When ALKEME’s head of AI Ryan Deeds (pictured) looked at how artificial intelligence was being sold into brokerages and agencies, he was unconvinced that glossy platforms or large vendors would drive real transformation. Instead, he believed the next shift would come from inside organizations – from producers, account managers and back-office staff who had never considered themselves technologists, but who could move quickly with the right tools and leadership backing.
After more than two decades in the agency world, Deeds joined ALKEME with a clear mandate: redefine what was possible inside a modern insurance organization, and do it in ways tightly aligned to day-to-day business problems, not technology for its own sake.
For Deeds, the starting point was an honest assessment of how fast AI was evolving compared with traditional software cycles.
“AI is going to get better faster than we can conceptualize utilization,” he said. At the same time, agencies could now “build software faster than we can buy, integrate, transform, and get adoption into it.”
That belief shaped ALKEME’s strategy. Rather than waiting on external platforms, the firm moved quickly on internal builds aimed at specific agency pain points, while encouraging teams to become what Deeds called “corporate developers” using AI-assisted tools.
One of those tools was an AI coding agent that Deeds deliberately pushed beyond IT. “It’s typically a development tool, but I’m pushing our accounting teams to use it because it is so good at Excel,” he said. Staff could upload complex financial statements, ask the system to flag inconsistencies and then have it walk them through the underlying logic.
Helping non-technical staff recognize those possibilities was “kind of the big deal,” Deeds said. Once people “started tasting that a little bit” and saw problems being solved in real time, AI stopped feeling like a black box controlled by a third party and became an extension of their own expertise.
Sales enablement became an early proving ground. Deeds focused on tools that helped producers identify the right companies and decision-makers, generate tailored outreach and research prospects more efficiently. “A lot of the tool sets that we’ve built – we’ve built five main products this year internally at ALKEME – solve for that kind of stuff,” he said.
Behind every build sat a broader strategic question. The goal was “really helping everybody understand what is possible today,” Deeds said, and then asking “how much risk are we willing to take as an organization either to leverage that, or how much risk do you incur by not leveraging that?”
That risk lens also shaped how ALKEME evaluated third-party vendors. The company did work with select external providers, but only when the benefit was narrow and clearly defined.
“I’m very reticent to do so today,” Deeds said. “We do use certain third-party vendors if a third-party vendor has a boxed solution that is tried and true and is solving specific tactical issues.”
The sequence mattered. “I’m way more about: what is the business problem that we’re solving for? And then let’s evaluate technology based on that,” he said. What he wanted to avoid was what he described as “shiny ball” syndrome – buying AI first and trying to retrofit it to the business later.
Deeds was particularly cautious about vendors that wanted to train models on ALKEME’s data and monetize it. In his view, agencies that outsourced too much AI capability risked losing critical “muscle memory” that would soon become table stakes, including the ability to interpret policies, dissect coverage and apply large language models to core insurance work.
“I truly do believe it’s as transformative as the internet,” he said. At the same time, he questioned how much of what was being marketed would simply become standard functionality in large language models over the next 10 to 12 months. That uncertainty drove a constant balancing act: how much control ALKEME should retain, how fast it could move internally and where partnerships genuinely made sense.
Keeping AI development in-house raised obvious questions about talent. For Deeds, the answer was not building a traditional, heavily engineered software team.
“I think talent is going to be different. You don’t need technical talent anymore,” he said. As an example, he pointed to his girlfriend, who worked in a prison and “builds applications today because she uses a AI coding agent.” “Is it the best code quality ever? No. Does it solve problems? Yes,” he said.
That speed came with trade-offs. “You incur technical debt quicker than you would if you had a whole production team,” Deeds said. But he was betting on the pace of AI improvement. “Our strategy is AI gets better over time,” he said.
“As a single person developer, I was able to build these five applications that are in production with over 400 of our users today,” he said. “And these are solving real-world tactical problems.”
As internal tools multiplied, codebase management emerged as the next challenge. ALKEME leaned on AI agents and supporting technologies to help refactor and scale its growing stack. “You start to get a code base that is pretty significant, and then how do you manage that well?” Deeds said. “You have AI agents that are helping manage that, you have different technologies that are scaling with that effectively.”
He described current workflows such as “vibe solutioning” as a bridge between old and new development models. Using AI tools 10 to 15 hours a day, he watched problems that were difficult to solve early in 2024 become far simpler later in the year. “Something that’s complicated today [is] super easy in 12 months,” he said.
None of this, Deeds said, would have been possible without leadership willing to tolerate uncertainty.
“By far the most important component [is leadership], right? Because you do have to have some risk tolerance to do this,” he said. One of the reasons he joined ALKEME was CEO Curtis Barton’s appetite for “new things” and his willingness to let teams be “as creative as we can be.”
“I’ve never seen an organization come over the top for a technical group to help them drive this into the culture and make it just a cornerstone,” Deeds said.
Producer Portal offered a clear illustration. “In January, we didn’t have something called Producer Portal,” he said. “Today, Producer Portal is the primary tool that our producers use – either through metrics or through trying to acquire clients or through carrier information. All these different components in this one tool.”
That pace of change reflected a shared philosophy. “If we don’t learn these lessons today and we wait until next week, or next year, or whenever, we’re so far behind the power curve that it’s hard to catch up,” Deeds said.
