural language understanding (NLU)? While plenty of new market entrants have built platforms from the ground up on large language models (LLMs), often hyping them as truly ‘Agentic,’ Kane and Henry agree: Generative AI alone isn’t enough.

“Newer entrants have an interest in innovation, but overly rely on LLMs,” said Kane. “You still need to use rules, which means you still need to use NLU. If you need to use NLU, you still need a robust enterprise infrastructure and integrations.

I’d rather be in boost.ai’s position of understanding what it takes to make enterprise conversations work, then layering in generative AI."

Kane pointed to Salesforce Agentforce, which went all-in on generative AI with its first version release. “And then version two included what? Variables, the ability to take entities out of utterances — you know, rules,” he said. “So it’s like, ‘Oh, brilliant, welcome to 2018.’”

That doesn’t mean organizations shouldn’t embrace generative AI.

"The secret to a really great experience is multi-modality"

Kane said. “You might need a bit of generative AI to handle variations in questions — some of the messiness of the conversation — but conversational AI to capture data. To handcraft and design the best experience at every turn, you need a mixture of the two.”

But how can organizations strike that perfect balance?

“Look at the use case,” Henry advised. “In general, more complex questions are better suited for generative AI, while NLU can handle simpler ones with predefined responses.”

The right approach can have major results. Kane explained to attendees that one UK-based bank automatically devolves disambiguation situations where NLU can’t understand a query down to a retrieval augmented generation solution. 

“Initially, [the solution] only answered 36 percent of questions,” Kane said, “but the more data they added, the better it performed. Now, they’re able to answer 80 percent of questions with 99 percent accuracy.”

AI’s hidden power is driving demand you didn’t even know existed

One of AI’s chief selling points is its ability to increase efficiency at scale without driving up costs. But what if AI also had the power to uncover new demand? That’s exactly what Kane experienced while helping a government agency digitize its services.

“Staff were concerned that if we put everything online, they were going to get more work,” he said. “I thought that wouldn’t be the case — that the demand was the demand, regardless whether it was over the phone or online. But, shortly after we went live, demand went up. People were going to the website anyway but, when they didn’t see a particular service available, they didn’t bother following up. Now, the government was able to capture that demand.”

It was a simple but stunning realization: that organizations have no idea how many people aren’t reaching out to them due to poorly designed support offerings, perceived complexity, a misunderstanding of available services or any number of other issues. 

Boost.ai’s experience provides further evidence to support this.

"Any time we work on a project,” said Henry, “we see an increase in traffic — anywhere from 10 to 15 percent

— simply because a service is available that wasn’t before.”

Even more unfiltered insights from two of AI’s top experts

The conversation covered significant ground, including key themes such as:

• Kane on what good conversational AI looks like: “The first principle of any good conversational AI system is to first and foremost understand what someone said, what their intent is, or what they’re trying to achieve.”

• Henry on the changing role of AI: “Our client, [Norwegian Tax Authority] Skatteetaten’s human advisors now escalate to AI for complicated questions. We used to talk about ‘high volume, low complexity’ issues, which we’d send to chatbots, but that’s changing.”
• Kane on the challenge of testing generative AI: “It’s so multi-faceted because it’s probabilistic. Google found there were 7,000 ways to express that you want to set an alarm. Generative AI can deal with that flexibility, but where do you even start with testing?

Anyone who isn’t putting in the proper effort to figure that out is going to struggle."

• Kane on trends in generative interfaces: “The more people interact with something like ChatGPT, which isn’t quite a generative interface but is full screen and immersive, the more they’ll transfer those expectations to other websites and apps.”

• Henry on the risk of prioritizing automation over resolution: “You can obviously automate 100%, but you’re going to sacrifice customer experience on that project. The best experience is to have a human waiting to answer, but that can be costly, and most people don’t want to make that tradeoff.”

The power is in your hands — what will you do with it?

In addition to a delicious dinner, Henry gave attendees serious food for thought when he asked them,

"If you could have a 1:1 relationship with each of your customers, what would you do with that connection?"

Now, it’s your turn. Picture your customers — their wants, needs and challenges — and what you might say to them or do for them if you could engage with each one on a personal level. With AI Agents, this kind of direct, meaningful relationship is no longer a possibility, it’s a reality. So, what will you do with that connection?

Let us know in the comments. And, when you’re ready to bring that vision to life, give us a call.