"Lobster" Swims Into Private Equity Circle: How Is OpenClaw Being Utilized by Private Funds?

Cailian Press, March 15 — Reporter Wu Yuqi A “Lobster” has recently been a frequent topic in the investment circle.

Some investors jokingly commented on social media that in the past, fund managers discussed AI mainly around computing power, models, and application chains; but this year, more and more people are directly deploying Agents, even private equity research teams are starting to explore how to “raise lobsters.” The so-called “lobster” refers to the open-source AI Agent framework OpenClaw. Its popularity on GitHub has rapidly increased, spreading quickly from the tech enthusiast community to a broader group of financial professionals.

It is not surprising that this enthusiasm has reached private equity firms. Over the past two years, AI in the investment industry has mostly stayed at the level of “observing industries and selecting targets,” but this year, a more noticeable change is that some institutions are beginning to treat Agents as part of their research infrastructure. Even if not yet officially integrated into core processes, they are at least in trial, comparison, and pre-research stages.

According to interviews with Cailian Press, some private equity firms are already internally experimenting with deployment, using Agents with similar functions in research and investment processes—mainly to assist with basic coding, data organization, and other standardized tasks. Some private equity professionals also said their organizations have not deliberately deployed such tools but view them more as observation samples rather than officially integrating them into core workflows at this stage.

A chairman of a private equity firm also tried installing “Lobster,” but remains cautious about integrating such tools into the organization. “Honestly, I don’t think the organization will use it easily. My own experience with ‘raising lobsters’ involved preparing a brand-new laptop, with no investment, trading, or client data—completely isolated, in a pure testing environment. That way, there’s no risk of issues.”

From a broader perspective, this wave of “raising lobsters” is not just a fresh experiment in the tech community. For private equity, it’s more like a touchstone—testing whether AI Agents can move from “chatting” to “getting things done,” while also forcing institutions to re-examine their research processes, data boundaries, and risk controls.

From Research Object to Research Tool, Private Equity Reassesses AI Agents

If last year’s AI trend mainly focused on listed companies, industrial chains, and thematic investments, this year a more noteworthy change is that AI itself is shifting from being a “research subject” to a “tool for use.”

In the private equity industry, this change is not necessarily reflected in large-scale, full integration but more in scattered and cautious exploration. Some start with coding assistance, others use it for data collection, document processing, and information retrieval; some simply see it as an entry point to observe new technological developments.

A private equity professional told Cailian Press that for most private equity firms, AI Agents are not yet the decisive core weapon but rather a “low-risk trial” efficiency tool. They are first placed in peripheral scenarios to see if they can save repetitive work.

This makes sense logically. In private equity research workflows, there are many standardized, fragmented, and decomposable basic tasks—such as initial material sorting, running simple scripts, building basic data processing pipelines, format conversions, and auxiliary searches. These tasks are not necessarily complex but often time-consuming and labor-intensive. If an Agent can reliably handle some of these basic steps, it may not immediately change investment decisions but could alter the allocation of research resources, allowing researchers to spend more time on framework judgments, industry comparisons, and transaction validation.

Therefore, private equity interest in “Lobster” is not purely out of curiosity. A private equity professional in South China told Cailian Press that what truly draws industry attention is not a particular open-source project suddenly going viral, but the gradual realization that Agents are different from traditional large-model tools. Their potential is not just in “answering questions” but in connecting tasks, calling tools, and replacing manual workflows. For private equity firms emphasizing efficiency and response speed, this capability is inherently attractive.

However, amidst the excitement, it’s important to recognize that private equity firms currently value these tools mainly in the context of high volatility and uncertainty. Their primary concern remains research frameworks, trading discipline, and risk management, rather than the flashy capabilities of any single tool.

He Li, General Manager of Zhizhi Investment, told Cailian Press that the “raising lobsters” craze sparked by OpenClaw essentially represents a paradigm shift from AI dialogue interaction to autonomous local execution—an industry milestone demonstrating AI’s transition from “talking” to “doing.” He sees this as a clear difference from the local deployment trend of DeepSeek a year ago, which was mostly about running models locally and manual invocation. OpenClaw, on the other hand, is moving toward autonomous AI gateways with real operational capabilities, with business models and industry impacts closer to true enterprise-level deployment.

