This document outlines the Data Team's strategic approach to AI adoption in the context of our evolving organizational role.
[TOC]
## Context: The Shifting Data Landscape
*Source: [Data Team Thoughts](https://docs.google.com/document/d/1dTDzt4J3V5U37JWHz9_Pm3E0Or3pmZCnHuu4SWhtHuI/edit?tab=t.0)*
The Data Team is experiencing a fundamental shift in how the organization approaches data. To understand this change, think of the data team as a restaurant:
We are currently set up as a **casual sit-down restaurant** expecting a regular set of customers. We are good at what we do and our regular customers know what to expect from us.
The first shift in the organization's approach to data—new teams asking us for data outside of our normal processes—is like new customers coming to our restaurant treating it as a **to-go or drive-through service**. We have been doing our best to accommodate the requests and the expectations of those making them, but it has been stressful as we are still primarily set up as a sit-down style of service.
What we see coming next—data being requested for new tools such as AI and bespoke reporting—is like a request for **pop-up stands with new menu items** to be out where people are and when people want them.
The path forward requires accepting this shift in direction and adopting the mental state that we have new customers wanting different things, which will position the team for solutions.
## Vision: AI-Assisted to AI-Directed
*Source: [Enterprise Data AI Strategy](https://docs.google.com/document/d/1pi1fXMNzk64LpS6FvZcpLKIICWZiitXONIHByg6q1p0/edit?tab=t.0)*
**Our goal this year: move every role from AI-assisted to AI-directed, while simultaneously shipping AI-powered data products that create value for the broader organization.**
The Enterprise Data team is not starting from zero—we are already building AI-powered data infrastructure. Key capabilities already in production or active development include:
-**Duo Agent Platform / OpenCode**: Taking well-scoped issues from concept to working MR in 10 minutes versus 1 hour manually
-**AI-generated model explanations**: Plain-English explanations of data models that are well-received by stakeholders
-**dbt MCP + Duo**: Full awareness of dbt model structure, lineage, and node details
Our focus is on AI tools as targeted accelerators for well-defined, mechanical tasks—freeing up hours of low-judgment work per engineer per week. The goal is not maximizing the percentage of AI-generated output, but measuring the mechanical time saved.
## AI as an Enabler: Quality and Foundation First
AI adoption offers a path to better serve our stakeholders in this shifting landscape—enabling us to move from sit-down service to pop-up stands, from predictable workflows to responsive, on-demand data delivery.
However, this transformation must be done with **data quality and data foundation at the forefront of our minds**. AI can accelerate our work and expand our reach, but it cannot replace the foundational discipline of well-governed, high-quality data. As we adopt AI-directed workflows, we must ensure that the speed and flexibility AI provides is built on top of solid data infrastructure, clear data governance, and rigorous quality standards.
The team's AI strategy is therefore dual-focused:
1.**Internal adoption**: Using AI to work faster and more flexibly within our roles
2.**External enablement**: Shipping AI-powered data products that create organizational value
Both must be grounded in the data fundamentals that make our work trustworthy and sustainable.
