TheModernCDO

For fifty years, the rule was simple: garbage in, garbage out. Bad data produced output that obviously looked bad, so someone caught it and fixed it. AI broke that bargain. Feed a copilot flawed data and it returns a fluent, confident, well-structured answer that happens to be wrong, and nothing on the screen looks broken. In this episode, Richard Muirhead makes the case that data quality is the o...

Most AI governance ends at the moment the real risk begins. The day a model goes live is the day it starts to drift. A model learns a snapshot of the world, and the moment it ships that snapshot starts to age. Accuracy and fairness degrade silently, with no alarm, because the governance was all spent at the gate and none on the operation. Watch the model for as long as it runs. Long-form whitepape...

Generative AI is most dangerous at the exact moment it is most convincing. It fails fluently. It produces a confident, polished, authoritative falsehood with no signal that it is wrong, leaks whatever an employee feeds it, and varies its output from one run to the next. Govern the real failure modes, not the ones traditional software taught you to expect. Long-form whitepaper is at https://themode...

A model can be 95 percent accurate and systematically unfair to the people who matter most. The aggregate number hides it. Treated as a late-stage checklist item, bias gets discovered too late to fix, rationalized away under deployment pressure, or missed entirely. A single accuracy number is not a fairness measure. Measure by subgroup, from the start, where you can still act on what you find. Lon...

The moment a real decision can be blamed on a model, you have built a machine for laundering accountability. A model makes a consequential decision and no human owns it. The data scientist points to the business, the business points to the model, the sponsor treats the output as objective. Everyone is adjacent and nobody is accountable, which means nobody is watching when the decisions drift wrong...

When a model starts making bad decisions and you cannot trace what it learned from, you do not have a problem. You have an outage with no off switch. A model runs beautifully for a year, then starts producing bad outcomes nobody can diagnose, because its lineage was never captured. An untraceable model is an unfixable model. Make lineage a deployment gate, the way tests gate code, and capture it a...

Standing up a separate AI governance function is how you rebuild data governance from scratch and get it wrong twice. Most AI risk is data risk with higher stakes: bias from training data, privacy from data use, unexplainability from missing lineage. Wall off a separate tower and you rebuild lineage, quality, and ownership badly, then watch the two versions diverge and open a seam where the worst...

Most companies wrote their AI policy before they had any idea how much AI they were running. Writing the policy is the easy part. The AI it governs is mostly invisible: shadow models, AI features inside SaaS tools, employees pasting company data into public generative tools. You cannot govern what you cannot see. Discovery comes first, and it never stops. Long-form whitepaper is at https://themode...

AI governance that only knows how to say no does not make AI safer. It drives the risky work into the shadows. A brake is not a steering wheel. A function that can only stop makes itself an obstacle, and obstacles get routed around. The risky AI work continues with no oversight at all, while the function congratulates itself on what it blocked. Build the steering wheel instead. Long-form whitepape...

MDM is never finished. It is operated. And the operating budget is the line item every program forgets. Every MDM business case has a detailed build budget and a stewardship line that is tiny or missing. Go-live is not the end. It is the start of the longest and most important phase. The exception queue that needs human judgment does not shrink after launch. It grows. Long-form whitepaper is at ht...

Master data management promises a single version of the truth. But a golden record built from wrong source data is just a single version of a wrong truth, and now it is official. In this contrarian standalone, Richard argues that survivorship rules pick a value, they do not verify one, so a golden record can launder bad data into a clean, trusted, authoritative error. The fix is not in the master...

Every MDM program assumes everyone already agrees what a customer is. That assumption is where it dies. Is one customer a person, a household, a legal entity, an account. The definition lives entirely in the edge cases the kickoff skips, and the buried disagreement erupts deep in the build, when the matching rules force a decision. Decide what one customer is before you try to master it. Long-form...

