Ask a Data Ethicist: How Should We Deploy AI in the Enterprise?

As Master of Ceremonies for the 12th annual DAMA Edmonton conference, I literally had a front row seat to many excellent presentations, questions, and observations about how AI is actually getting rolled out right now in a wide range of organizations. I thought I’d share a few stories from the field under this question…

How should we deploy AI in the enterprise? 

Also, I’m pleased to say I just earned my CDMP credential at the post-conference workshop. I’ll spend a little time unpacking that journey too.

You get a bot, and you get a bot, and you get a…

The most prominent way AI is marketed is as a trusty assistant that anyone can use. You can chat and vibe code your way to greater productivity, enhancing your outputs in meaningful ways. Maybe you’ll build a team of AI agents who will do all the hard work, while you “hang” with another AI avatar created as a “pal” to goof off with. One of the presenters – Stephane Contre – has built his solo-preneur consultancy, Storm Analytics, this way, complete with goof-off AI buddy Kevin. After a long career in management roles, Stephane is very happy not having to manage people, and as someone who is also an intentional solopreneur, I do see the appeal of this, though I’m not sure I see the appeal of having an AI “pal.” The audience also picked up on this, asking all the ethical questions I wanted to ask because it’s 2026 – and people, rightfully, have concerns. 

This type of agentic AI approach might be fine, even liberating, for a solopreneur because the workflow is simple. But, in an enterprise, the “bots for all” approach can create a series of non-scalable workflows, a message the folks from the Alberta Motor Association pointed out. It also introduces an inordinate amount of risk, as there can be a lack of appropriate tooling in order to help broaden the processes while respecting concerns for privacy, security, consistency, etc. 

And, as another presenter, MNP’s Caitlin Crowley, pointed out that “bots for all” isn’t really a strategy. Having a backlog of use cases is also not a strategy. Instead, she asks companies to define how AI is actually creating value. If you are simply doling out bots like Oprah gives away cars, you are likely spending a lot of money on licensing without clearly articulating any sort of value. Saved time does not necessarily translate into value because as humans, we’re pretty good at wasting time. More meetings, more emails, more scrolling. 

AI in the Hands of IT

A lunchtime conversation unpacked this question with a different take: Maybe AI is best deployed via IT first, and not directly to users. An attendee from a not-for-profit shared that they had built the necessary tooling and safety protocols around a suite of customized AI tools for their organization. This allowed their end users access to practical AI tools – not generic ones – without having to become prompt engineering experts. They were delivering demonstrated value to their organization, and it was much more cost-effective than a “bots for all” approach. An AI vendor representative, also seated at our table, agreed this made sense, but also acknowledged it might not be as economically fruitful as a sales strategy for AI vendors. Hence, “bots for all” gets pushed so much by the industry. I won’t share names of these folks because this wasn’t a presentation, just a table conversation. 

Yet, this message was also echoed by other presenters, including Tantus Solutions’ Jason Suriano, whose wide-ranging talk on shadow AI sparked a lot of audience engagement. He questioned who really benefits from time saved – something I’ve blogged about too. If you tell your employer you’re saving four hours a day, your reward is just more work, so best not to share that! Hence the “shadow use.” He also envisions a future line of business helping rip out failed AI projects – something I likened to everyone getting a little too much ink and then having regrets. 

Back to our friends at AMA, who shared a detailed description of building Teddy, their cutely named, enterprise AI rollout. Giovanni Mera and Mack Mackena laid out the details of building their solution using Amazon Bedrock Coreagent. Nods around the room followed their talk, along with a plethora of technical how-to questions. Their talk reminded me of the story about JP Morgan, who famously banned use of generic chatbots but got behind their own LLM Suite for its 200,000 employees. Mackena related the AMA approach to workflows, and the need to ensure scalability of workflows vs individual chats stuck in individual bot usage. At the same, time AMA has also allowed employees to develop their own customizations, aiming for the middle ground of sorts in terms of user experience and agency. 

Their story exemplifies AI in the hand of IT (aka technical builders). We don’t all need to become technical builders or vibe coders. Perhaps all we need are more thoughtful and strategic ways of deploying AI along with a super solid rationale (aka strategy) for how we plan to drive value.

Bonus: My CDMP Journey

How did I recover from two full on days of MCing a conference? By writing my CDMP exam. Each year, our DAMA chapter offers a “pay if you pass” proctored exam following the conference. I decided this would be the year I made it happen. I had taken part in various CDMP study groups before but I really needed a tune-up to refresh my memory. Here are some tips from my journey:

• Don’t leave too much time between learning the material and taking the exam.
• Consider taking a course. I’m a reader but I’ll admit – the book can feel dry. DATAVERSITY offers many great courses led by engaging instructors – lookin’ at you, Mark Horseman!
• Make a public commitment. Mine was made to 200+ conference attendees – I told them I would write the exam, so I had to follow through. Your commitment doesn’t have to be that public, but tell some folks – it really does help.
• Practice, practice, practice. 
• Have a strong “why” – my reason for wanting to learn the material was wanting to increase my practical data skills so I could have better conversations about ethical matters with data practitioners. Being CDMP certified gives me another level of proficiency and credibility in my role as a data and AI ethicist.

Send Me Your Questions!

I would love to hear about your data dilemmas or AI ethics questions and quandaries. You can send me a note at [email protected] or connect with me on LinkedIn. I will keep all inquiries confidential and remove any potentially sensitive information – so please feel free to keep things high level and anonymous as well. 

This column is not legal advice. The information provided is strictly for educational purposes. AI and data regulation is an evolving area and anyone with specific questions should seek advice from a legal professional.