---
title: "The agentic investor: What the new capital allocator actually looks like · ARC"
description: "The decision to buy your stock is still mostly made by a human. The decision to look at it increasingly is not. A taxonomy of the 2026 investor base, and what it means for how you communicate."
canonical: "https://arc.silba.ai/insights/agentic-investor-capital-allocator/"
markdown: "https://arc.silba.ai/insights/agentic-investor-capital-allocator/index.md"
author: "Rachel Carroll"
published: "2026-10-13"
modified: "2026-10-20"
language: "en"
---

# The agentic investor: What the new capital allocator actually looks like

The portfolio manager reading a research note over coffee still exists and still matters. But that investor is no longer the first to find your company or the fastest to act on what you publish. This piece maps the four segments that now make up the investor base, from fully automated quantitative readers to the AI assistant on a retail platform, and shows why one communications program serves all of them, if it is built to pass the machine first.

Rachel Carroll

- Number Insight 03
- Published 13 Oct 2026
- Reading 8 min
- Sources 10, all public
- Series Capital markets

When most IR professionals picture their target investor, they see a portfolio manager at a long-only institution reading a research note over coffee. That investor still exists. They still matter. But they are no longer the first to find your company, the fastest to act on new information, or, day to day, the most influential force on your share price.

It is worth being precise about what has changed, because the imprecise version is easy to dismiss. In June 2026, [Morningstar Manager Research](https://www.morningstar.com/en-gb/business/insights/research/AI-in-active-fund-management) reported that fewer than 20 asset managers show meaningful AI integration in how they build portfolios, against near-universal claims of AI adoption. Anyone who tells you the buy-side has handed its decisions to machines is overstating it.

What has changed is not who decides. It is what reaches the decider. Sixty-eight percent of US institutional investors in the [Center for Audit Quality's July 2026 survey](https://thecaq.wpenginepowered.com/wp-content/uploads/2026/07/caq_institutional-investor-survey-ai_2026-07.pdf) report extensive or moderate use of AI in research; 60% run earnings transcripts through it routinely; 45% use it for investment screening. Seventy percent of large buy-side firms in [SimCorp's January 2026 survey](https://www.deutsche-boerse.com/dbg-en/media/news-stories/press-releases/More-than-Two-thirds-of-Investment-Managers-Prominently-Using-AI-to-Support-Front-Office-SimCorp-Study-Reveals-4910704) have AI in the front office, meaning the portfolio managers, analysts, and traders who make investment decisions. And the peer-reviewed evidence on the other end of the market, [Blankespoor, Croom, and Grant's study](https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5053905), forthcoming in the Journal of Accounting and Economics, pairs more than 400,000 investor queries to a brokerage AI chatbot with a survey of over 2,000 retail investors. Nearly half of those surveyed have used generative AI, and the query data shows usage maturing from broad screening toward monitoring and interpreting specific company news.

The decision is still human. The filter is not. Most investor relations programs are built to persuade the human and have never been tested against the filter.

## A taxonomy of the modern investor base

Four segments now matter, and they discover, evaluate, and act on your company in four different ways.

**Systematic and quantitative funds.** Fully automated readers. They ingest your release, your transcript, and your filings the moment they publish, score them against history and peers, and trade the signal. Firms such as Man AHL, AQR and the quantitative arms of BlackRock, Goldman Sachs and J.P. Morgan sit here. This is the depth of integration Morningstar found at fewer than twenty managers, and they are the fastest-moving money around any reporting event. You do not persuade them. You either give them a clean signal or a noisy one.

**AI-assisted allocators.** The largest and fastest-growing segment: discretionary institutions whose analysts now start with a model. This is the segment the two largest AI labs built for in 2026: [Anthropic's financial-services agents](https://www.anthropic.com/news/finance-agents), launched in May with its models [already in production at JPMorgan, Goldman Sachs and Citi](https://www.fortune.com/2026/05/05/anthropic-wall-street-financial-services-agents-jamie-dimon/), and [OpenAI's ChatGPT for Financial Services](https://openai.com/index/introducing-chatgpt-financial-services/), launched in September with Morgan Stanley and Evercore as design partners. The human portfolio manager still decides, but decides among the companies and framings the agent surfaced, in the language the agent used to summarize them. If the agent's summary of your company is generic, the manager's attention goes elsewhere before you know you were considered.

