dbt Labs released its fourth annual State of Analytics Engineering report on April 14, and the headline finding is one that anyone running a data team should sit with for a moment: AI is making data work faster, but it is making it harder to trust at the same time.

The report surveyed 363 data practitioners and leaders across industries and regions, with responses collected in late 2025 and early 2026. The findings do not land as a warning against AI adoption. They land as a warning about what happens when you adopt AI without building the governance layer alongside it.

Speed Is Up, Trust Is Not

The most telling statistic in the report is this one: 72% of data professionals now prioritize AI-assisted coding in their development workflows. Only 24% prioritize AI-assisted pipeline management, which covers testing, observability, and the processes that catch problems before they reach stakeholders.

In other words, the tools that help you build faster have taken off. The tools that help you catch failures have not.

The consequence of that imbalance shows up elsewhere in the data. Seventy-one percent of respondents cite incorrect or hallucinated AI outputs reaching stakeholders as a top concern. That is not a theoretical worry. As autonomous agents operate on top of organizational data at scale, the cost of a bad output silently flowing through a pipeline and landing in a board presentation or a customer report becomes significant.

Trust Has Become the Top Priority

To their credit, data teams seem to know this is a problem. Trust in data has jumped to the most widely prioritized organizational objective, named by 83% of respondents, up from 66% in 2025. That is a 17-point jump in a single year.

The priority of shipping data products faster also rose, from 50% to 71%. So teams want to go fast and they want outputs to be trustworthy. The hard part is that right now, the tools and practices for one are much further ahead than the tools and practices for the other.

The Budget Problem Makes It Worse

There is a structural problem underneath the governance gap. Fifty-seven percent of data teams report increased warehouse and compute costs, but only 36% report increased team budgets. The cost of running AI-powered data infrastructure is outpacing the resources available to manage it properly.

This creates a squeeze. Teams are spending more on the infrastructure that generates outputs, and they have less left over to invest in the people and processes that validate those outputs. That is the wrong trade-off order.

What This Means for Business

If you are a business leader who relies on a data team to inform decisions, this report is a useful prompt to ask a few questions.

When your data team ships AI-assisted work, what is the process for validating it? Not at a high level, but specifically: who checks it, how, and at what point in the workflow?

When your analytics dashboards or reports are built with AI assistance, is there a paper trail that lets someone trace an unusual number back to where it came from?

These are not questions that require slowing down AI adoption. They are questions that require building the right foundations alongside it. The organizations that get this right will be able to move fast because they have earned the trust to do so. The ones that skip the governance work will eventually spend more time untangling data quality problems than they saved by speeding up.

The Bigger Pattern

The dbt Labs report is one data point in a pattern that is showing up across the industry right now. Stanford’s AI Index, released this week, found the same dynamic at the macro level: AI capabilities are hitting historic highs while responsible AI practices and transparency metrics are falling.

The pattern is not a reason to stop. It is a reason to build deliberately.

Data literacy, the ability for people across an organization to understand, question, and interpret data and AI outputs, is not a nice-to-have skill in this environment. It is a risk management tool. Teams that understand how their data pipelines work, where AI is involved, and what the failure modes look like are the ones that will catch problems before they become expensive.

---

Enterprise DNA’s learning platform is built for exactly this kind of moment. Whether your team is just starting with data skills or leveling up to work alongside AI systems, our courses and learning paths are designed to close the gap between speed and trust.