Bao Xiaohui, Chairman of Changli Assets, added a demand-side perspective. She told Cailian Press that compared to DeepSeek a year ago, which was mainly about “chatting and summarizing” with higher thresholds and stronger enterprise attributes, OpenClaw now has the ability to directly operate on files and execute tasks, with lower usage barriers and costs. This makes the current wave more pragmatic and closer to real-world scenarios.

Some Experiment, Others Observe

Although “raising lobsters” has become a buzzword in the industry, from what Cailian Press has learned, private equity firms’ attitudes are quite divided.

Some organizations have already begun to explore, even if they haven’t formed formal company-wide projects, with researchers, programmers, or TMT staff trying it out first.

These organizations tend to be pragmatic: they may not immediately integrate Agents into core systems but believe they should at least understand what these tools can do. Especially for private equity firms focused on technology, computing power, software, and AI applications, not experiencing the tools firsthand makes it hard to gauge their capabilities and understand industry progress. An industry insider said many firms are interested in these tools partly out of practical needs and partly to better understand how AI applications are evolving and what the market is actually trading.

Another group of firms remains cautious. A private equity professional told Cailian Press that their organization has not deliberately deployed such tools, mainly because their current workflows are not significantly hindered, and there are still concerns about access rights, data security, and compliance responsibilities.

Bao Xiaohui openly stated that while the trend itself is strong, institutions are unlikely to immediately incorporate these tools into real work environments. She believes OpenClaw differs from earlier models limited to chatting and summarizing, as it now has more powerful operational permissions, changing the risk profile. For investment firms, risk control is always the top priority. Client information, holdings data, investment strategies, and research materials are highly sensitive. Connecting high-permission Agents directly to work environments could, even with low probability, lead to prompt injection, permission breaches, or data leaks, potentially breaching risk thresholds.

Some industry insiders told Cailian Press that this “lobster” craze, to some extent, resembles previous years’ efforts by institutions to research large models, build knowledge bases, and connect research assistants—starting with early adopters and gradually spreading. But unlike before, Agents emphasize execution ability rather than just generation, which makes institutions more cautious when deciding whether to deploy. After all, a model that only answers questions is different from a tool with operational capabilities that can call external resources—under risk control considerations, they are not on the same level.

More Concerned About How It Will Reshape Research and Quantitative Profits Than About Replacing People

Looking further ahead, the deeper discussion triggered by OpenClaw in the private equity circle is not just about “whether to adopt,” but how it will change the industry’s workflow.

From a subjective research perspective, its most practical value remains as an aid rather than a replacement. Whether it’s basic coding, data collection, or initial research framework setup, these can be delegated to Agents for a first pass. But final investment judgments, industry comparisons, and risk assessments still rely on human experience and boundary awareness.

An anonymous industry insider believes that the real transformation in private equity may not be the research analyst role itself but rather how much time analysts spend on low-value-added tasks. Those who can quickly delegate these tasks to tools are more likely to reallocate human resources toward core judgment and validation.

From a quantitative perspective, this wave’s significance is more complex. Some quantitative private equity professionals pointed out that general AI will not easily displace top-tier quantitative firms, as their advantages in data, computing power, model iteration speed, and risk control are not easily matched by retail or small teams in the short term. But the problem is that as more participants use AI for decision support, market behaviors may become more homogeneous, trading points more concentrated, and volatility amplified. In this scenario, the most impacted may not be the top institutions but those middle-tier strategies that rely on following trends, lacking proprietary logic and risk control.

Many industry insiders believe that private equity will continue to explore and test similar tools, but the pace may not be as aggressive as discussions on social platforms. The reason is simple: investment firms can track frontiers but cannot easily turn cutting-edge technology directly into production. For private equity, the true value of Agents may not be in replacing many researchers today but in pushing institutions to rethink future research divisions. Humans will still make the final judgments, but machines may increasingly participate in preparatory work before decision-making.

(Cailian Press, Wu Yuqi)