Is AI an alpha engine?

Technology has long helped asset managers behind the scenes. It speeds up compliance work, automates reporting and, when used correctly, reduces operational risk, in theory freeing staff to focus on higher-value tasks.

But the front office is a different story: it remains a human territory with portfolio managers deciding strategy, selecting securities and negotiating trades. AI has been a useful auxiliary, but never a strategic partner.

There are signs that this mindset is beginning to shift. Some firms are placing AI deeper into core investment workflows, using it to help with traditionally-human roles like idea generation and portfolio commentary. 

In other words, early adopters treat AI as a partner for alpha, seeking new sources of return and improving decision speed, plus responding quickly and efficiently to sudden changes in complex markets. 

This then begs the question: has AI become a strategic partner in investment decision-making? If it can reliably generate alpha, then the answer is “yes” and early movers should be able to gain structural advantages. If “no,” then the challenges are obvious: over-investment, model failure and missed growth opportunities, as well as considerations related to data integrity, regulatory compliance and business planning for the future of work.  

Evidence suggests that while AI is not a guaranteed alpha engine, it has been a successful aide to performance generation for some managers. 

Generation alpha?

Some asset managers use reasoning models and generative AI in front-office workflows. This is not simply to summarize data, instead they have given algorithms the authority to suggest trades, for example.  

Gary Collier, CTO at Man Group, told Bloomberg’s Investment Management Summit: “We’ve got AI agents running…that can and have come up with their own independent alpha-generative ideas. We still have a human investment committee that will assess them, but in terms of the creative side of the process – idea generation and taking the idea and coding it up – that’s one of the most exciting spaces right now.” 

AI is moving from the realm of sub-editor to co-creator, identifying ideas and correlations humans might miss in vast datasets, however, human oversight is still essential (vetting signals, adjusting risk, validating models, making investment decisions). 

Research analysts spend more time validating AI-generated leads and less time originating them from scratch, while portfolio construction teams use AI for scenario analysis, stress testing and risk attribution. 

Intraday bond valuations show the potential on offer: AI can use millions of data points to price instruments faster than humans, enabling more informed decision-making. 

Jamie Ovenden, CTO at Schroders, explained the approach to workflows: “We deconstruct investor and researcher roles into component parts and focus on where you can have the most impact with GenAI, for example, in portfolio construction’s many steps. This all adds up to an incremental gain, enabling people to sit on larger information sets and be more effective in their roles.” 

Some caution is warranted. AI deployments are costly and data quality can be inconsistent, not to mention models that perform well in testing may fail in production. 

Collier said businesses should keep their feet on the ground: “There’s a tension between the dream of agentic AI and the need for predictable, deterministic workflows. In many cases, a traditional workflow may be more appropriate.”

Building an AI-ready business

In order for AI to generate alpha, firms should focus on strong tech foundations: high-quality data, scalable infrastructure and – perhaps most important of all – rigorous governance. Investment in these areas is non-negotiable. 

Data readiness is fundamental; firms need well-documented metadata, clear data lineage, and trust grading. Ovenden said Schroders is, “…consolidating to improve predictability and traceability.”  

Amanda Stent, head of AI Strategy & Research in Bloomberg’s Office of the CTO, highlighted a further challenge: “I don’t think many organizations have grappled with how AI agents can access data at speed and scale that humans cannot.” 

Stent outlined how AI is evolving from traditional AI that provides data and signals, as with intraday bond valuation, to AI that assists with tasks, like drafting portfolio commentary, attribution and risk reporting, as well as querying baskets of documents, such as analysts’ reports, over time. These capabilities accelerate decision-making and enable hypothesis testing on a much faster scale. 

Despite the hype, AI projects must not lose sight of their owner’s fiduciary duty. Collier recommended a pharma-style approach, with a hypothesis followed by a period of testing and then scale, all with robust guardrails.  

Ovenden emphasized that governance structures like committees and accountability frameworks are highly relevant in this process, with added education needed on AI-specific risks. Transparency with clients about AI use is becoming standard practice, while regulatory scrutiny continues to increase globally. 

Despite the promise of AI, ROI is unclear – possibly it’s too soon to gauge its impact in money terms. Indeed, immediate P&L impact may be limited, but efficiency gains and productivity improvements no doubt create value over time.  

Collier talked about the likely consequences of this for AI pick-up: broad experimentation with early adopters, followed by high-conviction scaling in areas demonstrating consistent differential value.  

Stent added that successful AI deployment should unlock entirely new products or workflows that generate revenue, not just create cost savings or efficiencies.

Asset Manager 2030: human, machine, a bit of both? 

Looking ahead, there is scope for asset managers to operate increasingly like tech companies. Front offices might choose to run AI-native workflows with integrated pods of quants, data engineers, portfolio managers and oversight specialists. Some boutique shops might go machine-first, while larger firms could use AI augmentation more selectively. 

Roles will change and routine analytical work will decline, while interpretation, oversight and strategic thinking rise up the skills agenda. As Ovenden said: “Investors will need to become pilots of big AI systems. They’ll learn to interact with them to specify designs, requirements, and outcomes, while the AI is actually doing the work.” 

Collier noted that better-than-human reasoning could impact the structure of front offices in profound ways, though humans will still be critical for deterministic decision-making and governance. 

Private markets pose particular challenges to the march of AI, presenting a potential oasis for organic intelligence. Data is less standardized, less accessible and risk assessments lean more towards human judgement – for now at least.  

Stent said: “We’re seeing a flight from publicly traded assets into private equity and private credit, and it’s much harder for AI to get access to that kind of data at scale. That’s where the humans are moving.”  

By 2030, the tipping point will likely be more obvious than it is today. As always, hindsight will be clear and firms that invested early in infrastructure, governance and AI-ready culture may outperform their peers who hesitated five years before.  

As in other industries, humans will retain their edge in strategy and creative judgement, plus in handling opaque or hard-to-reach datasets. But AI will likely become a valued collaborator in alpha generation – whatever form that takes.  

AI is already contributing to this cause in asset management. It is generating ideas, improving research, accelerating decision-making and refining risk assessment. For some firms, the gains are clear and obvious; for others, they are less so. 

Perhaps the differentiator is more about what goes in than what comes out. Clean data, reliable infrastructure, strong governance, and patient, disciplined implementation give asset managers the best chance of winning the AI race. Failure in this regard risks model error, bias, hallucination and, as a result, regulatory heat. 

On balance, the evidence leans towards AI being an alpha engine for managers who deploy it well. Thoughtful, strategic adoption today may well provide durable competitive advantage. Only time will tell.