The "Unsexy" Strategy: Why Data Governance is the Only AI Strategy That Matters Right Now
Re: Please Stop Buying Robots Until You Clean Your Room
Let’s talk some more about Artificial Intelligence.
It is the shiny new toy in the corporate toy chest. Executives are looking at Generative AI the way Ava, my dog, looks at a laser pointer: with absolute, unbridled, slightly predatory fascination.
We are currently drowning in press releases about how Company X is “leveraging AI” to “synergize paradigm shifts.” And that sounds great! It sounds futuristic! It sounds like something that requires a vest, a headset, and a very expensive subscription to a SaaS platform!
But here is the truth that nobody wants to hear because it is significantly less fun than asking ChatGPT to write a resignation letter in the style of a pirate:
If your data governance is garbage, your AI strategy is just a very expensive way to generate wrong answers faster.
Trying to build a sophisticated AI model on top of your current data infrastructure is like buying a brand new, top-of-the-line Honda Odyssey, the sensible, reliable chariot of the suburbs, and filling the gas tank with chunky salsa.
It doesn’t matter that the Odyssey has heated seats and a 110v outlet for your kids Nintendo Switch, Steve. The car is not going to move. The engine is going to seize. And the entire garage is going to smell like cilantro and regret.
The “Justin” Factor: A Case Study in Hallucinations
We have all heard the phrase “Garbage In, Garbage Out.” But AI turns this into “Garbage In, Confident Hallucination Out.”
Let me give you a hypothetical scenario. Let’s say you have an 11-year-old son. Let’s call him Justin.
Imagine you bring Justin to work one day. To keep him occupied during a backlog grooming meeting, you let him loose in your Workday Sandbox environment. Justin, being an industrious 6th grader (and the son of a Workday developer), creates a new Supervisory Organization called “The Skibidi Toilet Rizz Squad” and assigns himself the Job Profile of “Chief Minecraft Architect” with an executive compensation grade and a salary of $6.7 million (I believe I just completed my Gen Z bingo card - BINGO!).
Now, usually, this is harmless sandbox data. But if you lack Governance—if you don’t scrub your non-production data, or if you accidentally let test data bleed into your analytics pool, the AI sees this.
You ask your fancy new Executive AI Dashboard: “Who are our highest potential future leaders?”
Because the AI lacks context and only knows what it reads, it bypasses your VP of Sales and confidently recommends Justin. Why? Because Justin has a unique title, zero attrition risk, and the highest salary band in the tri-state area. The AI doesn’t know Justin is 11 and refuses to eat anything but Cheetos and chicken tenders. It just knows he is expensive and holds the title “Chief,” so he must be important.
That is what happens without Data Governance.
The Workday Reality Check
Let’s bring this back to your production tenant.
We all want to use the cool new AI features. We want Workday to “infer skills” or “suggest career paths.” But let’s look at the actual state of your tenant, shall we?
1. The Supervisory Organization Disaster
You cannot ask AI to “analyze span of control” if your Supervisory Organizations look like a family tree drawn by a historian who has had three martinis.
If your Sup Orgs are a mess—if you have managers reporting to their own direct reports, or “orphaned” organizations named Department_Old_DO_NOT_USE_v2 floating in the ether—the AI isn’t going to magically fix that hierarchy. It is going to assume that organizational chaos is your actual business strategy. It will look at that structure and think, “Ah, yes, infinite circular reporting. Very avant-garde.”
2. The Job Profile Junkyard
You want AI to match candidates to jobs? Great. But let’s look at your Job Profiles.
Good Governance: Your Job Profiles are clean, consistent, and have competencies and/or required skills attached.
Bad Governance: Your Job Profiles are a mix of PDFs, three sentences written by a hiring manager in 2016, and one profile just named “Manager” that is assigned to 4,000 people ranging from the Shift Supervisor to the VP of Sales.
If you feed that mess into an AI, it won’t give you “Talent Intelligence.” It will give you a digital shrug.
Governance: The Gutter Cleaning of the Tech World
Why do we ignore data governance? Because it is deeply, profoundly unsexy.
Data governance is the architectural equivalent of cleaning the gutters. Nobody wants to clean the gutters. We want to plant the pretty flowers in the front yard. We want the curb appeal. But if you don’t clean the gutters, the roof leaks, the walls rot, and eventually, your pretty flowers are covered in moldy drywall.
Governance is meetings about naming conventions. It is defining who owns the “Cost Center” field. It is the Monica Geller at the party who stops the music to remind everyone to use a coaster.
But here is the kicker: Monica and her coasters are the reason the expensive mahogany table isn’t ruined.
The Department’s Decree
If you are currently looking at a vendor proposal for an AI tool that promises to “fix your data as it learns,” throw that proposal into the nearest body of water.
AI cannot fix a lack of process. AI cannot intuit that when department code 101 means “Finance” in Workday, but 101 means “Cafeteria Services” in the legacy payroll system. Only a human - specifically a human with a very high tolerance for pain and Excel macros - can decide that.
At least at the start, your AI strategy shouldn’t be about Python libraries, AWS, or GPUs. It should be about:
Auditing your Sup Orgs: If the hierarchy isn’t real, the insights aren’t real.
Standardizing Job Families: You can’t analyze talent if every job title is a unique snowflake.
Keeping “Justin” out of the Analytics: Ensure your test data stays in the sandbox, or your succession plan will be led by a 6th grader who plays Roblox.
Clean the house first. Then buy (or build) the robot.