The (Still) Big Crisis For Data + AI Teams

Connecting technical execution to business value

Two years ago we wrote about the next big crisis for data teams saying:

“We’ve been building modern data stacks…Yet, much of the time we didn’t know whether or not these tools are actually bringing value to the business…Now when the waiter drops off the check, rather than plopping down the credit card without a second thought, the CFO is investigating each line item.”

Today, some things have changed and some things have remained the same. The CFO is now again eagerly plopping down the credit card as organizations prioritize their AI initiatives, but data + AI teams still need to solve the challenge of connecting technical execution to business value.

Otherwise we risk:

• Siloed teams rowing in different directions. All movement isn’t good movement.
• Burn out because ^
• Tactics becoming the goal. “It’s always been that way!”
• Bad bets on the wrong initiatives

Credibility is the most important currency for an executive, and it’s misspent with every misalignment. As data + AI teams start to apply radically new technology like AI to old problems it will require a leap of faith from all stakeholders. 

Let’s dive into some real ways data + AI leaders have been working to solve this problem.

How To Fix It (Or At Least Get Started)

There are many many possible paths, but I’m going to suggest three ways for leaders to get started based on what I’ve seen work for leading data + AI teams.

Domain Literacy Programs

Many data + AI teams create data literacy programs for their consumers. These help give data power users the foundations they need to self-serve while also understanding both limitations and art-of-the-possible.

What I almost never see are programs that go in the other direction, where everyone on the data + AI team gets briefed on the latest strategic initiatives and goals of a specific domain.

Morningstar doesn’t have a choice but to prioritize domain literacy. Their data and ML models are inseparable from the extremely complex IP they monetize. 

So they invest in knowledge management, prioritize retention, and most of all they clear the decks of any work that isn’t unique to their business.

“When we were rearchitecting our QA practices we took a step back and thought, what are the checks that are industry standard and we can outsource?” said Madison Sargis, head of Analytics at Morningstar. “And the stuff that was really bread and butter core to the methodology, core to the IP, we kept that.”

Enable Self-Service

Speaking of data literacy programs, if you can’t bring business expertise to the data team, sometimes you can bring data expertise to the business team.

For fast-growing M&T Bank, scaling meant investing in governance structures to enable self-service.

“You reach a point where tribal knowledge and deep expertise aren’t enough…” said Andrew Foster, chief data officer, M&T Bank. “…the opportunity was in federation and leveraging our governance model…A CDO should never promise, ‘we will fix your data problems for you.’ That creates a passive culture. Instead, bring the business into the process. The scale comes from business adoption.”

Warner Bros. data insights collection and engineering group (DICE) transformed from a centralized team with direct ownership to an enablement team. As part of this they created a number of collaboration mechanisms from the Data Quality Forum to Data Palooza and Data Days.

These efforts to democratize data + AI ownership are scaling more than ever before because of AI. It has never been easier for a data consumer to operate in English rather than SQL, which again, ultimately allows builders to build.

AI + Documentation

Let me ask you a question: how many people at your organization know the intricate details of every operation across the enterprise? Not many. 

It’s an unfair burden to place on the data + AI team.

The best teams have great mechanisms to make both data and business use cases easy to discover on-demand. For example, one of our marketing technology customers has just about every field, pipeline, and model documented. Business value was evident at a glance.

Unfortunately this is not how most teams operate. They are stretched thin or have not prioritized this level of operational rigor. And while documentation is still a valuable exercise (now more than ever because of AI!), there are some short-cuts.

One of the simplest yet highly productive AI use cases for data teams is the discovery and summarization of downstream assets. In other words, summarize business value on-demand.

AI is great at inferring use cases by drawing correlations from table names, lineage flows and an encyclopedic knowledge of common domain use cases. With just a prompt, I now know the Monte Carlo customer_hub table is critical to customer health scoring, consumption analytics, and customer support workflows.

Our lead analyst Terry Widger was able to dramatically expedite a planned schema migration by leveraging AI for this purpose.

“I didn’t know all 200 tables in this corner of our environment by hand and I was ready to dedicate my afternoon to learning,” said Terry. “Instead the Operations Agent gave me everything I needed to prepare for this migration in ten minutes.”

Focus On What Matters

When teams lack clear signals about what actually matters, they default to what they can control: pipelines, models, checks, and tools. Over time, tactics replace outcomes, alignment erodes, and credibility is quietly burned. 

Data + AI leaders can fix this today by removing work that isn’t core to IP,  by enabling self-service, and most of all by empowering teams with available AI tools. 

Business value isn’t something data + AI teams should be expected to magically “get.” It’s something leaders must intentionally design for.

Our promise: we will show you the product.