AI in Operational Due Diligence: The New Risk Landscape 

Just a few short years after LLMs first launched, artificial intelligence is now a becoming an established part of many companies’ day-to-day processes – Mercer’s 2026 AI in Asset Management Survey found that 55% of managers report AI is integrated into at least one of their investment processes. Further, 91% plan to increase their use of AI in the next 12 months*.

But in an environment of rapid evolution and differing capabilities, it can be difficult to establish uniform standards. As asset managers move beyond their initial phase of experimentation into implementation and scaling AI-based tools within their organizations, the question is how they are using AI within their organizations – and with what oversight.

The pressure on asset managers to show how they are adopting AI, and the improvements they have driven by doing so, could potentially create perverse incentives to ‘AI-wash’ – allege or exaggerate capabilities that are as yet unproven.

Plunging headlong into implementing AI presents its own concerns that these asset managers and their clients must address, from third-party risk mitigation to output validation. This marks an inflection point in how large investors approach their operational due diligence (ODD).

Across the investment landscape, AI integration and adoption varies widely.

Large, well-capitalized managers may have dedicated AI leads, internal development teams and structured implementation programs. At the other end of the spectrum, boutique and emerging managers may be operating with lean teams, limited budgets and far less formal oversight. Yet both increasingly compete for capital in the same ecosystem, particularly as private markets continue to democratize and reach broader investor segments. The difference in resources can result in profound differences in AI implementation and maturity from one manager to another.

For allocators, this presents a governance challenge. AI capability is becoming part of institutional credibility. But innovation alone is not enough. What matters is how it is controlled, monitored and explained.

ODD processes must now assess questions that were rarely raised two years ago:

• Is there a defined AI governance framework?
• Who is accountable for oversight?
• How are models tested and validated?
• What happens when systems fail?
• Do asset managers understand downstream AI dependencies and concentration risk among their vendors?

These are not theoretical considerations, and must be embedded into manager evaluation processes.

In the early phase of AI adoption, it was enough for investors to simply communicate their ambitions for AI usage. Today, the conversation requires something more tangible. Managers are being evaluated against structured AI principles covering governance, risk assessment, third-party exposure, incident response and ongoing monitoring. The shift mirrors ESG’s evolution: from awareness to disclosure, from disclosure to measurement, and from measurement to accountability.

After a year of observing implementation spreading across the market, clear differences in maturity are emerging. Some firms have invested meaningfully in governance and oversight. Others are still navigating basic questions around data exposure, the underlying technology stack and vendor risk – indeed, we have seen operational due diligence issues rise 5% year-over-year just in technological deficiencies alone.

The message is clear: AI adoption without structure is not sustainable.

A range of factors are required for responsible AI deployment, from data integrity, to domain experts-in-the-loop, to clear accountability and escalation paths. All of this ladders up towards explainability – the ability to see and understand how an output was reached.

Fiduciaries must be able to understand how decisions are reached, how data is used, and how outputs are validated. This is particularly important as AI tools become embedded in operational workflows such as document review and monitoring processes.

We do not yet know what the final regulatory environment will look like, but fiduciary responsibility does not pause while regulation takes shape

Operational due diligence sits at the intersection of innovation and accountability. Its role is not to slow adoption, but to ensure that adoption is disciplined.

At present, most AI implementation in investment management is spread across a range of areas such as research, portfolio analytics, cybersecurity monitoring, drafting investment memos and even coding copilots to assist with data analysis and tool development.

Boards, however, are already asking the next question: where are the measurable outcomes? And what were the costs to achieve them?

The reality is that we are still in the early innings, but the industry has already changed meaningfully in a short period of time. Just as electronic trading floors once replaced rooms full of traders, AI will likely become embedded and less visible over time. But the path from efficiency gains to investment alpha is neither immediate nor guaranteed.

Responsible adoption requires experimentation and stress-testing solutions before scaling up.

AI is not a passing trend. But its long-term impact will not be determined by speed of adoption alone. It will be defined by governance, transparency and the ability to explain how decisions are made.

In this series, we will explore how operational due diligence is adapting to this new landscape: from governance frameworks and third-party risk to monitoring, training and case studies of implementation.

Important notices

References to Mercer shall be construed to include Mercer (US) LLC and/or its associated companies.
© 2026 Mercer (US) LLC. All rights reserved. Please read here: Important notices