From AI Tools to AI Data Teams: What We Shipped This Spring

Most AI tools make the engineer faster. After our latest product launch, Maia does the work, migrations, builds, data quality, reverse ETL, across the lifecycle, in your architecture, with governance built i. The results CDAOs are reporting: pipeline migrations that took a week now take six minutes, platform consolidations that should have cost two years of engineering effort closed in a quarter, and the backlog clearing faster than headcount can grow.

The Wrong Half of the Job

Across the CDAO conversations we've had over the last twelve months, the same problem comes up in every one. Demand for data is rising. The team isn't growing to keep pace. AI initiatives are accelerating the gap, because every AI use case still needs clean, governed data behind it. S&P Global reported in 2025 that 42% of enterprises abandoned an AI initiative in the previous twelve months, with data readiness as the most-cited reason.

The pattern we see across data organizations is the same, fix the wrong half of the problem first. Make engineers faster at the manual data work in front of them. Coding agents. Copilots. Generative tooling for SQL and Python. Some of this is genuinely useful for individual productivity, but none of it removes the manual data work itself, because the bottleneck has never been how fast a single engineer types. It has always been how much work the team can move through the lifecycle in a quarter, against a backlog of brittle pipelines, repetitive maintenance, and legacy ETL sprawl that grows faster than headcount.

That gap is the question our Spring 2026 launch on May 20th was built around. Yesterday, we announced four new Maia capabilities in a 90-minute live session, including two going generally available. Together, they cross a threshold for the platform. Before this spring, Maia helped a data engineer get more done in a day. After this spring, Maia will become a member of the team, running the work, with your engineers managing execution.

Four Releases. One Direction.

We launched four capabilities: Context Engine, Migration Agents, Skills & Planning, and Mission Control. The temptation is to read this as a feature roundup. It isn't. These are the four legs of a system, and each one has been the missing piece in every other AI data tool we've evaluated.

Context Engine is how Maia learns your architecture. Not metadata in the schema sense. Your team's actual operating standards: naming conventions, architecture preferences, data contracts, security policies. Encoded once, and applied to every pipeline Maia builds going forward. New engineers on board to your standards in hours instead of weeks. Every pipeline Maia produces inherits your standards by default, with no manual review or cleanup.

Migration Agents read your existing Informatica, Alteryx, SSIS, Talend, or Qlik estate, understand the logic, and rebuild it natively in Snowflake, Databricks, Redshift, or BigQuery. They don't translate syntax. They preserve SCD types, CDC patterns, and medallion structure in the conversion. Those are the things that get lost in a manual rewrite and turn a six-week migration into an eight-month one.

Skills and Planning are the answer to two questions that have stopped agentic AI from going into production. Skills means Maia learns your team's patterns over time, so it stops asking the same questions every session. Planning means Maia produces an explicit build graph (sources, transformations, targets, dependencies, tests) before it writes any code. Your engineers approve the logic, then Maia builds.

Mission Control is the operating layer. It runs multiple streams of data engineering work in parallel under a single governance model, with full observability over what every agent is doing, what's queued, and what's gone to review. Migrations, new builds, data quality fixes, Reverse ETL syncs, FinOps optimization: all running concurrently, all visible on a single Kanban board, all moving through review checkpoints before deployment.

Before this spring, you could find one or two of these on the market. Context without orchestration. Code generation without context. Single-threaded agents without governance. After this spring, you can get all four, integrated, from one platform. This represents an inflection point for how data engineering is done.

What "Runs the Work" Looks Like in Production

The opening demo at the launch wasn't a feature tour. It was a Data Engineer (now turned Data Manager) opening Mission Control and watching their agentic data team work the board. The demo walks through Maia issuing an automatic data quality alert with proposed fixes, building a custom connector to data sources by simply passing in API documentation, and a reverse ETL job and legacy workload migration tasks being handled autonomously in the background.

This is what autonomous looks like in production. It isn't unsupervised execution; every change still moves through an approval checkpoint before it ships. The shift is that humans are in the loop where their judgment is the contribution, not where their typing is the bottleneck.

Outcomes You Can't Get by Speeding Up Engineers

The test we've landed on, after watching customers run these capabilities in production, is whether they produce an outcome a faster engineer could not have.

Balfour Beatty had an Informatica migration backlog and a hard compliance deadline. Manually, each pipeline took roughly a week. Mark Hume, their Head of Data, ran the migration with Maia. Per-pipeline time dropped to six minutes. They hit the deadline and eliminated £500K a year in systems integration fees in the process. Hume on the experience: "Maia makes the impossible, possible. We'd almost given up hope. This has given us a new hope that we can shortcut that process."

A faster engineer doesn't migrate a week of work in six minutes. The Balfour Beatty outcome required a system that could read source mappings, understand SCD logic and CDC semantics, and rebuild them natively in Snowflake. That is work an engineer did not have to do.

At St. James's Place, Kelly Maggs (Director of Data Architecture, Platform, and Engineering) and her team used Maia during a platform consolidation. The output: 22,086 engineering hours saved (the equivalent of 11 years of data engineer FTE effort) and $2.2M in immediate cost avoidance. Maggs's summary of the experience was the line you don't expect to hear about AI productivity at this stage in the cycle: "Some of the big productivity numbers you hear with AI can actually be real."

At Sophos, a pipeline task that combined execution, testing, documentation, and workflow updates went from five days to 30 minutes. An 80× change. Their CDAO presented at the launch on what it actually takes to make AI delivery predictable at enterprise scale.

None of these are single-task speedups. They are work that did not have to be done by a human at all. That distinction, work removed, not work accelerated, is what changed in this release.

From Engineer Productivity to Execution Capacity

The metric most data leaders still report up is engineer productivity. Story points completed, tickets closed, velocity per sprint. Those numbers describe how fast humans are doing the work, and they are no longer the right measurement.

The right measurement, after this spring, is execution capacity. How much data work moves through the lifecycle in a quarter, and what proportion of it is autonomous. That number determines whether your AI roadmap ships on time, whether the migration backlog clears before the Informatica contract renews, and whether the business gets a "yes" or a "wait" the next time it asks for a new data product.

The CDAOs we see operating at the new ceiling have already shifted off engineer productivity as their headline metric. The ones still reporting up on story points and velocity in twelve months will be measuring the wrong thing. St. James's Place is already operating there (11 years of FTE effort recovered in a single platform consolidation). Sophos is already operating there (an 80× compression on a workflow that used to define a sprint).

That is the shift this launch is for.

Start With the Recording. Then See It on Your Data.

The four capabilities are available today across Snowflake, Databricks, Redshift, and BigQuery.

The full session, including the Mission Control demo and the Sophos CDAO presentation, is now on our website. 90 minutes.

If you want to see Maia run on your own architecture, our team will set up a walkthrough, bring a workload from your migration backlog and we'll show you the conversion live.

Watch the Spring Launch recording here.