While AI has gone from an emergent technology to a core part of product strategy for product and development leaders, the complexities of building with AI—and the question of whether to build or fine-tune your own AI capabilties, or to partner with a trusted AI model provider remain top considerations in 2025.
Our 2024 insights report broke down founders’ biggest perceived barriers to integrating with AI: the steep learning curve, figuring out the integration with other tools, the time it takes to customize, and employee bandwidth.
“It’s like trying to race a rocket,” said a respondent. “You don’t compete with machines. You either build a machine or buy a machine.”
Many respondents also cited how difficult it can be to build or continuously customize AI models in-house. “Do a lot of research before deciding to go with your own AI because 90% of the time it’s the wrong decision. You have a huge amount of data to train a model to the standards that these other providers have,” said another respondent.
In our Assembly Required series, AssemblyAI founder and CEO Dylan Fox discussed the similar challenges that many companies face today, and leading AI founders gave their tips on making the right AI decisions for their unique needs.
Check out the entire Assembly Required series hereOn choosing the right AI strategy and finding longterm success
AI founders agree: choosing the right AI strategy that is best for both you as a company and for the longterm success of your customers can be tricky. It can be easy to get starstruck by new AI models and applications when instead, companies need to think about how the right AI can be used to solve the most pressing customer problems.
Jason Boehmig, founder and CEO of AI-powered contract management software company Ironclad, explains that the AI strategy that sounds the most compelling initially is often not the right strategy to pursue. For example, he describes the allure of a custom-trained model or LLM when, in reality, finding an AI partner or vendor is likely the best bet to staying on the cutting edge. But, as he shared, “the market is still playing catch-up to the reality of the results.”
Victor Riparbelli, co-founder and CEO of AI video generation company Synthesia explains that finding the right AI strategy, including deciding whether or not to invest your own resourcing in building a proprietary AI model or buying from a dedicated model provider, requires a lot of trial and error first because many AI models and applications are so new.
Riparbelli continues to describe that, despite making initial mistakes, speed to market has been one of the most important factors for determining AI success for Synthesia. AI moves in “step changes,” he explains, and while you may need to be one of the first to utilize a new technology, you don’t have to be the one to invent it in order to be successful. In fact, the right AI vendors can help companies move faster to capitalize on this AI innovation.
On how partnering with best-in-class AI providers can create better customer outcomes
Other founders describe the unlock they had when they realized that partnering with best-in-class AI providers can actually let their company focus on what it does best: solving customer problems.
Ironclad founder and CEO Jason Boehmig, for example, encourages other AI founders to have a strategic point of view on when it’s best to build from scratch in-house and to only build in-house sparingly. In most cases, he advises, integrating state-of-the-art AI models will be a better investment of resources.
Co-founder and CEO of the AI notetaker platform Fireflies.ai Krish Ramineni describes how his company has found the most success by partnering with best-in-class AI providers. “You realize you have to make these trade-offs,” he explains, citing that building AI models in-house can take too much time away from other areas of focus for your product and business as a whole. The true value of Fireflies.ai, he continues, comes from “how we delight and service our users in the fastest way possible.” Understanding this trade-off, he explains, has allowed Fireflies.ai to rapidly accelerate innovation and revenue growth.
On designing user-friendly, AI-first products that developers love
For many of the AI founders we interviewed, designing products for the developer experience was a key metric for success. Because everything in AI can seem so new, solving the needs of key developers unlocks access to new users and customers, and is a critical part of succeeding in today’s AI landscape.
Edo Liberty, CEO and founder of vector database platform Pinecone, explains how this process played out in the naming their product. Because the AI was in its infancy, Liberty explains how he didn’t even know to call it a “vector database” yet, but that one of their early customers described it that way, and the name stuck—making it easier for developers to find and implement their product.
Liberty continues to describe how Pinecone’s journey in uncovering what captures a developer’s attention led them to three main attributes: authenticity, love, and trust. First, he explains, customers must “love the experience” and think “these people get me” when they use your platform. Second, customers need to trust you as a company. For Pinecone, this means being transparent and open while maintaining a robust, authentic environment for developers to work with.
Founder and CEO of Replicate Ben Firshman describes how developers love building with AI because “the core technology is really magical.” At Replicate, the goal is to capture this magic by breaking down barriers to building with AI and inspiring developers to unlock AI innovation.
