Date: May 11 2026
Author: Bella Coles Gazolli

Artificial intelligence is increasingly embedded within the asset management industry, not as a speculative innovation but as a response to mounting structural pressure. Declining margins, intensifying competition from passive strategies, and the rapid expansion of available data have forced firms to search for new sources of efficiency and differentiation. In this context, AI has moved from a peripheral capability to a strategic priority.

This report combines secondary research from leading industry sources with primary interview evidence drawn from professionals across a range of asset management roles. The findings point to a clear and consistent pattern. While adoption is widespread, it remains shallow in many areas. AI is delivering measurable improvements in productivity, yet it has not fundamentally reshaped the core of investment decision-making.

Three conclusions stand out.

First, AI is primarily an efficiency tool rather than a decision-making engine. Across firms, the most common applications are summarisation, drafting, and information aggregation. These uses generate tangible time savings. In one case, an interviewee estimated time savings of around 20 per cent in specific workflows, but they rarely extend into final investment judgement.

Second, human judgement continues to dominate the investment process. Even where AI supports analysis, final portfolio decisions remain discretionary. Interview evidence suggests that investment decisions remain overwhelmingly driven by human judgement, with AI playing a supportive role rather than a determining one, reflecting both client expectations and the limitations of current systems.

Third, the principal barriers to adoption are not technological but organisational. Data security concerns, fragmented infrastructure, and limited trust in AI outputs constrain deployment. Firms are particularly cautious about exposing proprietary data to external systems, leading to a strong preference for internal, controlled AI environments.

Looking ahead, AI is likely to reshape how work is performed rather than who performs it. The industry appears to be moving towards augmentation, where machines handle scale and repetition, while humans retain responsibility for judgement, interpretation, and client relationships. The gap between AI’s theoretical potential and its current application remains significant, but it is unlikely to persist indefinitely.

Introduction and Industry Context

Artificial intelligence is often presented as the next phase of technological progress in asset management. In practice, its adoption is being driven by something less abstract. The economics of the industry have become more difficult.

Margins have been under sustained pressure for several years. Fee compression has intensified as passive strategies continue to capture market share, while the cost base has remained stubbornly high. At the same time, the volume of data available to investment teams has expanded rapidly, increasing the effort required to process information without necessarily improving outcomes. Against this backdrop, technology spending has risen, yet the link between that spending and productivity remains weak (McKinsey, 2025). The result is a growing sense that existing approaches are no longer sufficient.

This is the context in which AI is being adopted. It is not being introduced simply because it is new, but because firms are searching for ways to restore efficiency. Industry surveys reflect this shift. A large majority of asset management executives now view AI as critical to their future, and estimates suggest that its impact could be material relative to overall costs (Grant Thornton / ThoughtLab, 2025; McKinsey, 2025). The emphasis is less on innovation for its own sake and more on addressing a structural imbalance between rising complexity and constrained margins.

Progress, however, is uneven. Some large institutions have begun to embed AI across multiple parts of the business, building internal systems and integrating them into daily workflows. Elsewhere, adoption is more tentative. In many firms, usage remains confined to pilot projects or individual experimentation. Operational deployment, in particular, tends to lag behind strategic ambition. The gap between what firms intend to do with AI and what they are currently able to implement remains significant.

Rather than assuming a uniform trajectory, it is more useful to look at how AI is actually being used. This report takes that approach. It combines industry research with interview evidence from practitioners across different roles and firm types, focusing on three questions that recur throughout the analysis. Where does AI currently create value? What constrains its wider use? And how might these dynamics alter the way the industry functions over time?

Key AI Trends in Asset Management

The current wave of AI adoption in asset management is being shaped by a set of technologies that are often discussed separately, but in practice tend to overlap in how they are used.

