The Intersection of Regulation, Research and AI – Implications for Asset Managers Post MiFID II
This White Paper will consider the interaction of AI in the asset manager investment process with the legacy MiFID II research frameworks – the long-tailed impacts of which may have a major bearing on the successful implementation of AI strategies for MiFID II-scope managers.
Forthcoming White Papers will consider this from the perspective of upcoming, regulation, specifically the EU AI Act, which, despite being an EU regulation will likely affect all managers with any EU clients.
Recent History
Recent regulatory developments in the UK and EU highlight that policymakers increasingly recognize how divergent rules can affect the competitiveness of their domestic asset management industries and, by extension, asset owners. The rapid emergence of AI in asset manager investment processes will vastly increase the impact of these disparities for all market participants.
MiFID II reshaped the economics of investment research in Europe. The net result was that many UK/EU managers opted to fund research via their P&Ls. As research was often the manager’s largest cost after compensation it resulted in a ~75% decline in external research spending.
This clearly reduced research access for UK/EU managers. Most US managers continued to rely on client-funded research models with relatively limited change, creating a substantial trans-Atlantic divergence in research budgets and, consequently, a material information asymmetry.
In response, UK/EU regulators have made it easier for managers to use client money for research in order to level the competitive playing field with US managers and to reduce the market risks in P&L research budgets.
The Emergence of AI in the Investment Process
The arrival of AI into the investment process has the potential to vastly amplify existing research-related information disparities. AI’s impact on investment decision-making and operating models is likely to be so profound that the information gaps created in the MiFID II era may appear quaint by comparison. AI presents both significant risks and substantial opportunities for asset managers and pension funds.
Early manager AI use cases including client chat, marketing support and document summarization are rapidly giving way to integration into the investment process as AI auditability and governance have improved.
Asset manager executive committees have rightly been cautious about exposing core investment processes to AI. But the medium‑term question is no longer whether AI will be used in investment organisations; it is whether it will be used well, under governance that supports both performance and control. The potential productivity gains are unprecedented – and the competitive implications existential.
Key Requirements and Guardrails for AI Adoption
To capture AI’s benefits while mitigating its risks, a set of practical requirements and guardrails has begun to emerge:
- Walled Gardens: AI models are run over curated, vetted content sets, typically combining external research with internal reports. This enables managers to maximize the value of their historical research corpus and materially reduce hallucinations, as the content universe is pre‑approved and trusted.
- Source Attribution: AI systems should be able to identify immediately the specific documents and data underpinning their analysis, enabling transparency, auditability and easier validation by investment teams.
- Protection of Manager IP: Within a walled-garden architecture, managers can develop internal models without inadvertently training external AI systems on proprietary content, addressing a key weakness of some widely used tools.
- Robust AI Governance: Formal AI governance frameworks are now critical.
- Workflow Native Design: Generalist large language models may be suboptimal when applied to complex, technical asset management tasks. Specialist small language models (SLMs) designed for investment workflows can materially improve the quality, relevance and reliability of outputs.
This architecture also forces an important discipline: clear data and content rights over what is ingested and how it is used.
Practical Design Principles for Asset Manager AI Platforms
AI “point solutions” for asset managers that speed up one workflow step (accelerated analysis of corporate earnings releases for example), can be valuable, but the bigger productivity gains come when AI is embedded across the research lifecycle—from discovery to synthesis to internal knowledge reuse.
Key Design Principles:
- Customization/Personalised Interrogation of the Corpus: Allow PMs and analysts to embed their priorities into search and analysis, so the same body of content can be systematically interrogated through each team’s investment lens.
- Leveraging IP Across Complex Organizations: In multi-asset siloed enterprises, AI can ensure that collaboration/knowledge sharing in the investment process becomes systematic rather than accidental, aiding the discovery of high‑value internal IP (including other people) – a significant USP for managers.
Downstream Competitive Implications
AI usage and governance are set to become central elements of the competitive landscape for asset managers. Investment consultants, fund selectors and pension funds will need robust frameworks to evaluate managers’ AI strategies, including architecture, governance, data policies and integration with investment processes.
Implications for Asset Owners and Pension Funds
Beyond manager selection, pension funds with internal investment teams will need to determine how best to deploy AI within their own internal investment processes.
AI may also significantly reshape the traditional cost-versus-value debate for pension funds, particularly in the UK context.
The scale of potential AI-driven value creation suggests that modest increases in manager costs—whether to support research or AI development—could generate outsized improvements in investment performance and productivity.
These developments sit squarely within the FCA and UK Pensions Regulator’s “Value for Money” framework.
Are UK Pension Funds maximizing the value they receive from external managers by denying them a couple of basis points in research or AI costs?
If AI materially increases manager productivity and decision quality, then modest differences in fee levels may be trivial compared with the performance dispersion between funds that are frequently measured in thousands of basis points across multiple active categories
The key question for trustees and CIOs is: Are we optimising for the lowest visible cost line, or for the highest probability of sustained net outperformance delivered through a well‑governed Ai process advantage?
Historically, asset owners have been willing to pay high management fees for exposure to managers with demonstrable investment process advantages and/or superior information.
As AI adoption in asset management accelerates, it is likely to widen competitive gaps between managers that deploy it effectively and those that do not.
For both asset managers and pension funds, thoughtful, well‑governed engagement with AI is no longer optional.
The next paper will address practical steps asset managers can take to meet the EU AI Act requirements while simultaneously building the confidence of asset owners and regulators in the manager’s AI governance process.
Even the most powerful LLMs will not likely meet governance requirements “out of the box” when applied to asset manager investment processes. To understand how Contours supports research governance in regulated investment environments, contact us.
Contours is a KiteEdge product.
