The human spark behind Dovetail's AI-first approach

For our team, it's a blend of inherent curiosity, a love for complex problems, and the thrill of pushing boundaries.

Chakaveh discovered her passion early. “It goes back many years,” she recounts, remembering an optional course in artificial intelligence during her computer hardware bachelor's degree. “I fell in love with AI.” This initial spark led her through Master's and PhD degrees specializing in natural language processing. The recent explosion of large language models has been a natural progression for her expertise, positioning her perfectly to help build Dovetail’s customer intelligence platform.

Rod began his university studies in structural engineering, pivoting later in his degree to software engineering before beginning his career in product development. His journey into AI was self-directed and opportunistic. “It started with a desire for financial freedom. With my strengths in math and programming, and the need to invest my money intelligently, it made sense to try and write an algorithm to allocate my money intelligently.“ His venture into algorithmic trading, using image recognition technology to predict stock movements, predates the widespread buzz around multimodal models like ChatGPT and Gemini but helped to set him on a new course of specialization.

Adel transitioned to AI from within Dovetail itself. “I didn't start here as an AI engineer, I was hired as a software engineer,” he shares. After nearly a year working on different software teams, Adel's personal interest in AI—which included tinkering with tools like ChatGPT and Claude outside of work—converged with a new opportunity. “I went to my previous manager... and asked him, 'Hey, is it possible to get a trial working within the AI team? The AI team was only one other person then, but it was my foot in the door.” This trial, and some good timing, led him to become a core member of the burgeoning AI team. His experience highlights that a background in AI isn't always a prerequisite; a keen interest and a willingness to learn will not only open the door but ensure you stay on top of rapidly evolving tech.

Peter, who manages the AI team, found his way back to AI after a non-traditional start. His initial interest was sparked in high school, where he experimented with building neural nets to simulate fish in a tank, and later, experimented with algorithmic trading. At university, however, he studied both Computer Science and Law, pursuing a career in software engineering rather than AI. He believes this legal background now gives him an edge. “I view prompt and context engineering fundamentally as a communication challenge,” he notes. After the LLM boom happened, he saw a critical need within Dovetail. “For Dovetail to evolve into an AI-native leader, we needed a team whose top priority is AI quality and innovation, and drives that throughout the rest of the org,” Peter explains. And that’s how Dovetail’s AI team got started.

A conscious journey to AI-First

Dovetail's commitment to becoming an AI-first customer intelligence platform is a deliberate and ongoing journey that has led to a necessary mindset shift across the entire organization.

Chakaveh notes this transformation: “Dovetail, by nature, was a software engineering operation, but we needed to consciously move in the direction of AI-first and define our engineering problems in a different way.” This conscious effort involves educating other teams, a process she sees as rapidly picking up steam. “I see and hear people using terms that are specifically AI-related all the time across our product org. It’s becoming general knowledge, not specialized.”

Adel believes that being AI-first boils down to two core principles: being data-driven and evaluation-driven.“Now that we're getting thousands and thousands of data points, we can look at that mountain of agentic chat threads that we have in our database and make more informed decisions about how the product takes shape and what we need to build next.” For every feature built, rigorous testing against vast datasets ensures quality and identifies regressions.

Rod further elaborates on the philosophical shift in quality assurance. Unlike traditional software where “1 plus 1 equals 2” every time, AI deals in probabilities. “The difference between being an AI-first company and a traditional software company is the quality bar, the specification expectations around quality are more strictly defined." He established an evaluation framework that runs 200 fictitious conversations through the codebase, measuring performance against eight metrics. “If those measurements are worse than what was there originally, it can be used as an argument to stop a feature or stop a deployment,” ensuring the AI pipeline maintains its integrity.

Peter sees this transformation as a fundamental rethinking of product development, starting with the very definition of quality. “Quality is no longer just what you visually see—it’s not about being pixel-perfect. The focus is now on the content: Is this AI output accurate, reliable, and trustworthy?” This shift changes the finish line for every project. A perfect example is our contextual chat feature; while the UI was spot on, we’ve kept it in beta to do the deep R&D needed to nail the context engineering and ensure it avoids hallucinations. It even changes how we test our own features. “Effective dogfooding for AI means you have to use it as a customer would, with real-world inputs,” Peter adds. “You can’t just use dummy data anymore.” Ultimately, this new standard, combined with transparency about limitations and a commitment to rapid iteration, is how trust is built.

The embedded model: rapidly integrating AI across the product organization

A cornerstone of Dovetail's AI-first strategy is its embedded team model. Instead of the AI team operating in a silo, individual AI engineers are integrated directly into every product team ensuring that AI considerations are present from the earliest stages of product discussions, to design and roadmap planning, and that everyone benefits from mentorship and knowledge sharing.