At the most visible level, generative AI has attracted the most attention. Its appeal is straightforward. It allows users to generate and manipulate text quickly, which makes it immediately useful in areas such as drafting research notes, summarising earnings calls, and producing client-facing material. Many firms have already deployed internal versions of these systems, primarily as productivity tools rather than decision-making engines (McKinsey, 2024). The speed of adoption reflects how easily these applications fit into existing workflows.

Alongside this, more established machine learning techniques continue to play an important role, particularly in quantitative analysis. These models are used to identify patterns in large datasets, support portfolio construction, and refine risk management processes. Their influence, however, tends to remain in the background. Interview evidence suggests that even when these tools are embedded in the process, they rarely replace discretionary judgement. Instead, they operate as an additional layer of input, informing decisions rather than determining them.

A different set of capabilities emerges when looking at how firms handle unstructured information. Natural language processing has become increasingly important in this context, as it enables the
extraction of usable insights from sources such as news, company filings, and transcripts. The value here lies in scale. Information that would previously have required extensive manual review can now be
processed more quickly and consistently. Research indicates that incorporating alternative data alongside traditional financial inputs can improve forecasting accuracy (CFA Institute, 2025), although the practical impact depends on how these signals are interpreted.

More recently, attention has begun to shift towards systems that can do more than respond to prompts. So-called agentic AI introduces the possibility of automating multi-step tasks, where the system can gather information, process it, and produce outputs with limited human intervention. While still at an early stage, these applications are being explored in areas such as compliance and operational workflows, where processes are more structured and easier to standardise (Deloitte, 2025).

Viewed together, these developments point to a broad expansion in what AI can do across the asset management value chain. The limiting factor is less about capability and more about implementation. The extent to which these tools have an impact depends on how they are integrated into existing systems, and how they are used alongside, rather than instead of, human judgement.

Current AI Adoption Across the Value Chain

AI adoption varies significantly across different functions within asset management. A useful distinction can be drawn between front office, middle office, and back office activities.

In the front office, AI is primarily used to support research and idea generation. Analysts use it to summarise large volumes of information, generate initial drafts of reports, and explore potential
investment themes. Interview evidence indicates that these applications can significantly reduce the time required for early-stage analysis. However, they rarely extend into final decision-making. Portfolio construction remains largely discretionary, with AI outputs treated as inputs rather than instructions.

Client-facing roles show a similar pattern. AI is widely used to draft emails, summarise webinars, and extract key points from longer materials. These tasks are repetitive and time-consuming, making them well suited to automation. At the same time, relationship management remains a distinctly human activity. Interviewees consistently emphasised that clients are not simply buying performance, but also the judgement and credibility of the manager.

In the middle and back office, AI adoption is often more advanced. Operational processes such as compliance monitoring, reporting, and data reconciliation are particularly amenable to automation.
Industry research suggests that operational efficiency has become a primary focus of AI investment, reflecting the potential for measurable cost savings (Alpha FMC, 2025).

Despite this progress, adoption remains uneven across firms. Large global institutions have invested heavily in proprietary platforms, integrating AI into multiple workflows. In contrast, smaller or more specialised firms often take a more cautious approach. Some rely on AI primarily as an enhanced search tool, while others question its ability to generate differentiated insight.

This divergence reflects differences in resources, strategy, and investment philosophy. It also highlights the absence of a single, dominant model of AI adoption within the industry.

Where AI is Creating Value

The most visible impact of AI is not in headline investment performance, but in how quickly routine work can be completed. Across interviews, the same pattern appears in slightly different forms. Tasks that previously absorbed a meaningful part of the day, reading through research, pulling together notes, drafting communications, are now often reduced to a first pass generated in seconds. The change is not subtle. Several interviewees described a shift from building outputs from scratch to editing and refining something that already exists.

This is particularly evident in communication-heavy roles. Client-facing professionals are using AI to condense long-form material into a handful of usable points, or to rework emails into a more polished format without starting over. In one case, a webinar transcript that would previously have required manual summarisation was reduced to a short set of client-ready highlights. The underlying task has not changed, but the time required to complete it has been materially reduced.

