GenAI is changing investment processes faster than most firms can adapt. The release of Claude for Financial Services is the most recent step in the applying of GenAI to the investment industry. Its concentrate on domain knowledge and specialized workflows sets it aside from general frontier LLMs and raises vital questions on how financial workflows will evolve, how tasks might be divided between humans and machines, and what skills might be required to reach the long run of finance.
Financial firms are facing essentially the most profound overhaul of their technology capabilities in a generation. AI-driven digital transformation is changing work roles and investment processes, causing professionals to rethink the boundaries between human and machine cognition as firms work to enhance their technology stacks and human capital to stay competitive.
Amid this transformation, firms and professionals must re-evaluate the talents needed for fulfillment. Given the pace of technological advancement and uncertainty about transition paths, predicting how AI will transform workflows and job roles is difficult. Nevertheless, this assessment is needed for strategic planning for each industry leaders and individuals considering their profession path.
As we have now noted elsewhere, we consider the long run might be defined by the complementary cognitive capabilities of humans and machines, characterised by the “AI + HI” paradigm and the continued importance of skilled competence. To understand what this mixture looks like, it’s needed to first assess the present level of AI adoption in investment operations before identifying possible transition paths to future scenarios characterised by different mixes of human and machine interaction.
Current landscape
A key insight from this work is that investment professionals engage in a multihoming strategy, where they use multiple platforms and/or technologies to finish a task. In the Analytical job role category, three example workflows—assessment, industry and company evaluation, and research report generation—illustrate this pattern.
The table shows the proportion of respondents who use different technologies for every of those tasks. Not surprisingly, traditional tools like Excel and market databases remain essentially the most commonly used, but respondents also report integrating tools like Python and GenAI alongside traditional software. For example, while 90% of respondents reported using Excel for assessment tasks, 20% reported also using Python on this workflow. For analytical tasks, GenAI was mostly used to help within the creation of research reports, cited by 27% of respondents.[3]
GenAI in practice: A workflow example
Let’s consider conducting industry and company evaluation, where on the time of conducting our survey in 2024, 16% of respondents reported using GenAI on this workflow. Our Automation ahead Content series, within the episode RAG for Finance: Automating document evaluation with LLMsis a concrete example of how GenAI can improve this workflow.
The case study is complemented by Python notebooks in our RPC Labs GitHub repository. It shows how RAG can extract executive compensation and governance details from corporate proxy statements of portfolio firms and present the ends in a structured table – certainly one of several tasks performed on this workflow.
Such a task is traditionally manual and time-consuming, with the hassle required largely depending on the variety of portfolio holdings. With GenAI, the method may be scaled efficiently with little additional computational effort, freeing the analyst from manual data extraction and preparation of a tabular comparison.
Since the tasks of knowledge extraction and data presentation are outsourced to the GenAI model, the analyst can concentrate on interpreting the info reasonably than preparing it. Instead of analyzing the numbers, the analyst focuses on evaluating the outcomes by interrogating the model, checking the validity of the info, understanding the restrictions of the evaluation, correcting errors, and supplementing the outcomes with additional information or insights from other sources – all with the goal of identifying potential governance risks for all portfolio holdings.
Far from eliminating the necessity for a human analyst, this instance shows how greater value may be derived from human input by providing more time and capability for critical considering and decision-making. It also highlights the restrictions of AI (such tasks have insufficient accuracy rankings) and the continued need for human oversight and judgment.
evolution
Agentic AI has emerged as a strong tool that may further improve workflows and deepen human-machine interaction. These tools construct on a few of RAG’s limitations and incorporate thought chains and external function calls (see our article “Agentic AI for Finance: Workflows, Tips, and Case Studies“).AI agents expand the scope of tasks that machines can perform and may shape the long run direction of human-machine interaction.
In some ways, this development simply expands the multihoming strategy, combining multiple tools and platforms right into a single interface. Claude for Financial Services reflects this approach, connecting to market databases and traditional platforms resembling Excel to supply reports and evaluation for the user. In this manner, AI acts as an application layer on top of other software tools, providing an interface to the human analyst who maintains oversight and responsibility.
Professional judgment stays essential to check assumptions and validate data sources and references. In addition, the effective use of those tools also will depend on a sound fundamental knowledge of finance and investments, which allows analysts to trust and own model results and supply a sound basis for investment decisions.
Professionals also need soft skills that can not be outsourced to machines, including constructing relationships and exercising duties of loyalty, prudence and care based on ethical values.
[1] Our research inventory on AI includes:
AI in Asset Management: Tools, Applications and Limitations
AI pioneers in investment management (2019)
T-Shaped Teams: Organizing to Adopt AI and Big Data at Investment Firms (2021)
Ethics and Artificial Intelligence in Investment Management: A Framework for Professionals (2022)
Handbook for artificial intelligence and massive data applications in investing (2023)
Unstructured Data and AI: Fine-tuning LLMs to Improve the Investment Process (2024)
AI in Investment Management: Ethics Case Study (2024); AI in Investment Management: Ethics Case Study Part II (2024)
Creating value from big data within the investment management process: A workflow evaluation (2025)
Synthetic data in investment management (2025)
Explainable AI in Finance: Addressing the needs of diverse stakeholders (2025)
Automation Ahead: Content Series (2025)