“This definitely helps with the education and problem-solving aspects. It help us to influence decision-making and guide things in a certain direction using AI-thinking,’” says Chakaveh.

Adel highlights how this model promotes collective ownership of the AI platform. “We started with just the AI team working on the AI platform, but after we made the switch to being embedded and supporting the infrastructure that supports this product, it's become a collaborative process across Engineering.”

Peter agrees, emphasizing that this collaborative model means that the foundational AI infrastructure is built with real-world product needs in mind, streamlining development across the board. “The embedded model is how we diffuse best practices and set the quality bar high across all teams consistently,” he says. “But the most important part is that it empowers every team to take that ownership. For us to deliver AI features that customers truly love, we need to throw the entire weight of the product organization behind AI innovation, and this model makes that possible.”

“Communication is one of the most important factors, sharing new knowledge, new learnings,” says Chakaveh about how they stay connected despite working on different projects. Outside of product team standups, the team holds weekly meetings, shares documentation, and hosts brainstorming sessions to work through challenges, ensuring that knowledge transfer takes place and to help prevent duplicated efforts. This constant exchange also helps everyone stay abreast of rapidly changing AI developments and ensures that individual learnings benefit the collective.

New innovations, same customer-centric DNA

Dovetail’s shared value of putting the customer first in product development is a powerful motivator. Adel was involved in creating magic insights, now called AI docs—Dovetail's first agentic product, and one he’s particularly proud of. “The reason I was bullish on that idea was that when you create an insight there are millions of things that customers want from it. And you can't really hardcode these intents, so why not just let the LLM actually decide what the research path is?”

For Rod, it was feedback from a customer call regarding the trustworthiness of AI outputs that reinforced the importance of deep-linked citations, allowing users to trace AI-generated information back to its source. The customer felt that a chat agent is useless unless you can believe that what it's saying isn't a hallucination and that means referencing back to specific data sets to build context and trust.

Chakaveh thinks the biggest benefit is being able to more efficiently use her time. “One big benefit of using AI is automation. What might have been hours checking and rechecking the data, can now be done in one-third of the time. I’ll caveat that we shouldn't ignore that AI is not reaching human accuracy - I still need to check the work. But no doubt, it saves me time and frees room for me to do something AI can’t, like putting the customer first, one of our core values.”

Peter believes that while data and metrics are vital for measuring progress, a level of empathy that comes from direct conversation is required to uncover crucial blind spots. “I talk to dozens of customers, and by paying attention to what they’re saying about hallucinations or reliability, I’ve developed a gut feeling for where the real quality gaps are—you feel it on an emotional level,” he explains. “You don’t always know what to measure upfront with a non-deterministic technology like AI. That’s why you have to be completely obsessed with your users. We need to own the entire quality experience, and that comes from staying incredibly close to them.”

Thriving on the bleeding edge: keeping up with AI

The pace of innovation in AI is relentless. Our AI engineers embrace this challenge by staying fiercely up-to-date.

Adel emphasizes the importance of dedicating time daily to keep pace with new models and technologies. Beyond listening to AI news every day, Adel advocates for hands-on experience: “How you get 80 percent there is to do your own side hustle projects and allow yourself to fail over and over again to increase the pace of learning.” This experimental mindset is key to rapid skill development.

Chakaveh reinforces, referencing the proliferation of content about new AI technologies, she notes the shift from deep dives into single sources to needing to “just skim through, scan through the titles” of multiple blogs to discern what's most relevant to keep on top of it. The team has created a ritual to help with this, leading a bi-weekly AI reading club open to all engineers.

The future is collaborative: humans and AI

What does the future hold for AI, and for Dovetail's role within it? Our engineers believe AI will amplify human capabilities, making customer intelligence more efficient and insightful.

“No matter how sophisticated AI gets, human-in-the-loop collaboration will always be fundamental to Dovetail’s product experience,” Peter says. He believes the core principles of any trusted collaboration—the need for transparency and the ability to verify outcomes—are universal. He envisions AI graduating from smaller tasks like summarization and synthesis to taking on more responsibility as agents integrated into a business’s core processes. “The industry is moving toward AI that is more powerful and autonomous,” Peter concludes. “Our commitment is to ensure that as AI becomes more powerful, it remains a partner that you can collaborate with and, most importantly, trust.”

At Dovetail, we're not just building AI; we're building a team and a culture that is ready to embrace the vast potential of artificial intelligence to redefine how businesses understand their customers. Want to be part of Dovetail's journey? Check out our careers page to learn more about opportunities to join our innovative team!