How Maven AGI’s CX leader-turned-CEO approaches AI agents
Long before he started selling an artificial intelligence (AI) platform to customer experience (CX) professionals, Jonathan Corbin was among those desperately in need of one.
The CEO of Boston-based Maven AGI has joined the ranks of executives touting the promise of AI agents that can help companies offer fast, accurate answers to their customers’ most common and complex questions. In his prior role, that’s exactly what Corbin said he and his team lacked, despite his then-employer’s reputation and pedigree.
At HubSpot, where he worked as global vice-president of customer success and strategy, Corbin said he saw the marketing software-as-a-service (SaaS) firm grow from hundreds to nearly thousands of people, and from hundreds of millions to more than two billion in revenue. With all that success came a common challenge.
“We were really struggling to get the information we needed in order to engage with our customers at scale,” Corbin told 360 Magazine amid the launch of Maven AGI’s enterprise AI agent platform. “The foundation of being able to deploy AI agents to address that, though, is that you need to bring together disparate data sources, otherwise you’re going to be in a much worse state than you were before.”
Maven AGI’s platform includes more than 50 built-in integrations designed to unify all that data, along with the ability to proactively identify gaps, contradictions and outdated information.
Corbin offered some advice on how his peers in the CX community can assess the AI agent technologies they need, and how best to deploy them:
Map out the three most common customer inquiries
Most AI agent solutions are aimed at helping customers answer a questions around what Corbin described as “How do I do a thing?” This could be setting up a product, making use of its features or managing a product return, for example.
That’s table stakes, Corbin said. Another key area of inquiry involves a search for information that’s relevant to an individual customer.
“At HubSpot, about 30 per cent of inquiries came from customers who were trying to find out more information about their bill,” he said. “You have to be able to leverage different systems, pull information out and talk specifically about that persons’ account. The complexity is challenging for other systems to be able to pull that information.”
The third bucket – and potentially the one with the biggest opportunity for so-called agentic AI – includes questions that require a company to take action on a customer’s behalf. Corbin used the example of a streaming service, where customers run into difficulty cancelling or even pausing their subscription.
Beware AI based on generic knowledge
There’s a danger in training AI agents on large language models (LLMs) that have simply scoured the public Internet, Corbin warned. He suggested it was akin to doing basic research on a search engine.
“If you’ve ever Googled anything and been directed to a random Subreddit, oftentimes it’s definitely not accurate, and you wouldn’t want to use it to serve your customers,” he said. “We’ve been working with banks where we have to go in and find out answers to things like, ‘What’s your average deposit every month?’ or ‘What was the last check drawn from your account?’ This is not generic knowledge.”
Strive for push and pull, rather than rip and replace
Some CX leaders might be wary of deploying AI agents that require them to not only introduce new technology but shut down existing toolsets. This doesn’t make sense, Corbin argued. After all, many firms have already deployed technology from Salesforce for their sales teams, Adobe’s Marketo for the marketing department and Zendesk for their support team.
“It’s really challenging to go in and rip all that out,” he said. “Those tools are functioning and in many cases, companies are locked into multi-year agreements. You need to be able to deploy AI to customers but make sure they’re still getting to use the legacy systems that are there. It’s more about allowing AI to pull information out and push information in where it needs to go.”
Respect the three most common customer segments
Even if a company has yet to deploy AI, Corbin said they are likely serving people who gravitate towards one of three patterns.
The first are the digital natives. These are the people who come to your web site and search for their own information, he said. The second might want to do the same thing, but they need to see some form of validation – either tangible evidence they’re getting the right information or even that they were asking the right question in the first place.
The third group includes those who will want to speak to a human employee. A solid approach to AI delivers an experience that respects all three scenarios, he said.
