Sunday, 6 September 2026
Artificial intelligence is beginning to alter one of the most established parts of the investment industry: fundamental equity research. What started with machine learning and automated document analysis has progressed through generative AI into a new generation of agents capable of gathering information, updating models, filtering news and supporting large parts of the investment research workflow. The result may not be the replacement of portfolio managers, but a fundamental change in where investors can still find an advantage over their competitors. That was one of the central conclusions from the AI4 2026 discussion “From Autocomplete to Agents: The Evolution of AI in Fundamental Investing,” moderated by John Divine, Assistant Managing Editor of Investing at U.S. News & World Report, with a senior investment research executive from Wellington Management.
The development can broadly be divided into three stages. Before the arrival of widely accessible generative AI, sophisticated machine-learning techniques within investment organisations were largely controlled by quantitative researchers, engineers and data scientists. Natural-language processing could already be used to examine sentiment, analyse filings and process large datasets, but the technology remained largely invisible to traditional fundamental investors. Analysts generally consumed the resulting research without necessarily building or operating the systems themselves.
Generative AI changed that relationship. When conversational models became widely available, fundamental investors suddenly had something resembling an on-demand research assistant. Public filings could be summarised, documents compared and initial drafts of investment research prepared far more quickly. Tasks that previously consumed hours could sometimes be reduced to minutes. The limitations were immediately apparent, however. Models could invent information, lacked sufficient knowledge of an investment firm’s internal research and frequently struggled to retain the context required for longer analytical processes. They were useful assistants but not reliable investment colleagues.
Agentic AI represents the next stage because the technology is beginning to interact with the workflow rather than simply responding to individual questions. Agents can potentially retrieve regulatory filings, examine earnings information, work with financial models, monitor news, generate screens and repeatedly update their analysis as new information appears. The significance is not simply that research becomes faster. It is that AI could compress the distance between asking an investment question and obtaining the evidence required to investigate it.
A fundamental analyst covering hundreds of companies normally spends substantial amounts of time collecting information before making a judgement. AI agents could increasingly perform much of that preparatory work, allowing the analyst to concentrate on determining whether the information actually changes the investment thesis. That could alter the economics of research departments, with the investment professional’s time gradually moving away from collecting information and towards evaluating it.
The change is already affecting who can build investment technology. Earlier machine-learning systems generally required significant programming expertise. New AI coding tools increasingly allow professionals without traditional software backgrounds to construct basic dashboards, screening systems and analytical applications by describing what they want the technology to produce. This could narrow the historical divide between investment professionals and technology teams, allowing analysts who understand companies and markets to customise more of their own research infrastructure rather than relying entirely on dedicated engineering teams.
The longer-term vision goes significantly further. Agents could ultimately participate across almost the entire investment chain, including research, trading, risk management, portfolio construction and client reporting. Routine research could operate continuously in the background, with systems identifying developments that require human attention rather than analysts manually monitoring every source. Full automation of fundamental investment decisions, however, appears much less likely.
The technological question of whether an AI system can technically select securities is different from whether an asset manager should delegate responsibility for investment decisions to it. Asset managers have fiduciary responsibilities towards clients, and accountability becomes difficult when an algorithm makes a decision that produces substantial losses. Investment managers can question a portfolio manager about why an investment failed. The same accountability becomes considerably more complicated when the explanation is that an autonomous system selected the security.
Human involvement therefore remains important not simply because current AI technology has limitations but because investment management requires responsibility, judgement and explanation. Perhaps the more important question is what happens to investment alpha once every major manager has access to similarly powerful AI.
Many traditional sources of investment advantage have historically depended partly on speed. An analyst able to examine new information quickly, understand its implications, update a financial model and reach an investment conclusion ahead of competitors could potentially benefit from that information advantage. AI threatens to compress that advantage dramatically.
If agents can read the same corporate filing, update models and identify the important changes within seconds, the ability to process publicly available information quickly becomes much less distinctive. What could be described as process alpha, or investment advantage created primarily through superior information processing, may consequently become smaller. This has happened before. Technologies including spreadsheets, financial terminals, electronic communications and quantitative databases progressively made capabilities that were once specialist advantages available across the investment industry. AI could represent a substantially larger version of the same phenomenon.
Alpha would not necessarily disappear. Its location could move. As information processing becomes increasingly automated, proprietary information and differentiated judgement may become more important. An investment firm’s internal datasets, access to company management, understanding of industries and ability to identify changes in corporate behaviour could become more valuable precisely because publicly available information is becoming easier for everyone to process.
Understanding management quality provides a useful example. Financial statements can reveal margins, cash generation and balance-sheet conditions, but determining whether executives are credible, whether corporate culture is deteriorating or whether management has genuinely changed strategy often depends on interpretation rather than calculation. AI can contribute evidence, but human experience, intuition and emotional intelligence may remain significant when assessing situations that are difficult to quantify.
