How AI forces asset managers to rethink their strategy

I previously argued that artificial intelligence is not primarily a tooling problem for asset managers, but an operating model problem. The significance of AI lies in its impact on how organisations learn, scale decisions, and embed responsibility.

Click here for the previous article in this series: 'How AI forces asset managers to rethink their operating models'.

This second article builds on that foundation. If AI reshapes operating models, then strategy in asset management must also be reconsidered. The central claim developed here is that strategy can no longer be separated from an organisation’s ability to organise data, learn from it at scale, and manage growing complexity responsibly.

Executive summary

In this second article, I further develop the argument that artificial intelligence in asset management fundamentally reshapes operating models and strategy. If AI reshapes how organisations learn, scale decisions, and manage complexity, then strategy can no longer be separated from how data is organised and how operating architecture is designed.

The article introduces a different lens on technological change, framing it as a collision rather than a traditional disruption narrative. AI-first firms operate according to fundamentally different economic and organisational logics, shaping how they compete. These structural differences often remain hidden for extended periods, particularly in regulated and trust-based industries such as asset management, before manifesting abruptly in competitive outcomes.

The strategic distinctiveness of AI-first operating models is analysed through three reinforcing dimensions: digital scale, digital scope, and learning effects. AI does not alter the financial scalability of asset management, but it changes how operational complexity scales by embedding decision support and monitoring into digital systems. It enables reuse of data and insights across data products and teams without undermining governance, and it institutionalises learning so that improvement becomes continuous and organisation-wide.

At the centre of this system lies data. For incumbent asset managers, the binding constraint on AI strategy lies in data organisation rather than in technology itself. Most organisations possess large volumes of data that remain difficult to mobilise in practice. Fragmented architecture, inconsistent definitions, and limited reusability prevent AI initiatives from moving beyond experimentation. As a result, AI maturity is more closely linked to data engineering than to algorithmic sophistication.

Crucially, data is not “just IT”. It embodies business knowledge such as investment assumptions, risk interpretations, and governance choices. Effective AI-first operating models therefore require business ownership of data, technically enabled by IT, rather than IT-driven data management alone.

Finally, the article positions governance as a strategic variable. Transparency, traceability, and accountability underpin sustainable AI-enabled growth rather than standing in its way. In an AI-driven environment, strategy increasingly emerges from operating architecture: how data flows, how decisions are coordinated, and how learning accumulates across the organisation.

The concluding implication is clear. In the age of AI, competitive advantage in asset management depends on the organisation’s ability to design an operating architecture that allows learning, governance, and responsibility to scale together on a performant data platform.

From disruption to collision: a structural perspective

Technological change is often framed as disruption: new entrants innovate, incumbents react. This framing suggests speed and surprise as decisive factors. Yet it fails to capture what happens when firms built on fundamentally different operating logics begin to compete.

What we are witnessing today is better described as collision. AI-first firms differentiate through fundamentally different economic and organisational principles, extending well beyond better products or lower prices. Their cost structures, learning mechanisms and scaling dynamics differ structurally from those of traditional firms.

History shows that such collisions rarely result in immediate displacement. Instead, they lead to gradual shifts in industry structure. Firms may remain competitive for extended periods while the basis of competition quietly changes underneath. When outcomes eventually diverge, they often do so abruptly.

Asset management is not immune to this dynamic. Regulation, long investment horizons and trust-based client relationships can mask structural changes, but they do not prevent them.

What makes an AI-first firm strategically different

To understand the strategic implications, it is useful to step away from specific technologies and focus on underlying economic mechanisms. AI-first firms differ from traditional organisations along three reinforcing dimensions: digital scale, digital scope and learning effects. All three rest fundamentally on how data is organised and mobilised.

Digital scale: managing complexity instead of headcount

In asset management, growth is not mechanically tied to headcount. Managing €10 million or ten times that amount does not automatically require ten times more professionals. The business is, by nature, scalable.

That scalability, however, is not absolute. As assets grow, operational complexity increases: more instruments, more markets, more regulatory oversight, more reporting requirements and more internal coordination. These elements introduce less scalable components into an otherwise scalable business model.

AI-first operating models leave the scalability of asset management largely intact, while fundamentally altering how operational complexity scales. By embedding decision support, monitoring and process logic into digital systems, they reduce the marginal cost of managing additional complexity. Growth remains scalable as human attention is increasingly directed towards judgement, oversight and exception handling.

This shift is only possible if data is accessible, consistent and reliable across the organisation.

Digital scope: from repetition to reuse

In asset management, scope has traditionally expanded through products, mandates and asset classes rather than through platforms. Knowledge, models and insights are often developed within specific teams and remain closely tied to those contexts. This has supported accountability and fiduciary clarity, but it also limits reuse.

AI-first operating models retain product-based structures while adding an additional layer of organisational reuse. Data, analytical components and decision-support capabilities are designed to be shared across strategies where appropriate, without collapsing necessary distinctions in risk ownership or responsibility.

The strategic shift concerns a move away from repetition towards systematic reuse, without abandoning specialisation. Scope expands as learning generated in one context increasingly informs decisions in others. In an industry that is already financially scalable but operationally complex, this form of AI-enabled scope becomes a meaningful differentiator.

Again, this presupposes that data can be reused both technically and organisationally, without eroding governance.

Learning effects: institutionalising learning

In asset management, learning effects focus on strengthening judgement rather than automating or replacing investment expertise. They are about systematising how organisations learn from their own activity.

Traditionally, learning is episodic and human-centred. Insights emerge through post-mortems, investment committees and expert experience accumulated over time. This model can be effective, but it scales imperfectly. Learning remains unevenly distributed, difficult to codify and slow to propagate.

