Hello. This is Alex Bakker and Olga Kupriyanova standing in for Stanton Jones with what’s important in the IT and business services industry this week.
If someone forwarded you this briefing, consider subscribing here.
AI ROI: Timing is Everything
In our 2025 ISG Market Lens Data and AI Programs Study, we found that the activities most closely tied to data quality — data governance, master data management and data quality and insights — are also the ones most likely to be handled by in-house staff. (see Data Watch)
Data Watch
Other areas of data management are much more likely to be handled with third-party support. For example, only 43% of enterprises run data warehouse and data lakehouse analytics with in-house FTEs (57% use contractors or MSPs), and just 44% keep data observability in-house (56% use third-party). That makes it clear that enterprises treat data governance, master data management and data quality differently, choosing to keep them closer to the vest than other parts of the data stack.
The Disconnect with AI Adoption
Our study also makes clear that this approach to data programs has not solved the core challenges of AI adoption.
The number one data challenge enterprises cite is data usability for AI applications, with 53% of enterprises placing it in their top five challenges. The number one AI challenge is demonstrating ROI to the business (33% put it in their top three) and the number two AI challenge is data quality, accuracy and consistency (32% put it in their top three).
Taken together, these results show that, even with heavy reliance on in-house teams for foundational data work, enterprises are struggling to align data relevance, usability and quality to deliver AI outcomes.
A Capacity Challenge, Not Just a Capability Challenge
AI adoption is uniquely time-bound. If organizations don’t deliver ROI quickly, projects lose momentum and funding. And AI needs data to deliver ROI. Which means data issues that could be addressed gradually in the past now have to be solved under tight time pressure. This is why businesses need to focus on identifying which data is the most aligned and relevant to the business outcomes.
The issue is not whether in-house teams have the skills — many enterprises have capable data staff. The issue is capacity. The scale and speed required to fix data quality and governance problems often exceed what internal teams can do on their own, and because AI often tackles problems that require significant data integration, it places extra pressure on profiling data correctly before it can be made available to AI.
That’s why third-party support matters in this context. While enterprises rarely outsource data governance, master data management or data quality today, external partners can provide surge capacity and specialized expertise to help accelerate improvements. The driver is not cost savings — it’s time-to-value for AI.
What’s Next?
For enterprises aiming to achieve differentiated AI outcomes, the real question is not whether these functions remain in-house. It’s whether in-house alone is fast enough.
The data shows that, while enterprises continue to hold data quality functions much closer, they are comfortable using third parties for areas like data warehouses, observability and integration. And, as AI adoption pressures mount, this imbalance may prove unsustainable.
Without faster improvements to underlying data, AI initiatives risk missing ROI targets and losing support. That sets up a clear decision point: continue relying almost exclusively on in-house teams, or bring in outside expertise that can address the capacity problem that AI has made impossible to ignore.
About the authors
Alex Bakker