The same applies to identifying turning points. Markets frequently move around changes in expectations rather than existing conditions. Recognising when an industry’s economics are about to change can require combining incomplete evidence with experience and judgement. This means the human investor could become more important in some areas even as machines perform more of the research process.
The shift could also push investment horizons further into the future. If AI dramatically accelerates the market’s ability to process immediate news, competing over information released today becomes increasingly difficult. Investors may instead need to concentrate on questions that cannot easily be resolved from existing information, including how businesses, industries and economies could develop several years ahead. That does not automatically create an advantage, because if every investor moves towards longer-term forecasting, competition simply shifts there as well. Nevertheless, it illustrates how technology can redistribute investment opportunities rather than eliminate them.
Investment firms are also beginning to consider whether their own research methods should become proprietary AI systems. As third-party platforms offer increasingly sophisticated investment research tools, many asset managers could end up using similar data, similar models and similar workflows. That creates a new risk: technology designed to increase investment differentiation could inadvertently make investment processes more alike.
Managers may therefore seek to codify their own research philosophies, analytical frameworks and investment rules into internal AI environments. Instead of asking a generic system to analyse a company, the AI could examine the company according to the particular questions, valuation disciplines and risk criteria used by that investment team. The technology then becomes a mechanism for scaling a firm’s investment philosophy rather than replacing it.
This development could have significant consequences for recruitment. Only a few years ago, investment teams building advanced data capabilities placed considerable value on people with exceptional programming, statistical and engineering skills. Those abilities remain important, but AI coding tools are beginning to reduce the scarcity value of some technical skills. Curiosity, communication ability, investment judgement and the capacity to generate differentiated questions may therefore become relatively more important.
The shift could be described as moving some of the emphasis from technical intelligence towards human judgement. If AI can increasingly write software, process datasets and construct analytical tools, the more valuable employee may be the person who knows which questions the technology should be answering.
New hybrid positions are also emerging inside investment organisations. These professionals sit between conventional investment analysts and technology engineers. They understand the investment process sufficiently well to know what analysts require while also understanding AI sufficiently well to construct prompts, agents, libraries and workflows around those requirements. Such roles could effectively become the architects of an investment team’s AI infrastructure.
Traditional junior analyst positions may face a more complicated future. Much of the work historically used to train young investment professionals involves precisely the activities AI is becoming capable of automating: updating spreadsheets, reading company reports, gathering news, preparing comparable-company analysis and maintaining financial models. If those tasks disappear, firms will have to reconsider how inexperienced analysts develop the judgement expected from senior investors later in their careers.
The same question is emerging across other professional industries. Removing repetitive work can increase productivity, but repetitive work has also historically functioned as training. Investment firms may therefore need more deliberate methods of teaching company analysis, financial reasoning and portfolio judgement if junior employees no longer acquire those abilities by completing thousands of basic research tasks.
AI is already particularly useful in helping investment teams manage the enormous amount of information surrounding financial markets. Analysts increasingly face corporate announcements, economic data, regulatory filings, news, alternative datasets and social-media information simultaneously. The problem is no longer simply obtaining information but determining what deserves attention.
Agents can help separate potential signals from background noise and narrow very large investment universes to a smaller group of opportunities for human analysis. They can also be used to challenge existing investment positions. Rather than merely searching for evidence supporting a portfolio manager’s view, an AI system can be instructed to identify weaknesses in the thesis, examine alternative scenarios or search for evidence suggesting that the investment team may be wrong. That could make AI particularly valuable as a research adversary.
Investment professionals inevitably develop assumptions from previous experience. Artificial intelligence can rapidly search historical information and alternative interpretations, potentially exposing evidence that does not fit those assumptions. Used correctly, the technology may therefore improve investment decision-making not by providing the answer but by forcing investors to consider questions they might otherwise overlook.
AI itself has biases, however, and its conclusions remain influenced by its training data and system design. Investment managers therefore cannot treat machine-generated analysis as an independent source of objective truth. The more effective model is likely to be a combination. Humans contribute experience, market context and accountability, while AI supplies speed, breadth of information and the ability to test large numbers of possibilities.
This partnership could ultimately redefine fundamental investing. The first stage of financial AI helped specialists analyse data. The second gave almost every investment professional a digital assistant. The emerging third stage is beginning to insert intelligent agents directly into investment workflows. The next question is no longer whether those tools can save analysts time. It is what investment professionals should do with the time that remains once much of the mechanical research process has been automated.
For active managers, that question goes directly to the future of their business. If every firm can process public information almost instantly, superior technology alone will not guarantee superior returns. Competitive advantage will increasingly depend on proprietary knowledge, differentiated research processes, access to companies, longer-term thinking and the quality of human judgement applied to the evidence AI produces. In that environment, AI may commoditise some of the old sources of investment alpha while simultaneously making the genuinely difficult parts of investing more valuable.
Source: CIJ.World Research & Analysis Team
