AI and Human Decisions in Re-platforming

Richard Suttons 2019 essay, “The bitter lesson” says that in AI development, novel solutions designed to compensate for the limitations of models will become obsolete too soon to create value. So, does this mean that our focus on tooling in support of legacy system re-platforming will be solved by more capable models? Yes, and no.

Yes, models will get better at analyzing codebases, inferring business rules, and implementing technical changes. But re-platforming efficacy isn’t just about code translation - it’s about helping people rapidly comprehend, evaluate, execute and retain accountability for complex decisions that cascade through interconnected systems, processes, and teams.

Re-platforming decisions must remain human-driven because they fundamentally shape an organization’s future: which capabilities to prioritize, what technical debt to address, how to phase changes without disrupting critical operations. Services bridge the gap between AI tools and these human decisions, with judgment remaining essential for balancing competitive needs, risk tolerance, and organizational readiness.

The challenge isn’t compensating for model limitations - it’s ensuring AI outputs reflect technical leaders’ priorities, architectural imperatives, quality expectations and specific organizational needs. Our tooling focuses on providing the right proprietary information about what systems do, how they do it and why, and how people need it to change, enabling increasingly powerful AI models to generate relevant, contextual responses that accelerate transformation while preserving human strategic vision and engineering craft.

Richard Sutton concludes his essay, “(W)e want AI agents that can discover like we can, not which contain what we have discovered.” We’re not trying to encode human knowledge into AI - we’re building interfaces that help humans use AI’s discoveries effectively.

#EnterpriseAI #LegacyModernization #DigitalTransformation