Blog
Field notes from the build.
Practical writing on getting data and AI into production, from the senior practitioners doing the work. No hype, no boilerplate.
AI AgentsDo You Need RAG? Usually Not for Your First AI Use CaseRetrieval, vector databases and orchestration sound like the serious version. For most first use cases they're upkeep you don't need yet. Here's the question we ask before building any of them.4 min read
AI AgentsThe Four-Line Test: Which Steps in Your AI Workflow Belong in CodeA worksheet you can fill in for one workflow in about fifteen minutes. It shows which steps the model should own, which belong in code, and whether you're ready for production.4 min read
AI AgentsWhy AI Agents Fail in Production: It's Scope, Not the ModelMost agents that break after launch were given too big a job. Here's how we split a workflow so the model reads and code decides, and why that removes most hallucinations by design.5 min read
AI AgentsI built the same agent twice. Here's what Managed Agents changed.Claude Managed Agents and the Agent SDK run the same agent loop. The choice is who runs it, and where your data is allowed to sit.Oct 21, 2026 · 10 min read
AI AgentsThe Burden of ProofPutting an agent on real work takes a clear set of steps and a way to check what it did. What I've learned: the agent's own summary of its work is the one thing you can't check it with.Oct 13, 2026 · 8 min read
Data & AI StrategyHow to Run an Internal AI Readiness Assessment Before Your Sponsor DoesTwo weeks, six seats, one page. The runbook the framework post summarizes.Sep 28, 2026 · 10 min read
AI AgentsJev explained: what a System One model is and where it fits in your stackTypeSafe AI's Jev returns typed, calibrated decisions instead of text. Here's what that means, how to call it, and where it fits in an agent harness.Sep 23, 2026 · 5 min read
Data MigrationsThe Legacy Billing Job Ran One Last Time. The New Pipeline Ran the Same Day.What a parallel run proves that a runtime number can't.Sep 23, 2026 · 5 min read
Data & AI StrategyThe AI Exit Premium: Why Data Infrastructure Is Now a Valuation VariableAI-readiness is becoming an M&A diligence line item. Portcos with instrumented, auditable AI in production are commanding exit premiums because strategic buyers can underwrite the revenue contribution. Ones with ungoverned AI pilots are getting haircut on valuation. Here is what the gap looks like, and what to build during the hold period to close it.Sep 10, 2026 · 7 min read- Data & AI StrategyThe AI Exit Premium: What Buyers Are Really Checking in the Data RoomStrategic buyers are now running structured AI diligence. Portcos that built clean data infrastructure and measurable AI ROI during the hold period are commanding higher exit multiples. This is what the technical review team is looking for when they open the data room.Sep 8, 2026 · 7 min read
Data & AI StrategyHow AI-Ready Is Your Organization? 5 Questions Every Executive Should AskAI readiness is not about picking the right model. It is about whether your data, talent, processes, and governance can support the investment. These five diagnostic questions will help your leadership team find out where you actually stand before you spend anything.Aug 11, 2026 · 4 min read
AI AgentsAgentic Automation Comes for Diligence Before the PortcoThe operating partners we talk to want to use AI to transform their portfolio companies. Automate RevOps. Wire up customer support agents. Build an AI-powered field service routing layer. Get the portco on the right AI stack before the hold period runs out.Aug 10, 2026 · 7 min read
Data & AI StrategyBuilding the Oversight Wrapper: What a PE Operator Actually Has to Ship Before the Portco’s AI Is AuditableOnce you've decided AI oversight is required, the question becomes what to build and in what order. An operational guide to the four components every PE-backed company needs to ship: audit trail architecture, human-in-the-loop gate placement, decision-logging schema, and vendor assessment checklist.Aug 10, 2026 · 8 min read
Data & AI StrategyBuild-vs-Buy for AI Governance Is the Wrong First QuestionPortfolios rushing into AI governance vendor evaluations are solving the second problem first. The decision that drives the vendor is not build or buy. It is centralized or distributed. Get the architecture question right and the vendor decision almost answers itself.Aug 10, 2026 · 8 min read
Data & AI StrategyAI-Ready Isn't a Score: The 3 Operational Signals That Predict Whether AI Compounds or CannibalizesAI readiness is not a technical maturity score. Whether a portfolio company will capture AI upside comes down to 3 measurable operational signals a deal team can read before close. Here is the framework.Aug 7, 2026 · 8 min read
Data ArchitectureData Infrastructure Is the AI Moat: Why PE Portfolio AI Fails at the Data Layer, Not the Model LayerMost PE portfolio AI initiatives stall before production, and the post-mortem rarely blames the model. Gartner projects 60% of AI projects unsupported by AI-ready data will be abandoned by end of 2026. Clean, governed data infrastructure compounds AI value; bolt-on AI on top of ungoverned data does the opposite.Aug 6, 2026 · 7 min read
Data & AI StrategyYour AI Pilot Didn’t Fail Because of the Model: The Data Infrastructure That Has to Come FirstMost mid-market AI initiatives fail not because of the model but because of what feeds it. Here are the four data infrastructure preconditions that determine whether any AI initiative succeeds.Aug 5, 2026 · 7 min read
Data & AI StrategyYour portfolio company's data stack is a valuation problem, not an IT problemPE operators who treat data infrastructure as an IT overhead line are miscategorizing the problem. Data maturity is now an explicit diligence criterion, and the operating partners who get ahead of it during the hold period close at a premium.Aug 1, 2026 · 8 min read
Data ArchitectureThe Property Data Problem Every CRE Platform Eventually HitsEvery commercial real estate platform hits the same ceiling eventually. You build solid integrations for the first few data sources. Lease abstracts, rent rolls, market comps flow cleanly into your analytics layer. Then someone asks for building sensor data from a new IoT vendor, or the team wants to pull in a third-party brokerage feed, or a large tenant requests integration with their own facilities management system. The answer, more often than not, is: that is a three-month project.Jul 31, 2026 · 24 min read
Data & AI StrategyAI Governance for PE Portfolio Companies: What Actually WorksEnterprise AI governance frameworks break under PE portfolio conditions. Here are the three controls that actually work, and the 90-day sequence to install them when you are starting from zero.Jul 29, 2026 · 6 min read
Data & AI StrategyBefore You Buy an AI Model, Score the CompanyMost PE-backed companies buy the AI model first and discover the readiness gaps second. A structured assessment inverts that sequence before you write the check.Jul 25, 2026 · 3 min read
Data Engineering5 Signs Your Data Stack Isn't Ready for AIAI pilots rarely fail because of the model. They fail because the data foundation underneath it is not ready. Here are five signals that tell you where you actually stand before you invest.Jul 25, 2026 · 6 min read
Data & AI StrategyYou Scored Your Portfolio Company's AI Readiness. Now What? An Operator's Playbook for Closing the GapsThe assessment is done. The score is real. Now what?Jul 25, 2026 · 10 min read
Data & AI Strategy5 Questions Your Board Is Asking About AI (And What a Real Answer Looks Like)Boards are asking about AI strategy, peer benchmarks, investment priorities, risk, and ROI. Here is what a credible answer looks like for each, from someone who has sat on both sides of that table.Jul 21, 2026 · 5 min read
Newsletter
Past the Pilot
Next step
Let's find where AI pays.
Read-only, with an engineer who has shipped this. You leave knowing what level you're at and what's worth funding.