The Mishkin Method
From business ambiguity to production software.
The Mishkin Method is an AI-first approach to designing, building and operating production software. AI participates throughout the lifecycle. Human judgment owns the outcome, architecture and risk.
The six-stage cycle
01 Define the Outcome
Begin with the business outcome, the people involved and the operational reality. AI helps interrogate the requirement, expose assumptions and identify unanswered questions.
02 Iterate Through the Specification
Requirements, workflows, business rules, exceptions, permissions, data relationships and integration boundaries are refined collaboratively with AI as the team learns.
03 Iterate Through Possible Solutions
AI helps explore alternative workflows, architectures, data models and implementation patterns. The objective is not to defend the first answer. It is to reach the right answer faster.
04 Implement at Production Grade
AI accelerates implementation across application logic, data models, permissions, integrations, tests, deployment configuration, documentation and operational controls while established engineering practices remain in force.
05 Validate Against Reality
Working software is reviewed against its specification and the wider operating system. When reality reveals a needed change, return to the appropriate layer, adjust and validate again.
06 Go Live and Continue
Production is an operating milestone, not the end of learning. The same AI-first cycle continues after launch to correct, improve and extend the system.
Rapid does not mean rushed.
Traditional development often separates discovery, specification, design, implementation and testing into long sequential phases. The Mishkin Method compresses that distance. AI helps preserve context as work moves from business problem to specification, from specification to solution and from solution to running software.
The result is a faster development cycle without pretending that complex systems are simple.
The method evolved.
The Mishkin Method began with high-fidelity AI prototypes that pulled validation forward and reduced the requirements gap. AI is now used across specification development, solution design, implementation, testing, documentation, infrastructure, integration and production adjustment.
The workflow is no longer “prototype, then hand off for the real build.” The AI-first process can carry a system all the way into production when paired with disciplined engineering and clear human ownership.
Proven in production
The clearest demonstration of the method is the delivery of an enterprise-class platform over seven months of AI-first development. The system spans multiple code repositories and supports complex operating workflows, self-service experiences, integrations, governance, testing, documentation and production infrastructure.
It was not a disposable prototype reconstructed by another team. The work moved iteratively from business requirements to specifications, from specifications to solution options, and from selected solutions to production implementation. Testing and operational feedback drove further adjustments through go-live.
This is what the Mishkin Method has become: a practical way to use AI across the complete lifecycle of serious software, not as a shortcut around engineering, but as a force multiplier for it.
Guiding principles
- Start with the outcome, not the code.
- Treat the specification as an evolving product.
- Challenge the first solution before committing to it.
- Use AI throughout the lifecycle, not only for generation.
- Build for production from the point the solution is understood.
- Keep humans accountable for decisions, architecture and risk.
- Let working software and real operations drive the next iteration.
Have a software initiative that needs to become real?
Start with the outcome. We can work forward from there.