AI-first operating models introduce a complementary mechanism. Decisions, outcomes and contextual data are captured continuously and fed back into analytical systems. Patterns become visible that would be difficult to detect through isolated human processes alone. This approach preserves discretion and accountability while extending their effectiveness.

The strategic value of learning effects is expressed through greater consistency and adaptability over time. Learning accumulates institutionally rather than residing primarily in experts.

Scale, scope and learning as a coherent strategic system

Taken together, digital scale, digital scope and learning effects form a coherent strategic system rather than separate advantages.

Asset management is financially scalable, but operational complexity grows with size. Digital scale determines how efficiently that complexity is absorbed. Digital scope determines whether insight developed in one part of the organisation can be reused elsewhere without undermining governance. Learning effects determine whether the organisation improves as it grows, rather than merely becoming larger.

AI-first operating models align these dimensions as part of a coherent system. Scale without learning leads to fragility. Scope without governance leads to risk. Learning without structure leads to inconsistency. At the centre of this system lies data.

Data as the hidden constraint (and enabler) of AI strategy

Many discussions around AI implicitly assume that data is readily available and usable. For incumbent asset managers, this assumption rarely holds.

Most organisations possess large volumes of data that remain difficult to mobilise in practice. Data exists across portfolio systems, risk platforms, market data feeds, client systems and reporting tools. Yet it is often fragmented, inconsistently defined and difficult to activate at scale.

In practice, the constraint lies in how data is organised rather than in its availability. Without well-structured, well-governed and accessible data, AI models cannot be trained, validated or deployed reliably. There is no AI-first operating model without a deliberate effort to make data usable across the organisation. Many AI initiatives stall because foundational data work is underestimated.

Data engineering before AI engineering

AI strategies frequently fail because they start too late in the stack. Organisations invest in models, proofs of concept or vendor solutions before addressing underlying data architecture.

In practice, the most demanding part of AI transformation is not model development, but data engineering:

creating consistent data definitions,

designing data pipelines rather than point-to-point integrations,

ensuring data quality, lineage and traceability,

and making data accessible without compromising control or compliance.

This work is technically complex, slow and often invisible. It rarely produces immediate business wins, which makes it difficult to prioritise. Yet it is precisely this layer that determines whether AI can move beyond experimentation.

For asset managers, AI maturity is therefore far more closely linked to data readiness than to algorithmic sophistication.

Data is embedded business knowledge

A second, more subtle challenge concerns ownership.

In many incumbent organisations, data is primarily managed by IT. This is logical from an infrastructure and security perspective. But it obscures a critical reality: much of an asset manager’s business intellectual property is embedded in data.

Data encodes investment assumptions, risk interpretations, portfolio construction logic, client segmentation and regulatory understanding. These elements reflect domain expertise and strategic intent.

AI-first operating models therefore require a shift from IT-owned data to business-owned data, technically enabled by IT. Domain experts must actively shape data semantics, quality thresholds and usage constraints. Without this involvement, AI systems risk being technically sound but strategically misaligned.

Why incumbents struggle and why this is not a failure

For non-AI-native asset managers, reorganising data is often harder than adopting new technology.

Legacy systems were designed for stability, reporting and control, not for reuse and learning. Data models reflect historical organisational boundaries. Regulatory requirements have reinforced fragmentation rather than integration.

As a result, incumbents often face a paradox: they possess decades of high-quality data, yet struggle to mobilise it for AI-driven learning. This outcome reflects the constraints of operating models optimised for a different era.

Crucially, none of the strategic accelerators discussed below can materialise without first treating data as a shared, governed and reusable organisational asset.

Strategic accelerators (once data has been organised)

Once data is organised and mobilised, AI-enabled operating models unlock several accelerators:

learning velocity, as feedback loops shorten,

experimentation speed, as hypotheses can be tested cheaply,

architectural defensibility, as advantage becomes structural,

strategic optionality, as new services and partnerships become feasible.

These accelerators reinforce one another. Together, they redefine competitive dynamics.

Governance as a strategic variable

As AI systems scale, ethical considerations such as bias, transparency, privacy and inclusion become operational risks. Governance determines whether AI-enabled growth is sustainable.

Here again, data is central. Governance is only effective if data lineage, usage and accountability are transparent. For European asset managers, strong governance traditions are a strategic asset if embedded early rather than retrofitted later.

Why operating architecture now defines strategy

In an AI-enabled environment, strategy can no longer be treated as a matter of vision statements or long-term ambition alone; it increasingly emerges from the operating architecture that determines how data flows, decisions are taken and learning accumulates across the organisation.

Historically, strategy in asset management focused on positioning: product offerings, market segments, distribution and performance differentiation. These choices remain important, but they no longer fully explain competitive outcomes in an environment where learning speed and complexity management matter as much as intent.

Operating architecture translates strategy into reality. It determines whether data can be mobilised across the organisation, whether insights can be reused without undermining governance, and whether learning remains episodic or becomes institutional.

In the age of AI, this distinction becomes decisive. Ambition without architectural alignment produces experimentation without impact. Conversely, well-designed operating architecture often generates strategic options that were never explicitly planned.

For asset managers, this does not imply abandoning long-term vision or fiduciary discipline. It implies recognising that competitive advantage increasingly depends on how the organisation is built to learn, rather than on what it intends to achieve.

Conclusion

Artificial intelligence challenges asset management primarily through its impact on how organisations manage complexity and institutionalise learning, rather than through task automation.

The real constraint on AI strategy is rarely technology. It is the organisation’s ability to treat data as a strategic, shared and business-embedded asset and to design an operating architecture that allows learning, governance and responsibility to scale together.

In my upcoming article, I move from strategy to execution: what a realistic transformation journey looks like for non-AI-native asset managers, and which steps matter most in practice.