**High-agency investors.** Sophisticated individuals and family offices running institutional-grade research on retail platforms with AI built in. In March 2026, 62% of US retail investors [surveyed by Investing.com](https://www.investing.com/news/stock-market-news/survey-nearly-twothirds-of-retail-investors-use-ai-to-inform-market-decisions-4598846) were already using AI for investment decisions. On September 29, 2026, [Robinhood launched Agents](https://robinhood.com/us/en/newsroom/hood-summit-2026/) inside its app, free to use on one OpenAI model until the end of the year, with trade approvals on by default and adjustable, having reported more than 150,000 agentic trading accounts since May. The free tier runs a smaller model than the frontier systems institutions use, which means the retail reader is the weakest reader your material will meet, and the standard it sets is the one that holds. The same study shows this population's usage deepening from screening to company-specific monitoring. They are numerous, fast, and, because they query AI assistants directly, the segment most exposed to whatever your agentic footprint currently says.

**Traditional institutional allocators.** The long-only manager over coffee. Still the anchor of most shareholder registers, still reached through roadshows and sell-side introductions, and still the segment most IR programs are exclusively designed for. Increasingly, though, the research note over coffee was drafted with an AI assistant, and the companies on the meeting list were screened by agents.

## What this means for investor targeting

Traditional targeting (ownership analysis, peer-holder screens, conference calendars, non-deal roadshows) remains legitimate. It is no longer sufficient as the primary mechanism for building a shareholder base, for one structural reason: it is periodic, and agentic discovery is continuous.

A roadshow puts you in front of thirty investors in a week. An agentic research query about your sector runs thousands of times a day, and your company is either surfaced accurately in the answer, surfaced generically, or absent. That outcome is determined not by the meeting you booked but by the quality, consistency, and structure of everything you have published, which is to say by communications engineering rather than by access.

The practical test is simple and I run it for every company I assess: ask an AI research assistant to describe the company as an investment. If the answer is a sector description that could apply to three peers, the filter is working against you, and no amount of roadshow scheduling corrects it.

There is a second-order risk worth naming. Research agents increasingly read each other's output. A simulation study published on arXiv in September 2026 ([WolfSociety, Zhang et al.](https://arxiv.org/abs/2609.05591)) modeled communicating financial agents and found that distorted information caused limited damage until it crossed a threshold, after which it propagated sharply, and the threshold fell as the agent population grew. It is a simulation, not a market. But the issuer-level inference is sound: an outdated or inaccurate characterization of your company, once it enters the agent layer, can be repeated across research workflows without anyone independently re-checking it. The correction has to be made at the source, and it has to be made early.

## Engineering for the full spectrum

The good news is that there is one program to build, not four. The disciplines required to perform for agentic systems (precision of language, consistency of narrative, explicit thesis articulation, freshness of content) are the same disciplines that make investor communications compelling to the best human investors. A release that opens with a clear thesis restatement, a website that states the case for the stock in one place, a terminology framework that holds across every channel: these serve the systematic fund, the AI-assisted analyst, the high-agency individual, and the traditional PM equally, because they are simply what excellent investor communication looks like.

What changes is verification. In the traditional model, you learned how your story landed from the questions in the meeting. Now you can learn it before the meeting exists, by testing how the filter represents you, and fixing what it gets wrong.

## The companies that will win

The investor base has changed faster than most IR programs have adapted. The 83% of institutional investors who expect their AI use to rise within two years are describing a filter that gets stronger every quarter. The companies that take this seriously earliest, engineering their communications for the full spectrum of how capital is now allocated and testing the result, will capture a disproportionate share of the flows that determine valuation. The rest will keep optimizing for the reader who arrives last. At the institutional end the decision remains human. At the retail end, execution is beginning to be delegated to agents with a human approval step, and the direction is set.

> The decision is still human. Reaching the human now requires passing the machine.

## ARC tests how AI research tools read your company.

In a workshop, we present our current findings and apply them to your own materials.

[Request a workshop](https://arc.silba.ai/contact/)

## Keep reading

### [The first investor to read your earnings release isn't human](https://arc.silba.ai/insights/earnings-release-agentic-ai/)

### [Why your IR website is the most undervalued asset in your capital markets program](https://arc.silba.ai/insights/ir-website-agentic-optimization/)

[All insights](https://arc.silba.ai/insights/)