https://vladimirpetrishchev.com/blog/Practical notes on building dependable AI systems, data products, and teams inside real organisations.enThu, 10 Sep 2026 08:00:00 GMThttps://vladimirpetrishchev.com/blog/real-time-ai-features-need-write-contracts/https://vladimirpetrishchev.com/blog/real-time-ai-features-need-write-contracts/Thu, 10 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)data productsMLOpsreal-time AIdata governancePartial feature updates make online ML systems faster, but they also turn shared records into a concurrency problem. Define who may write, how time orders changes, and what evidence survives.https://vladimirpetrishchev.com/blog/ai-data-agents-need-an-investigation-contract/https://vladimirpetrishchev.com/blog/ai-data-agents-need-an-investigation-contract/Wed, 09 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsagentic analyticsdata productsanalytics leadershipWhen an AI agent can explore data faster than a team can review it, define the question, evidence, budget, and stopping rule before the first query.https://vladimirpetrishchev.com/blog/ai-agent-controls-need-a-portability-test/https://vladimirpetrishchev.com/blog/ai-agent-controls-need-a-portability-test/Tue, 08 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsAI governanceplatform engineeringAI securityA policy that works in one agent framework is not yet an enterprise control. Test whether decisions, enforcement, and evidence survive a runtime change.https://vladimirpetrishchev.com/blog/ai-throughput-is-not-ai-progress/https://vladimirpetrishchev.com/blog/ai-throughput-is-not-ai-progress/Mon, 07 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)enterprise AIAI operationsanalytics leadershipevaluationAgents can generate code and experiments faster than teams can judge them. Manage review debt before output volume becomes the wrong success metric.https://vladimirpetrishchev.com/blog/an-ai-agents-reasoning-is-not-an-audit-log/https://vladimirpetrishchev.com/blog/an-ai-agents-reasoning-is-not-an-audit-log/Sun, 06 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsobservabilityAI governanceauditabilityReasoning traces can become shorter or less revealing as models improve. Build the audit trail from actions, policy decisions, and receipts instead.https://vladimirpetrishchev.com/blog/your-ai-agent-registry-needs-an-expiry-date/https://vladimirpetrishchev.com/blog/your-ai-agent-registry-needs-an-expiry-date/Sat, 05 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsAI governanceplatform engineeringenterprise AIA searchable agent catalog can spread stale trust. Add admission evidence, review-by dates, and retirement before reuse scales.https://vladimirpetrishchev.com/blog/ai-monitoring-needs-a-data-contract/https://vladimirpetrishchev.com/blog/ai-monitoring-needs-a-data-contract/Fri, 04 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI operationsAI governanceobservabilitysecuritySafety monitoring creates a sensitive data system of its own. Define custody, detection, review, and deletion before production traffic begins.https://vladimirpetrishchev.com/blog/an-ai-agent-handoff-is-a-transaction/https://vladimirpetrishchev.com/blog/an-ai-agent-handoff-is-a-transaction/Thu, 03 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsinteroperabilityoperationsreliabilityWhen one AI agent delegates work to another, reliability depends on explicit acceptance, completion evidence, and a recovery path.https://vladimirpetrishchev.com/blog/treat-ai-evaluations-like-production-systems/https://vladimirpetrishchev.com/blog/treat-ai-evaluations-like-production-systems/Wed, 02 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI evaluationAI agentssecurityoperationsAn agent test can reach beyond the lab. Build evaluation harnesses with verified containment, live tripwires, and incident ownership.https://vladimirpetrishchev.com/blog/your-ai-agent-needs-a-job-description/https://vladimirpetrishchev.com/blog/your-ai-agent-needs-a-job-description/Tue, 01 Sep 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsgovernanceidentityleadershipAI agents need explicit identity, permissions, approval rules, and audit trails. Define their authority before they enter production.https://vladimirpetrishchev.com/blog/your-ai-agent-has-a-meaning-problem/https://vladimirpetrishchev.com/blog/your-ai-agent-has-a-meaning-problem/Fri, 07 Aug 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI agentsdata productssemanticsgovernanceWhen an AI agent gives inconsistent answers, adding more documents can make the real problem worse: the business has not defined what its data means.https://vladimirpetrishchev.com/blog/the-ai-pilot-worked-so-why-did-nobody-use-it/https://vladimirpetrishchev.com/blog/the-ai-pilot-worked-so-why-did-nobody-use-it/Tue, 04 Aug 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI productadoptionoperationsleadershipA good demo proves that AI can produce an answer. Adoption depends on whether the answer fits the work.https://vladimirpetrishchev.com/blog/the-first-question-before-adding-ai/https://vladimirpetrishchev.com/blog/the-first-question-before-adding-ai/Sun, 02 Aug 2026 06:00:00 GMTvladimir.petrishchev@gmail.com (Vladimir Petrishchev)AI productstrategyleadershipBefore choosing a model, ask what decision or workflow should become meaningfully better.