A similar dynamic is visible on the investment side, although the implications are slightly different. AI allows analysts to move through information more quickly, whether that is earnings commentary, market news, or internal research. The benefit here is not simply speed, but coverage. With less time spent on initial processing, there is scope to look at a broader set of companies or themes. One interviewee noted that this could allow analysts to expand the range of stocks they follow, although the extent to which this translates into better decisions remains unclear.

What is noticeable is where the impact begins to taper off. AI is consistently used at the start of a task, where the objective is to organise or compress information. Its role becomes less certain once interpretation is required. Outputs are often good enough to work from, but not strong enough to rely on without revision. This creates a pattern where AI accelerates the early stages of work but does not remove the need for judgement at later stages.

There are also early signs of value in operational contexts, although these tend to be less visible. Processes such as reporting, compliance checks, and internal data handling are gradually being automated or streamlined. In some firms, AI is being used to review large datasets or produce draft reports that would previously have required manual assembly. The gains here are more about consistency and scale than speed alone, particularly where processes are repeated across large volumes of data.

What emerges from this is not a picture of transformation, but of compression. Work is being done faster, and in some cases by fewer people, but the structure of the process remains largely intact. The distinction between low-value and high-value activity becomes sharper. Tasks that are repetitive or formulaic are increasingly absorbed by AI, while tasks that require interpretation, judgement, or interaction remain largely unchanged.

This helps explain why the impact, while noticeable, is often described as incremental. AI is not replacing the core of the investment process. It is reducing the effort required to reach the point where that process begins.

Limitations and Trust Constraints

The constraints on AI in asset management are not abstract. They are visible in how professionals choose to use, or avoid, these tools in practice.

Reliability is the most immediate issue. Interviewees did not describe occasional minor errors, but situations where outputs were confidently wrong. One example involved a query on central bank rate decisions that returned answers for the wrong year, despite appearing plausible on first reading. In isolation this is not unusual, but in an investment context it creates a practical problem. Outputs cannot be taken at face value, even for relatively straightforward tasks. This introduces a layer of verification that limits the extent to which AI can replace existing workflows. Time is saved in drafting or summarising, but part of that gain is lost if the result must be checked line by line.

A related constraint is that AI performs unevenly when context matters. It can process structured information effectively, but struggles where interpretation depends on tacit knowledge. Several
interviewees pointed to areas such as management quality, credibility, or the tone of a meeting, where judgement is formed through experience rather than explicit data. These are not edge cases. They sit at the centre of many investment decisions. As a result, AI tends to be used at the beginning of the process, to
organise information, rather than at the point where conclusions are formed.

There is also a noticeable pattern in the type of output produced. AI-generated material is often coherent and well-structured, but rarely distinctive. One fund manager described it as reinforcing “accepted positions” rather than challenging them. This matters more than it might initially appear. In strategies that depend on identifying mispriced assets or taking positions that differ from the market, consensus-level analysis is of limited value. If widely adopted, these tools risk narrowing the range of views rather than
expanding it.

Trust is further complicated by how the tools are used internally. Less experienced staff are often more willing to rely on AI outputs, particularly when those outputs are presented fluently. Several interviewees expressed concern that this can mask underlying errors. The issue is not simply accuracy, but judgement. If the user lacks the experience to question the output, the presence of AI can give a false sense of confidence rather than improving decision quality.

Taken together, these limitations help explain why AI has been absorbed into some parts of the workflow but not others. It is well suited to tasks where speed and scale matter more than precision, such as summarisation, drafting, and information gathering. Its role becomes more constrained as the task moves closer to decision-making. This is not due to resistance from practitioners, but to a recognition that, at present, the technology does not consistently meet the level of reliability and contextual understanding required.

Barriers to Adoption

The constraints on AI adoption in asset management are not primarily technical. In most cases, the limiting factors relate to how the technology interacts with existing data, processes, and risk frameworks.

