Your Pipelines Got Faster. Your Checks Did Not.

Two numbers from the same survey sit a page apart and tell opposite stories. 72 percent of data teams say they are prioritising AI-assisted coding in their development workflow. Only 24 percent say they are prioritising AI-assisted pipeline management, which covers testing, observability and everything else that tells you when a pipeline broke.

Both figures come from the State of Analytics Engineering report that dbt Labs published in April 2026, based on 363 data professionals surveyed between late 2025 and early 2026.

So the writing got faster and the checking did not. You can probably guess what happens next.

The same report found that 71 percent of data professionals named incorrect or hallucinated outputs reaching stakeholders as a top concern. That worry is rational. When a team triples the number of models it ships in a quarter without touching its test coverage, the failure rate per model stays flat while the absolute number of failures climbs.

Trust Went up, Which Is the Strange Part

Here is the finding that complicates the story. Trust in data jumped from 66 percent to 83 percent year on year, the largest single-year rise the report has recorded, and speed as a stated priority climbed from 50 percent to 71 percent over the same period.

What do you get when you read those together? An organisation that is moving faster, feels better about its data, and has fewer people watching the pipes. That combination has a shelf life.

Ownership Is the Quiet Failure

41 percent of organisations in the survey reported ambiguous data ownership. This is the one that rarely appears in a postmortem but sits underneath most of them.

Ask a plain question inside your own company. When the revenue dashboard is wrong on a Monday morning, who is expected to notice, and who is expected to fix it? If the honest answer is "whoever gets pinged first", you do not have an ownership model. You have a rota that happens to hold while nothing serious breaks.

Ambiguity here is expensive because it delays detection, not because it delays repair. Repair is usually quick. Detection is what takes three days.

The Budget Squeeze Is Real

One more pair of numbers from the report. 57 percent of respondents reported increased warehouse and compute spending, while only 36 percent reported increased team budgets. That means more data, more models and more compute handled by roughly the same number of engineers.

This is the pressure that has pushed a lot of organisations toward external Data Engineering Services for the unglamorous layer, meaning contracts and tests on critical tables, freshness monitoring, lineage a non-engineer can actually read, and ownership records that survive someone changing jobs.

What to Do This Quarter

Start with your five most-viewed dashboards and trace every table that feeds them. Put a test on each one. Nothing elaborate, just a not-null check, a row-count range and a freshness threshold.

Then name an owner for each of those tables. Name a person, not a team.

Lastly, measure how long it takes you to notice a break rather than how long it takes to fix one. Most teams have never measured the first number and are quietly shocked when they finally do.

None of this is exciting work. It does not demo well and it will never earn a slide in the quarterly review. But it is the difference between a data team that gets faster and a data team that only gets louder.

May your freshness checks fire before your CFO does.

About Author

Shefali Vasave

Shefali manages content marketing at Opus Technologies, a domain-native engineering partner for banks, payment providers, and fintechs, and writes on the various aspects of financial institutions navigating change in a real-time, digital-first world.