On December 10, 2025, I had the privilege of moderating an opening keynote panel, 'AI and Information Management in the Capital: Framing Today's Challenge Landscape,' at AIIM's AI+IM Regional Forum in Washington, DC. Joining me were four distinguished information leaders who brought perspectives from government, nonprofit, and legal sectors: Thad Lurie from American Geophysical Union, Mark Patrick from the US Department of Defense, Melvin Baskin from ICG Consultants & Advisors, and Leigh Zidwick from DLA Piper LLP.
Three years after ChatGPT's public launch, AIIM's research shows that AI adoption has reached 91% critical mass when combining current use with planned implementation for both generative AI and agentic AI. With this increase in adoption, this panel discussion revealed that we need practical strategies for AI success grounded in solid information management principles.
We opened with a simple question: What about AI keeps you up at night?
The answers ranged from undefined boundaries for AI to managing the gap between expectations and reality. But a common thread emerged—there's significant opportunity in AI governance if we position ourselves correctly.
This shift from experimentation to implementation is critical, especially given that a recent MIT study found 95% of AI projects fail. Success requires focusing on business requirements and problems rather than "shiny objects." Thad recommended starting with small footprint projects where organizations can evaluate risk, make modest investments, and carefully monitor ROI. As a panel, we also discussed responsible AI. AGU has established AI ethics frameworks and plans to focus on monitoring the application of those ethics in 2026.
One of the most important themes that emerged was the need to transition information management from a compliance engine to an AI enabler. Melvin highlighted how IM professionals should be embedded throughout the AI lifecycle, from helping with system selection to positioning the IM program as vital to AI success.
One panelist brought this into sharp focus with a question many organizations are grappling with: who determines whether AI output qualifies as a record?
Panelists agreed that anything that is an artifact that could be used for a decision or process should be considered a record. However, the process of establishing and cataloging what constitutes a record needs to be automated and happen by default. It will not be possible for humans to manually manage AI output as records.
This speaks to the fundamental role information professionals must play. As Melvin noted, "Trust is key and information professionals are the gatekeepers of trust." Creating that trust costs money, but it's essential to AI adoption and success.
Leigh shared insights from DLA Piper's AI working group, emphasizing the importance of bringing diverse teams to the table. Key practices include developing a framework for decision-making, establishing a known and approved catalog of AI tools, meeting regularly, and maintaining transparency in decision-making processes. A primary focus of the working group is to efficiently and transparently assess the risk of AI use cases and tools.
This governance approach addresses a common challenge: how to mitigate the risks of rogue AI and shadow AI without becoming the department of "no." Melvin drew a parallel to shadow IT, noting that the best course of action is to position yourself as close as possible to stakeholders so you understand and anticipate their needs.
Information professionals often find themselves in a difficult position: responsible for identifying and communicating risk, yet potentially perceived as bearing sole responsibility for that risk. This model doesn't work with AI. Responsible AI requires shared responsibility, which means the risk can't sit with one person or one team.
Panelists agreed that data curation leads to AI success. Information professionals need to help stakeholders and leadership understand that records are created when artifacts are absorbed into business processes, and ensure they understand that their role in governing AI artifacts is the same as governing other records. We are all responsible for data quality.
With expanding tech stacks and increasing data demands on RIM teams while resources stay flat, Leigh emphasized the importance of understanding organizational strategy, defining clear roles, and leveraging AI working groups to monitor and shape evolving roles. Most importantly, share with leadership what will happen if you don't invest in resources to support AI.
As we closed the panel, I asked each speaker for one piece of advice for the audience. Their responses captured the essential mindset information professionals need. Their advice included:
As Thad memorably put it, "AI is a taker, but it's also an imperative."
The path forward requires information professionals to move beyond compliance and position themselves as essential enablers of AI success. The opportunities are significant, but only if we're ready to embrace new skills, stay close to stakeholders, and be the "yes" our organizations need.
The AI+IM Regional Forum brings together information leaders to address the critical challenges at the intersection of AI and information management at a local level. Learn more about our series of regional events and see where we are headed next.