Data security emerges as the most immediate and consistently cited constraint. Across interviews, there was a clear reluctance to use open or external AI systems where internal information could be exposed. Firms instead favour enterprise environments, often built on external models but deployed within closed networks. This is not simply a preference for caution. It materially shapes how AI can be used. In several cases, interviewees noted that they were unable to upload internal research or client materials into otherwise useful tools, significantly narrowing the range of practical applications.

Data quality presents a more structural limitation. While AI models are capable of processing large volumes of information, their effectiveness depends on how that information is organised. In practice, asset managers typically operate across fragmented systems, with data spread across asset classes, legacy platforms, and teams. Industry research identifies data quality and integration as leading barriers to deployment, but the interviews suggest a more specific issue. AI tends to deliver the greatest value where datasets are already complete and well-structured. Where this is not the case, the technology often adds complexity rather than reducing it.

Reliability and trust form a further constraint, particularly in front-office contexts. Several interviewees highlighted instances where AI outputs were factually incorrect or lacked sufficient context. In one case, an incorrect response to a relatively simple macroeconomic query raised concerns about the need for constant verification. This creates a trade-off. If outputs must be checked at every stage, the efficiency gains associated with automation are partially offset. As a result, AI is often used for drafting and summarisation, but avoided in areas where precision is critical.

Regulation reinforces these limitations. Asset managers are required to justify and explain investment decisions, particularly where client capital is involved. This creates friction when using models that cannot easily be audited or interpreted. Regulatory expectations around transparency and model risk do not prevent AI adoption, but they do slow it, particularly in areas such as portfolio construction where accountability is central.

Organisational factors are equally significant. The challenge is less about access to tools and more about how they are used. In some firms, multiple AI systems have been introduced in parallel, leading to fragmentation rather than integration. In others, adoption is uneven, with usage depending on individual initiative rather than firm-wide processes. Training also remains inconsistent. Interview evidence suggests that even where tools are available, employees are not always clear on how to apply them effectively within their specific workflows.

These constraints point to a more grounded explanation for the pace of adoption. The issue is not that AI lacks capability, but that its effective use depends on conditions that are unevenly met across the industry. Where data is structured, workflows are clearly defined, and governance is established, adoption can be relatively rapid. Where these conditions are absent, progress is slower and more limited in scope.

Industry Implications

The immediate effect of AI adoption is not a transformation in investment outcomes, but a compression of how investment work is carried out. Tasks that previously required significant time and junior resource, including first-pass research, drafting, and data gathering, are increasingly standardised and accelerated. Interview evidence consistently highlights the use of AI in summarisation, information aggregation, and early-stage analysis, suggesting that a growing share of baseline output is becoming uniform across firms.

This has direct implications for competition. If firms rely on similar tools, often trained on overlapping datasets, improvements in processing speed are unlikely to generate a sustained advantage. As these capabilities become widely available, the marginal benefit of faster information processing diminishes. Differentiation shifts instead towards how information is interpreted and applied in decision-making.

There is, however, a tension between efficiency and differentiation. While AI can standardise elements of the research process, outperformance still depends on forming views that depart from consensus. One interviewee operating a concentrated, contrarian strategy noted that AI-generated outputs often reflect prevailing market narratives rather than challenging them. If this pattern is widespread, greater reliance on AI may contribute to more crowded positioning rather than improved investment outcomes.

The effects are unlikely to be uniform across the industry. Larger institutions are better positioned to build proprietary systems and integrate AI across workflows, benefiting from scale and data access. At the same time, reduced costs of information processing may allow smaller firms to compete more effectively in areas where scale was previously an advantage. This suggests a more heterogeneous competitive landscape rather than a simple shift towards consolidation.

The implications for talent are more complex. Many entry-level tasks are being partially automated, including activities that traditionally formed part of the training process for junior analysts. Interview evidence raises concerns that early reliance on AI may limit the development of independent judgement, particularly where outputs are accepted without sufficient validation. Over time, this could affect the depth of expertise within the industry.

