SEARCH
AI adoption is accelerating, yet as many as 95% of enterprise pilots fail, per a recent MIT study. The pilots that succeed tend to embed AI into existing workflows, focus on automatable use cases, and leverage commercial, off-the-shelf (COTS) products.
What we see every day in our work is that AI success starts with strong data practices. Too often, organizations treat new tools as a shortcut for deeper gaps in how they manage and use data. We’ve worked with teams that invest heavily in advanced analytics platforms, only to find that adoption stalls because the underlying data is untrusted. When data consumers don’t believe the data, they simply don’t use it—no matter how powerful the tool sitting on top of it may be.
That’s why the difference between an AI pilot that scales and one that fizzles is rarely the tool itself. Often, the differentiator between an AI pilot that scales and one that fizzles out is the maturity of an organization’s data governance and data quality (DQ). Fortunately, a new generation of AI-enabled DQ platforms, many backed by leading venture capital firms, is helping organizations close the data maturity gap.
This article outlines a framework for preparing enterprise data operations and selecting the right AI-enabled DQ tools.
Get Your “Data House” In Order
Across industries, fragmented governance frameworks and poor data quality prevent enterprises from building a single source of truth, which can stall AI initiatives before they have a chance to succeed. Around 90% of data professionals report difficulty scaling data, citing governance and compliance as the primary barriers.
Leaders that prioritize governance initiatives. set themselves up for success. At CarMax, for example, CTO Shamim Mohammad built data governance into the business fabric. He focused on three essentials:
On the other side of the spectrum, VC investors see the same imperative. In a recent Technoventure podcast (a spin-off of Technovation), Greylock’s Asheem Chandna spoke to the future of data infrastructure and the invaluable role of DQ in enabling scalable AI. Chandna foresees a future in which agentic-enabled workflows are redefined via a unified data platform. He believes that the enabling technologies to reach these technologies lie in the next generation of software infrastructure, namely novel data platforms and data observability technologies.
Ultimately, organizations that treat governance and DQ as integral to both people and process, not just technology, are the ones that unlock durable value from AI.
We explore three guiding questions that answer how to address enterprise issues across processes, people, and technology:
When: Readiness to Adopt
Organizations can use the following frameworks (Figures 1 and 2) to assess data readiness before integrating data quality tools.
Figure 1 depicts the data governance value chain, showing the flow of governance activities across people, process and technology. It’s important to evaluate where in the chain AI tools can deliver the most value.
Figure 2 illustrates the data governance operating model, including key roles and reporting structures. Defining this structure and ways of working is a key step toward giving governance efforts real traction and turning potential bottlenecks into managed workflows.
Figure 1: The Governance Value Chain
Figure 2: Data Governance Operating Model and Roles
Choosing when to leverage new AI-enabled tools is a matter of readiness. Foundational skills ensure governance and data quality tools don’t just exist on paper but deliver measurable impact. Analysts and engineers provide the technical baseline through SQL expertise; stewards and architects bring accountability through ownership, metadata, and lineage practices; and business and technology teams create alignment by agreeing on shared definitions and standards.
Readiness also shows up in the ability to measure success against the right key performance indicators (KPIs). Teams must not only fulfill requests quickly but also track whether they are solved correctly the first time. They should gauge user satisfaction with the quality and timeliness of data, monitor accuracy and completeness through DQ metrics, and, most importantly, link the work to business impact: cost savings, efficiency gains, or revenue growth.
Table 1 highlights a checklist teams can use to assess readiness across people, process and technology. The key is to diagnose and remediate bottlenecks that slow speed to insight before layering in a new tool. If these prerequisites (single source of truth, stable pipelines, codified governance rules) are missing, adoption of a DQ tool is premature.
Table 1
Other considerations includemapping how data moves across platforms and where reconciliations occur, establishing a baseline for time-to-insight, and expanding self-service analytics so business users can access trusted data without overreliance on IT or data engineering teams.Once these bottlenecks are addressed, a DQ tool can operate as intended.
Where: Placement in the Stack
The data governance value chain (Figure 1) highlights where automation can occur, such as data health monitoring, enrichment, and reporting, and clarifies where AI-enabled tools can drive measurable value.
Regarding technology, the central question is where these tools fit. Figure 3 shows DQ tools positioned as a dedicated governance software layer above the company’s lake, warehouse, or lakehouse. The tool serves as a checkpoint to enforce standards and monitor quality before data is consumed by analytics, AI models, or business processes. The “where” is therefore not across the entire stack but in the governance layer itself, bridging raw data sources and the systems that consume them. This placement ensures governance remains consistent even as architectures evolve.
Figure 3
NOTE: Metis does not endorse a vendor over another; companies depicted here are illustrative in nature
What: Prioritizing the Business Case
The “what” often comes down to specific use cases. Organizations may automate lineage tracking, standardize definitions, or generate quality metrics that previously required manual effort. In some cases, leaders may also prioritize metadata management or compliance monitoring, where automation can reduce both cost and risk. These examples clarify not only what the tool should do, but also what business needs it must serve.
Moreover, developing and prioritizing business cases for a new initiative can help teams identify what data sets to target, leading them to identify which may require DQ augmentation. As we’ve seen through our work, the majority of distinct business cases today have unique, underlying data sets, allowing teams to define the “what” they are targeting.
Picking the Best Tool: Decision Matrix
Table 2: Tooling Decision Matrix
Selecting the right AI-enabled DQ tool is complex. With a flood of options and executive pressure to scale, tech leaders often lack the time and resources to differentiate between similar offerings. The decisioning matrix provides a structured way to evaluate tools across people, process, and technology.
For example, when evaluating a tool of interest, leaders would walk through how it performs across the major dimensions of the matrix. They might start by looking at the technical foundations: does the tool integrate cleanly into the existing ecosystem, or does it require workarounds because it supports only a narrow set of warehouses and APIs? From there, they would examine whether it can realistically handle the organization’s data patterns, the frequency of jobs, the expected volumes, and the balance of batch and streaming needs. Governance naturally follows: can the tool enforce the level of auditability and SLA rigor the enterprise expects, or does it stop at basic event logging? After grounding the assessment in these functional areas do the people-focused considerations come into view, how intuitive the interface is for stewards, how quickly teams can onboard, and whether support extends beyond documentation.
Scoring each of these areas on a Likert scale allows teams to quantify qualitative measures, creating a consistent basis for comparison that ensures decisions reflect functional readiness rather than isolated, flashy features.
In the current environment, most AI-enabled tools, including DQ-focused ones, are funded by VCs. It’s important for organizations to assess vendor durability alongside feature fit, as well as seek solutions aligned with both their immediate needs and long-term strategy.
In addition to the above matrix, enterprises can assess alignment by:
Parting Thoughts
Adopting a DQ tool is not a standalone technology choice; it’s a decision rooted in governance maturity, people readiness, and business alignment. Organizations that treat the choice this way are better positioned to realize lasting value from their AI investments.
