Thank you to everyone who joined CENTRL’s inaugural AI Practitioners Discussion, A CIO Discussion: Actual Use of AI in Investment Diligence.
Our goal was to make this a practitioners’ discussion, not another AI hype session - starting with the problem and then thinking about AI, not a technology solution looking for a problem. Rather than focusing exclusively on AI’s enormous potential, we wanted an honest, open conversation about what is actually working, and the significant risks as investment professionals implement AI.
Together with investment leaders Jim Smigiel and Andy Conner, who ran two elite OCIOs and foundations respectively, we explored the practical realities of AI adoption across investment research, manager due diligence, and client service.
The most pronounced takeaway from the discussion is that an AI platform is worthless without structure, security, and organization.
Below are seven key themes, highlighting both the opportunities and the challenges discussed:
1. Productivity Gains vs. Implementation Complexity
The opportunity: AI is already delivering meaningful productivity improvements by automating repetitive work, including meeting notes, research summaries, investment reporting, and client communications. Concrete applications included various “delta analyses" - quickly identifying subtle differences across managers or across time, and supporting data-intensive 15(c) advisory-renewal work.
The challenge: Moving beyond isolated use cases to integrated enterprise workflows across the entire lifecycle is considerably more complex. Without structured platforms, governance, and coordination, proliferating AI agents can quickly become difficult to manage and unreliable.
2. Expanded Research Capacity vs. Investment Quality
The opportunity: AI lets investment teams process substantially more information, evaluate more managers, and pursue opportunities they might otherwise miss because of limited time and resources.
The challenge: More research does not necessarily produce better investment decisions. Greater volume and speed must be accompanied by rigorous human analysis and genuinely differentiated opportunities. In practice, AI-assisted diligence needs a defined thesis-driven framework, fully fleshed-out positive and negative views, and multistage refinement and challenging of the thesis.
3. Objective Analysis vs. Confirmation Bias
The opportunity: AI can synthesize extensive research, identify potential risks, and help investment professionals evaluate and challenge investment theses by presenting alternative views and devil’s advocate perspectives.
The challenge: Not properly configured, AI can also reinforce existing assumptions, creating an echo chamber of seemingly well-supported conclusions. Investment teams must deliberately seek disconfirming evidence rather than use AI simply to validate their views. Practical ways to challenge a thesis include producing a “pre-mortem” memo - write the theoretical termination thesis prior to the hire. Additionally, compare overlap between the actual investment memo and an AI-generated memo written purely from manager-provided materials.
4. Democratization of Research vs. Homogenization of Thinking
The opportunity: AI gives smaller investment teams access to analytical capabilities and research capacity that historically required substantially greater resources.
The challenge: As more firms rely on similar converging underlying models, research and recommendations risk becoming increasingly standardized. Independent judgment must be explicitly built into the logic to distinguish from the crowd. We all know alpha, by definition, is having a divergent point of view.
5. AI as an Analytical Assistant vs. Erosion of Human Judgment
The opportunity: AI can operate like a highly productive analyst, rapidly preparing drafts, organizing information, and freeing experienced professionals to focus on higher-value decisions. Practitioners described the usefulness of using AI for identifying subtle changes over time that can be indicative of thesis creep.
The challenge: Investment theses, manager assessments, and final investment decisions must remain human responsibilities. Excessive AI dependence could also weaken the development of junior professionals who need hands-on analytical experience to build investment judgment. There must be explicit platform tooling for human sign-off and scoring of all AI-generated output.
6. Historical Knowledge vs. Forward-Looking Insight
The opportunity: AI excels at synthesizing historical information, identifying patterns, and making institutional knowledge more accessible.
The challenge: Just like any machine-oriented investment strategy, AI can run the risk of overfitting history. Markets, investment strategies, and managers evolve. Heavy reliance on historical patterns can inhibit the recognition of structural changes, emerging strategies, and opportunities that depart from conventional thinking. An appropriately configured and tested approach can create checks to ensure there is not overreliance on history alone.
7. Bottom-Up Innovation vs. Enterprise Governance
The opportunity: Encouraging experimentation allows for creativity across the investment teams to identify valuable applications organically, while organizations can scale successful solutions across the enterprise.
The challenge: AI introduces substantial risks around data confidentiality, proprietary intellectual property, output accuracy, and process integrity. Secure, well-structured platforms with clear controls, traceability, and human oversight are essential. A well-structured platform can strike the balance between innovation and creativity with governance, oversight, and security. Practitioners also highlighted the value of an explicit AI usage policy and external compliance support, and reported being extremely careful about client information in models.
The Bottom Line
The discussion reinforced that AI’s potential in investment management is substantial, but deploying agents is precarious, and structure is everything.
AI can make investment teams faster, more productive, and thorough. But speed does not guarantee accuracy, more information does not guarantee better decisions, and automation does not replace experience or judgment.
The winners will be firms that successfully balance these competing forces, capturing AI’s efficiency and scale while preserving the independent thinking, disciplined processes, and fiduciary accountability that drive sound investment decisions.