AI could be reducing the value of routine information processing while increasing the importance of interpretation, judgement, and decision-making under uncertainty. Firms that rely primarily on AI to improve efficiency may find that their outputs converge with those of peers. Sustained differentiation is more likely to depend on how effectively firms combine these tools with independent analysis and investment conviction.

Future Outlook and Conclusion

Over the next few years, the most meaningful impact of AI in asset management is unlikely to come from fully automated investment decisions. Instead, it will emerge in less visible but more pervasive ways, reshaping how research is conducted, how information is processed, and how time is allocated across the investment process.

The evidence in this report suggests that the core of investment decision-making will remain human for the foreseeable future. Portfolio managers are not simply reluctant to hand over control; in many cases, AI systems are not yet capable of handling the ambiguity, context, and judgement required. Several interviewees pointed out that even when AI produces technically correct outputs, it often fails to capture nuance or generate genuinely differentiated views. As a result, its role is still bounded. It accelerates analysis, but it does not replace conviction.

What is changing more quickly is the structure of the work itself. Tasks that once defined junior roles, including first-pass research, drafting, and data gathering, are already being absorbed by AI tools. This creates an uncomfortable tension. Firms benefit from efficiency gains, yet risk weakening the training pipeline that produces experienced investors. One interviewee explicitly raised this concern, noting that over-reliance on AI may leave junior staff less able to identify errors or challenge outputs.

At the same time, the widespread availability of similar AI tools introduces a more subtle shift in competition. If most firms can access comparable models, trained on broadly similar data, the informational edge that once differentiated managers begins to erode. In that environment, simply producing analysis more quickly is unlikely to be enough. The differentiator becomes how that analysis is interpreted, where judgement is applied, and whether the resulting decisions depart meaningfully from consensus.

This has implications for industry structure. A plausible outcome is a divergence between firms that lean heavily into AI-driven efficiency and those that position themselves around high-conviction, human-led strategies. Rather than converging on a single model, the industry may fragment, with different approaches appealing to different client preferences.

For now, AI is best understood as a force that compresses the lower end of the value chain. It makes routine work faster, cheaper, and more standardised. What it does not yet do, and may struggle to do in a meaningful way, is replace the elements of investing that rely on judgement under uncertainty. That distinction matters. It suggests that the long-term impact of AI will be less about removing humans from the process, and more about narrowing the space in which human skill actually adds value.

Primary research/ interviews

All interviews have been anonymised due to confidentiality considerations.

• Interview with Client Director, Investment Trusts, 10 April 2026.
• Interview with former Senior Client Portfolio Manager, 11 April 2026.
• Interview with owner of investment management firm, 13 April 2026.
• Interview with Vice President in Client Advisory and Investment Trust Sales, 17 April 2026.
• Interview with Director of investment management firm, 20 April 2026.

References:

Alpha FMC, Global Operations Survey 2025.
Amundi Investment Institute, AI Investment Research, 2025.
BattleFin / Exabel, Alternative Data Buyside Insights & Trends 2025.
CFA Institute, AI and Investment Professionals Survey, 2024.
CFA Institute, Alternative Data and Forecasting in Asset Management, 2025.
Deloitte, Tech Trends for Investment Management 2025.
Deloitte, AI Adoption Challenges Survey, 2025.
Grant Thornton / ThoughtLab, The AI-Powered Investment Firm, Q3 2025.
IOSCO, Consultation Report on Artificial Intelligence in Financial Services, 2025.
McKinsey & Company, How AI Could Reshape the Economics of the Asset Management Industry, July
2025.
McKinsey & Company, State of AI 2024.
Mercer Investments, AI Integration Survey, April 2024.
PwC, AI Risk & Compliance, 2025.
Traders Magazine, Preparing Capital Markets for the Age of Agentic AI, October 2025.
Verified Market Research, AI in Asset Management Market Analysis, 2025.