Data governance cannot be an afterthought, though for many organizations, it is.
In a recent episode of So You Think You Can Dev, we sat down with Tom Altman, a fractional CTO with decades of experience, to discuss the increasingly complex world of AI, data governance, and organizational preparedness. As businesses grapple with the rapid evolution of AI technologies, Tom offers a grounded perspective on where organizations stand today and where they need to go.
Understanding Your Own Systems First
Before diving into AI implementation, businesses need to take stock of their existing systems, a step often overlooked in the rush to innovate.
Tom refers to this as systems mapping, a “know thyself” approach. Systems mapping is the process of identifying what technologies are in place, how they interconnect, and how workflows actually happen across departments.
Surprisingly but no uncommonly, many organizations are unclear about their own tech stack, sometimes relying on legacy knowledge hidden within individual employees rather than formal documentation.
As more autonomous AI is developed, organizations must understand their systems to effectively use this technology to automate processes, which requires a granular understanding of specific workflows and how and why they’re structured the way they’re structured.
The Real Barrier to AI Adoption: Foundation
One of the major inhibitors to AI integration is a lack of foundational readiness. Without a clear understanding of workflows, data flows, and existing tools, introducing AI could automate inefficiencies or, worse, create new risks.
“If you’re going to feel confident enough to say, ‘I want this agent to replicate this workflow and this job,’ you have to make sure that all the components of that workflow are set up correctly, or you’re going to potentially exacerbate problems.”
This is especially crucial as agentic AI—AI systems that autonomously complete tasks—become more common. These agents will not only execute tasks but will also traverse existing systems looking for patterns. If the underlying systems are poorly understood, agents might automate flawed or unsecured processes.
Documentation and Governance: The Hidden Levers of Efficiency
Documentation and process visualization aren’t just nice-to-haves, but vital components of your data and AI strategy. Documenting systems exposes technical debt, inefficiencies, and opportunities for automation.
“I like to think in workflows. I can trace it through the process. You can start to see where some things might start to break down.”
While technical debt is often associated with outdated code or infrastructure, unused or redundant software and scattered data permissions are equally problematic. This isn’t just an IT issue, but a larger business one.
Efficiency gains through simple tasks like rightsizing SaaS subscriptions or cleaning up permission creep can be immediate. However, many IT departments are already stretched thin, often sacrificing long-term maintenance in favor of urgent projects. As such, documentation and clear governance can hold the mindshare of these systems, freeing up IT bandwidth and keeping quick efficiency gains at top of mind.
Data Governance in the Age of AI
Many organizations rely on “security by obscurity”—data isn’t secured, it’s just buried.
“What we find, is the co-pilots and the Gemini’s and all these of the world…can traverse your whole network. [They] can check out all these files and get the information you need.”
The risks here are obvious. Sensitive data that was hard for humans to locate becomes trivially accessible to AI. More critically, AI agents might be granted broad permissions without carefully managing access based on user roles.
“We tend to give those bots high degrees of security access because they need to see all the things. But we have to find a way to govern the agent’s security based on the person that’s asking the question.”
Data governance must go beyond technical controls. It needs to be embedded in the culture of the organization, with clear ownership, training, and ongoing management.
Building Toward AI Value
Despite the challenges, AI can bring immense value, if implemented thoughtfully.
Tom advises leaders to identify processes that are simple but high-value, where AI can free up teams from repetitive tasks without introducing complexity or risk.
Here’s what to look for in identifying automation prospects:
“It’s an easy decision. It could be high value because we spend a lot of time on that easy decision. And when you have those two things match, it’s an absolutely great place to look.”
Final Advice to Leaders
Tom leaves listeners with a simple but powerful challenge: be honest about where your organization stands today.
“Are we really in a place where we have our systems mapped? Do we feel like the processes that we have documented already are something that we feel comfortable about?”
If the answer is yes, you’re in a strong position to confidently pursue AI and automation. If not, now might be the perfect time to step back and build the foundation first.
TAGS