AI has, in many ways, reached a fork in the road — like a choose-your-own-adventure book, leaders can either flip to one page or another, depending on the journey they choose.
The quest is a big one: find the next big shift in AI before it happens. Some note that AI can only work well if it’s being fed real-time data. For the biggest AI players, this means solving the question of supplying enough computational power or finding ways to run AI that don’t need as much juice to power it. For other tech companies, the question ahead involves deciding what large language models to use or new methods of data management.
In generative AI alone, two-thirds of companies are increasing investments, according to a recent Deloitte study. In mid-2024, Cisco launched a $1 billion global investment to develop secure and trustworthy AI solutions. Private funding in the United States has put $249 billion into AI research. Numerous big tech companies are funneling millions into supporting AI education and open-source AI efforts.
Regardless of the size or even the AI problem that each one is trying to solve, many tech leaders are in the same predicament — they need research and development teams to scout what’s ahead.
Built In Boston spoke with two AI leaders who are building dedicated time for their teams to research what is coming next in AI. For MacPaw and Hometap, finding and sharing new ideas regarding AI and machine learning is paramount.
“In our company, there is an AI research department that focuses on researching the latest advancements in AI even before they become widely popular,” said Volodymyr Kubytskyi, head of AI at MacPaw.
Kubytskyi noted that having a dedicated research and development team is by far the most effective approach if a company has the means to do so.
Robert Johnson, director of data products at Hometap Equity Partners, tasks individual team members who are experts in data science with keeping the entire team up to date on new developments — integrating research and continuous learning into the entire team culture.
“The rest of our engineering team is also constantly sharing new ideas and tools, and we have strived to keep up with all of these changes by creating a focused but collegial atmosphere,” said Johnson.
The two AI leaders share their strategies and approaches to AI research below.
MacPaw is a macOS and iOS software developer and distributor that builds apps for a variety of purposes, from cybersecurity to file encryption.
How is your team integrating AI and ML into the product development process, and what specific improvements have you seen as a result?
When it comes to the product development process, we at MacPaw actively use a variety of AI/ML tools, such as ChatGPT, Claude, Copilot, Midjourney, DALL-E and others. This depends on the specific tasks and roles within the team.
For example, product managers use ChatGPT and Claude to plan key product development stages and create process descriptions in Confluence. Engineers who write code proactively use Copilot to speed up development and improve quality, while QA specialists use it to write tests, which enhances the overall quality of the code. And designers use Midjourney and DALL-E to quickly generate visualizations of their concepts. While these images are not the final product, they help rapidly demonstrate ideas to the team.
The key improvements are speed and quality. Speed increases because each specialist has an AI assistant that helps with creative processes, from coding to design.
What strategies are you employing to ensure that your systems and processes keep up with the rapid advancements in AI and ML?
Much depends on the size of the organization and its ability to allocate dedicated units for such tasks. In our company, there is an AI research department that focuses on researching the latest advancements in AI even before they become widely popular. Our researchers begin studying relevant technologies as soon as academic papers appear on platforms such as arXiv or after key conferences like NeurIPS or CVPR. The main areas of research include computer-human interaction, productivity, utilities and cybersecurity.
“Our researchers begin studying relevant technologies as soon as academic papers appear on platforms such as arXiv or after key conferences like NeurIPS or CVPR.”
Our researchers explore possible applications of these innovations and their integration into our products and workflows. The key strategy is having a specialized research department. For us, these are the AIR and also technological research and development center departments, which continuously work on implementing the latest technologies. This is the most effective approach if the organization can afford it.
If a dedicated research team isn’t possible, organizations should assign at least one person in the team for investigations. Otherwise, they should split time — 50 percent on core tasks, 50 percent on tech research. This helps keep up, though fully catching up with the rapid pace remains tough.
Can you share some examples of how AI/ML has directly contributed to enhancing your product line or accelerating time to market?
Yes, we have examples of how AI solutions have significantly improved the user experience. For instance, in Setapp — our curated app subscription service for macOS and iOS, which offers a wide range of subscription-based apps — we noticed that users are not specifically searching for apps but rather looking for solutions to their problems. In other words, they need a problem-solving tool, not just an app. When the search is based solely on app names or keywords, the relevance of the results decreases. However, we integrated AI into the search process, allowing it to better understand the user’s intent. As a result, the number of relevant and successful recommendations has significantly increased, improving the overall user experience.
Another example is our malware detection technology, Moonlock Engine, which powers CleanMyMac. AI is now being used in malware detection, and we believe this will greatly enhance the product’s functionality in the future.
Hometap offers solutions that enable people to get more from homeownership so they can get more from life. Its home equity investment product allows homeowners to access the equity in their home as debt-free cash in exchange for a share of their home’s future value.
How is your team integrating AI and ML into the product development process, and what specific improvements have you seen as a result?
AI coding tools, such as Microsoft Copilot, allow us to augment and add leverage to our development process; a data engineer noted that the adoption of AI has expedited our testing and iteration process tenfold from two years ago, prior to wider-scale large language model usage. Other engineers have seen a 10 to15 percent increase in overall productivity in their day-to-day coding. We’ve even seen instances where product managers, who are not full-time programmers, have been able to use LLM tools to create working prototypes during the software development process. For a small company like Hometap, doing more with less is critical as we continue on our growth trajectory.
What strategies are you employing to ensure that your systems and processes keep up with the rapid advancements in AI and ML?
Hometap has a dedicated data organization with data scientists, data engineers and data analysts who keep the rest of the business apprised of changes in this area. This department works on critical roadmap items but is also available to consult with any part of our business that wants data-driven insights — and at Hometap, that’s just about everyone.
“The data department works on critical roadmap items but is also available to consult with any part of our business that wants data-driven insights — and at Hometap, that’s just about everyone.”
The rest of our engineering team is also constantly sharing new ideas and tools, and we have strived to keep up with all of these changes by creating a focused but collegial atmosphere.
Can you share some examples of how AI/ML has directly contributed to enhancing your product line or accelerating time to market?
Staying true to our mission of helping make homeownership less stressful and more accessible, Hometap wants to be able to say “yes” to as many homeowners as possible. Most traditional home equity products look only at one’s FICO score when evaluating creditworthiness; however, Hometap is able to use a proprietary supervised model in our underwriting process that takes a more holistic view of a homeowner’s financial situation in order to paint a more accurate picture and better inform our investment decision.
LLMs have also played a critical role in understanding trends in our business. We use LLMs to analyze all communications we send to our homeowners in order to identify patterns and needs. We’re also beginning to explore how we can use LLMs within other contexts across our business — namely, making sense of all of the data we have at our disposal to improve our operational prowess and provide homeowners with a best-in-class experience in service to our mission.
