AI delivery
By Andrew Korolov
Workflow or agent? Choose the least autonomy that works
How to decide whether a fixed AI workflow is enough or an agent should choose the next step.
Read the articleField Notes
What it takes to build production AI.
Articles and opinions from MavenSolutions leadership on choosing worthwhile problems, testing what works, and operating AI after launch.
Featured articles
Business case
By Andrew Korolov
Measure the workflow before you automate it
A practical baseline for deciding whether an AI project changed cost, speed, quality, or capacity.
Read the articleAI evaluation
By Andrew Korolov
Build the evaluation set before the prototype
Why historical cases and explicit pass criteria should define the system before prompts and tools do.
Read the articleAI operations
By Andrew Korolov
Production AI needs a runbook, not just a dashboard
What operators need to detect failures, reproduce them, recover safely, and know when to escalate.
Read the article
Production AI
By Andrew Korolov
Read the articleYour AI pilot works. Why hasn’t it shipped?
A diagnostic for the ownership, integration, reliability, and adoption gaps between a strong demo and production.
AI strategy
By Andrew Korolov
Read the articleBuild, buy, rescue, or stop?
A decision framework for choosing the next move when one costly workflow looks ready for AI.