When Kate Grasman (pictured) joined Heffernan Insurance Brokers three and a half years ago as chief information officer, she arrived with a mandate that went beyond maintaining IT systems. Her focus was transformation. “I run all of the IT aspects, from infrastructure to implementing new business solutions,” Grasman said. “But the key change when I came in was really a focus on transformation and innovation.”
That shift was supported with a dedicated innovation budget and a structured experimentation program designed to test emerging technologies quickly. “We do at least six innovations a year, and we are constantly watching technology,” she said. The approach has allowed the brokerage to move early on artificial intelligence, both AI and Agentic AI, to streamline operational processes, and rethink how producers and account managers interact with data.
Grasman’s early AI strategy focused on operational workflows - the repetitive tasks that consume time for producers and account managers. “We adopted ChatGPT over two and a half years ago, which was pretty early,” Grasman said. “We locked it down through getting our own license of ChatGPT.”
From there, the team began embedding AI into specific operational processes. One of the most significant implementations involved an AI platform called Fulcrum, which automates policy checks, proposal generation, and policy comparisons. “We used to do a lot of that work offshore,” Grasman said. “We’ve used this AI solution to save over $400,000. We’ve already quantified it. We’ve already canceled our work with an outsourced provider in India.”
The system now produces proposals in seconds and checks policy documents with high levels of accuracy. “We’ve trained the models with this innovation startup where it’s actually more accurate than a human doing it,” she said. “We’re finding 96 to 98% accuracy for policy checks and policy proposals.” The result is a faster turnaround for frontline teams. “What we’re finding is our account managers and our producers are able to more quickly get the work done,” Grasman said.
While early deployments focused on task automation, Grasman said the next stage involves what she calls agentic AI - systems designed to orchestrate entire workflows while keeping humans involved. "I differentiate AI from agentic AI,” she said. “AI is straight-up tasks that you can make more efficient. Agentic AI is building a workflow that puts humans in the loop of the tasks of the workflow for AI.”
One example currently being developed is a proprietary carrier appetite database with Stitch Studio designed to help producers identify insurers most likely to write specific risks. “There’s no data out there really to collect how to match a person who needs insurance with a carrier,” Grasman said. The system automatically gathers data from internal communications and submissions. “It takes the data from their emails, pulls it, and then puts it into this database to show where there was a successful match for a carrier,” she said.
The database has also been augmented with proposal data and external datasets, with most information entering the system automatically through AI ingestion. “Our account managers and producers almost have to do nothing with it,” Grasman said. “If something looks good, they send an email to it. It automatically ingests it through AI.”Early testing suggests the tool could significantly improve placement success rates. “In our testing, we’ve found about a 50% improvement in success rate,” she said.The project itself began as an experimental initiative. “Our high school interns started that project for me, which is pretty amazing if you think about it,” Grasman said.
The innovation program at Heffernan follows a staged framework designed to surface promising technologies while limiting risk. "At Heffernan, we do what are called tests and learns, and then pilots,” Grasman said. The first stage typically involves just a handful of users. “A test and learn is taking two or three people to help us say, ‘Does this solution even hunt?’” she said. Many ideas fail at that stage. “The ‘bot meets buyer’ solution I mentioned didn’t hunt,” Grasman said, referring to experiments with AI agents on the brokerage’s website. “We found that it was frustrating for our users.”
If a concept shows potential, the firm moves to a broader pilot involving roughly 20 employees from across the organization. “We get different people involved: people who are not as used to using technology, people who are kind of in the middle, and those cutting-edge people who love to use technology,” Grasman said.
An internal AI committee of roughly 25 employees also helps evaluate new tools. Even with strong technology, however, adoption often becomes the biggest challenge. "With our proposals, we have been working on them for two years,” Grasman said. “I thought it was a slam dunk. It’s the most amazing tool I’ve ever seen." Instead, the team discovered an unexpected barrier. “What we found was it was more work for people to do a new proposal from the ground up the first time,” she said. To overcome that hurdle, the brokerage temporarily added staff augmentation to help generate initial proposals until the models were fully trained and workflows stabilized. "The change, adoption, and figuring out why the tech isn’t skyrocketing was critical,” Grasman said.
Despite developing some proprietary tools, Grasman described herself as a “buy” CIO rather than a builder. “I am not into becoming a tech company where we sell insurance,” she said.
Her team scans the market continuously, reviewing dozens of emerging vendors each year. “I would say we evaluate 50 AI companies a year and pilot six per year,” Grasman said. Vendor selection relies heavily on hands-on demonstrations and technical transparency. “If they only have a video of their demo, I’m done,” she said. “They must show me the solution. "Equally important is the team behind technology, I like the people I’m going to co-develop with, and do I think that they have what it takes?” Grasman said. Industry networks also play a role. Grasman regularly collaborates with peers and innovation groups to surface new startups and ideas.
