My journey with artificial intelligence began at Jam City during an era when AI was still called ML, and most executives weren’t lying awake at night worrying about their AI strategy. As CTO, I oversaw the development and maintenance of our AI-optimized games in a time when this was a bleeding edge application of AI/ML. We approached AI development like software development — a discipline I’d mastered over two decades.
While I’m incredibly proud of what we built, the process was challenging and often painfully slow. Despite our expertise in building and shipping mobile games, applying traditional software development processes to AI projects proved surprisingly ineffective. Part of that challenge was not including game designers and product managers earlier in the process.
Now, as Generative AI advances faster than any technology mankind has ever seen, the reason for those early struggles has become clear:
AI demands a fundamentally different production paradigm than traditional software development.
This realization drove us to create Salt, where our mission is to enable a collaborative process for building AI that hasn’t been necessary — or possible — until now. Because AI development hasn’t previously been a foundational business need, we haven’t yet invented the most effective way to build it. My team and I are committed to changing that.
This shift in development approach isn’t just theoretical — it’s a practical necessity. While traditional software development evolved to build deterministic systems with clear inputs and outputs, AI development requires us to embrace uncertainty and deep collaboration with domain experts. This fundamental difference demands new methodologies, especially as AI becomes central to business strategy.
AI development requires more than sophisticated algorithms and robust computing power. Domain experts bring crucial insights that shape how AI systems understand, interpret, and interact with real-world scenarios. For example, a medical researcher’s deep understanding of disease progression is invaluable in developing AI systems for diagnostic medicine. Similarly, climate scientists’ knowledge of complex environmental systems is essential for creating accurate climate modeling algorithms.
Modern AI development platforms are beginning to recognize this reality. For instance, our work at Salt has shown that when domain experts can actively participate in the AI development process through visual programming interfaces, while still allowing developers to leverage the full power of Python and ML libraries, the resulting solutions are more closely aligned with real-world needs and constraints.
The traditional engineer-centric model of software development, while successful for conventional applications, proves insufficient for AI’s unique challenges. The complexity of modern AI systems demands a highly nuanced understanding of the domains they serve, particularly when addressing multifaceted problems in healthcare, environmental science, and social systems.
Subject matter experts — often with decades of experience in their fields — provide context that raw data alone cannot capture. Their involvement ensures that AI systems:
This expertise becomes particularly crucial when AI systems must operate in highly specialized or regulated environments. For example, in healthcare, understanding clinical workflows and patient care protocols is as important as the underlying machine learning algorithms.
We built the Salt platform to provide an environment where both technical teams and domain experts can effectively collaborate on AI development. By providing visual tools that bridge technical complexity with domain understanding, platforms like Salt enable organizations to unify their team’s effort in one platform.
Effective AI development integrates domain expertise throughout the entire process:
Organizations successfully implementing this collaborative approach typically establish formal frameworks for ongoing dialogue between technical teams and domain experts. This might include regular cross-functional meetings, embedded subject matter experts within AI teams, and structured review processes that incorporate diverse perspectives.
As AI tackles increasingly complex challenges in healthcare, climate science, social systems, and enterprise strategy & execution, the need for cross-disciplinary collaboration grows. Success requires a development approach that values both technical excellence and domain expertise equally. Organizations must create frameworks that facilitate meaningful partnerships between AI developers and field experts, ensuring that AI solutions are not only technically sophisticated but also practical, ethical, and truly impactful.
The future of AI lies not just in advancing its technical capabilities, but in our ability to harness these capabilities through informed collaboration. In the realm of complex global challenges, as well as for-profit enterprise environments, partnership between AI technologists and domain experts becomes not just advantageous but imperative for creating solutions that genuinely serve society’s needs.
At Salt, we’ve observed this evolution firsthand. When domain experts can visualize and interact with AI workflows alongside technical teams, the resulting solutions are notably more robust and practical. The future of AI development platforms must prioritize this collaboration, making sophisticated AI development accessible to cross-functional teams while maintaining the technical depth that complex solutions require.
This collaborative model represents more than an ideal — it is a prerequisite for AI’s continued evolution and its ability to effectively address society’s most pressing challenges, as well as gain a competitive advantage in business. Organizations that embrace this approach position themselves to develop more robust, practical, and ethically sound AI solutions that can make meaningful contributions to their respective fields.
Learn more about Salt AI’s collaboration platform on our website.