MDM does not fix bad data. It gives bad data a gold star and removes the warning label. MDM matches and merges. Neither operation creates correct data. Feed it bad inputs and it selects one of them, stamps it golden, and broadcasts it everywhere with more authority than it ever had. Quality comes before matching, or the master just promotes your mess. Long-form whitepaper is at https://themoderncd...

Writing code was the scarce skill in software for fifty years, and vibe coding just made it nearly free. Andrej Karpathy named it in early 2025; by November it was Collins Dictionary's word of the year. Google and Microsoft now say a quarter to a third of their new code is AI-generated. So here's the uncomfortable question for every engineering leader: your developers can produce more code...

The most consequential decision in your MDM program is being made by whoever is configuring the tool. Which source wins when two systems disagree decides whose version of reality is authoritative. Too often an implementer picks a sensible default under deadline pressure. That is a business judgment, attribute by attribute, and it belongs to the domain owner, validated against real records before g...

Without governance underneath it, your MDM hub is an expensive argument that runs on a server. You reached the agreement, configured the hub, and the golden records were trusted at launch. Six months later they have drifted back into disorder, because mastering data is not a thing you finish. The survivorship rules are business judgments that need an owner forever, not just at setup. Long-form whi...

Mastering customer, product, vendor, and asset at the same time is how you master none of them. Each master data domain is its own deep, contentious negotiation, and contentious negotiations do not parallelize. Start them all at once and eighteen months later you have four half-mastered domains, which is to say four nobody trusts. Master one completely first. Long-form whitepaper is at https://the...

Master data management is a business agreement enforced by software. Skip the agreement and the software just automates the fight. The hard part of MDM was never the hub. It was getting the business to agree on what a customer is and which source wins. Automate before you agree and you do not get a single version of the truth. You get an automated version of your disagreement, running at machine s...

Governance built only to pass an audit dies the moment the auditor leaves the building. A program born from an audit gets optimized to survive audits. It passes, the regulator moves on, and it loses its only reason to exist. Use compliance as the trigger. Then build for the value, because the same inventory and lineage the regulator wants are what the business runs on. Long-form whitepaper is at h...

If your data quality work has an end date, you are paying to clean data that will be dirty again by next year. The cleanup crosses its finish line, everyone celebrates, and a year later the data is back where it started, because the sources that made the errors never changed. Quality is not a state you reach. It is an operation you run, with monitoring and a loop back to the source. Long-form whit...

A steward with no time and no authority is a name in a spreadsheet, not a control. You can name forty stewards in a kickoff. It feels like progress. It is a list. A steward needs protected time, real authority to decide, and recognition in how they are evaluated. Take away any one of the three and you have built a costume, not a control. Long-form whitepaper is at https://themoderncdo.gumroad.com/...

A data catalog is not a governance program. It is a filing cabinet for a program you never built. A tool scales a working operating model. It cannot create one. Buy the catalog first, hoping it does the hard work for you, and you get an empty machine harvesting technical metadata into a stale, misleading glossary. Build the program. Then file it. Long-form whitepaper is at https://themoderncdo.gum...

IT cannot tell you what an active customer is. So IT cannot own your governance. Ownership drifts to IT because the catalogs and the tooling live there. But whether someone counts as an active customer is a business judgment, not a technical one. The business owns the what. IT owns the how. Invert that and the catalog is complete and quietly ignored. Long-form whitepaper is at https://themoderncdo...

Governing four domains at once is how you govern none of them. Named stewards everywhere, untrained, with no time. Processes defined in parallel with no proven template. A year later, there is no single place you can point to and say governance works here. Prove the operating model in one high-value domain. Then scale the proof. Long-form whitepaper is at https://themoderncdo.gumroad.com/l/s2e9-th...

If people describe governance as something done to them, they are already building the shadow systems that will gut it. Built as a tax, governance gives the data user nothing they can feel, just forms and approvals, so they route around it. Every shadow spreadsheet and unofficial extract is a vote against your process. Make the right thing the easy thing, or get evaded. Long-form whitepaper is at...