Do Data Roles Have a Future?

Open original

Joe Reis argues data engineering will survive, but the day-to-day tasks that defined the job are being automated, so the work shifts toward interpreting business meaning and owning outcomes. He says roles are blurring into generalists and warns that junior engineers need deliberate paths to build real judgment.

  • AI makes producing code, pipelines, and docs much cheaper, but interpreting business context and taking accountability remain human work.
  • A 2026 survey found 80% of respondents use AI daily, yet only about 10% have built it into organizational workflows, and nearly 89% still struggle with basic data modeling.
  • Reis expects leaner teams and convergence of roles toward "data practitioners," and says companies must create training paths so junior engineers still learn diagnostic skills.

AI is making the mechanical parts of data work cheap. Writing SQL, dbt models, Python pipelines, dashboards, and docs can increasingly be handed to agents, and the author, who co-wrote *Fundamentals of Data Engineering*, says he doesn't know anyone still hand-writing most of it. The value shifts to three things: production (building artifacts), interpretation (figuring out what the data means and why metrics disagree), and accountability (deciding what's right and owning it when it breaks). His example is revenue. An AI can produce clean code for "revenue," but it can't tell you whether the business means gross or net, or recognized or invoiced. That's a sociotechnical question that gets settled by people arguing over definitions, not by better code generation. He cites the 2026 Practical Data Community Survey as evidence of the gap: about 80% of respondents use AI daily, but only around 10% have built it into organizational workflows, and nearly 89% are still struggling with basic data modeling. Tools are moving fast while semantic clarity lags far behind. On roles, he expects data engineers to shift toward system design, reliability, and cost; analytics engineers toward business semantics and data contracts; analysts toward sharper strategic questions; data scientists toward experimental design and causal inference; architects toward cross-environment and agentic workflows; and managers to stay hands-on. He also thinks the lines between these roles keep blurring, pushing people toward "data practitioner" as a catch-all, with specialists in distributed systems, performance, and security as the exception. His main worry is the pipeline of junior engineers. People used to build diagnostic judgment by writing code from scratch and breaking things, and entry-level automation removes much of that practice. His advice to newcomers is to use AI aggressively but read the output closely, understand what's underneath, build real projects, and be able to defend design choices. He expects leaner teams, more fluid roles as org charts flatten, and growing demand for governance, lineage, and semantic consistency as autonomous agents become both producers and consumers of data, with humans still needed to resolve ambiguity and validate what gets generated.

Ask about this article