Big Tech is Panic Building Its Own OpenClaw

Openclaw appears to have sparked a new race in Silicon Valley, with Microsoft, Google, Meta, and Anthropic all building AI agents that can autonomously execute tasks instead of simply responding to prompts. 

At NVIDIA GTC 2026, CEO Jensen Huang quipped that every software company now needs an OpenClaw strategy.

 To begin with, Microsoft has introduced new capabilities in Copilot Cowork, allowing users to delegate and complete work tasks across applications, data sources, and devices.

Copilot Cowork is built on what Microsoft calls Work IQ, an intelligence layer that uses organisational data, tools, and workflows to plan and execute tasks. 

Much like Openclaw, which allows users to manage multiple agents through their smartphones, Copilot Cowork will also be available on iOS and Android. This will let users assign tasks on the go while execution continues in the cloud. “You can delegate work the moment you think of it—on your commute, between meetings, or away from your desk—and come back to a finished outcome,” Microsoft said.

Moreover, the company also introduced skills, which are reusable instruction sets that capture structure, tone, and processes, allowing Cowork to repeat a specific way of working across tasks. 

Jay Parikh, Executive Vice President of Microsoft’s CoreAI, in an earlier interaction with AIM, described his experience of using OpenClaw as “eye-opening” and “very fun”. He said it introduces a far more persistent and personalised AI interface than traditional chatbots. “It can be in WhatsApp, and can do various things,” he said, adding that it can also perform real work depending on how users configure it. 

He further noted that Copilot Cowork offers similar capabilities, but without many of the risks enterprises may face with OpenClaw.

The appeal of these systems is not limited to Big Tech. Rahul Agarwalla, Managing Partner at SenseAI Ventures, tells AIM that OpenClaw gained popularity because it brings AI closer to functioning as a personal assistant.

Speaking about his personal experience with the tool, he said he uses it to manage his communication channels. “I have too much WhatsApp, so much email, which is difficult to track,” he says, adding that conversations often happen across multiple channels and groups, making organisation challenging. “The first thing I’m trying to do is organise my data.”

Vibe coding startup Emergent also recently launched its own version of OpenClaw called Wingman. Built as a messaging-first interface, Wingman allows users to interact with it through platforms such as WhatsApp, Telegram, or iMessage, assigning tasks conversationally.

“We are moving into a world where you will soon have agents which can run for 24 hours autonomously without any guidance,” Mukund Jha, CEO of Emergent labs, tells AIM.

He further shares that the company will soon launch a new agent that allows users to simply give an instruction while the system handles the rest. “It will be able to run for eigh to 10 hours without any instructions,” he adds. 

Anthropic Comes After Wall Street

Anthropic is aggressively pushing into enterprise AI with Claude Cowork, Managed Agents, Finance Agents, and workflow automation.

Notably, OpenClaw’s original name was Clawdbot. Anthropic reportedly objected to the branding, which eventually led to the project being renamed OpenClaw.

Anthropic recently unveiled 10 ready-to-run AI agent templates built for some of the most time-consuming tasks in finance.

The template list reads like a junior banker’s nightmare to-do list: pitch builder, meeting preparer, earnings reviewer, financial model builder, market researcher, KYC screener, valuation reviewer, general ledger reconciler, month-end closer, and financial statement auditor.

Each template is a packaged architecture combining task-specific skills, governed data connectors, and subagents—specialised Claude calls that handle functions like comparables selection and methodology checks.

The agents run on Claude Opus 4.7, which Anthropic claims now leads Vals AI’s Finance Agent benchmark. They can be deployed either as plugins inside Claude Cowork or as fully autonomous Managed Agents.

Anthropic has also enabled Claude to work directly inside Excel, PowerPoint, Word, and Outlook through native add-ins.

Anthropic CEO Dario Amodei previously described the future of AI as “a country of geniuses in a data centre”, with autonomous agents working together across coding, research, analysis, and operations at a massive scale.

The launch builds on Anthropic’s January release of 11 open-source plugins for Claude Cowork, which helped spark the SaaSocalypse debate across Silicon Valley. 

Those plugins targeted enterprise workflows spanning productivity, sales, marketing, finance, customer support, product management, data analysis, and biology research.

Meta & Google Join the Race

While Microsoft is busy building AI coworkers, Meta has entered the race as well. The company is building a highly personalised AI assistant for its more than three billion users, according to Financial Times. Powered by Meta’s new Muse Spark AI model, the assistant is currently being tested internally by a group of employees. 

Meta is reportedly building something similar to OpenClaw, with AI agents capable of autonomously completing tasks instead of simply responding to prompts. The company also appears willing to push the trust boundary further than most. Meta ultimately wants users to share sensitive information, including health and financial data, with these assistants if they choose to.

CEO Mark Zuckerberg, during the company’s recent earnings call, said that OpenClaw offers “a very exciting glimpse” into the future of AI agents, but argued that current implementations remain too complex for mass adoption. “To set up OpenClaw, you need to install it on a computer locally and then get into a terminal and configure a bunch of things,” he said, adding that such workflows are currently accessible only to “a small number of millions of people.”

According to Zuckerberg, Meta is trying to make these experiences polished and simple enough for billions of users. He added that the opportunity extends beyond personal assistants and into business applications as well. “A lot of people’s goals are to create things—to create websites, products, grow their products,” he said. 

Not to be left out, Google is reportedly building its own answer to the agentic assistant race. Business Insider reported that the company is working on a project codenamed Remy, a Gemini-powered AI agent currently being tested internally through a staff-only version of the Gemini app.

Remy is described as a 24/7 personal agent for work, school, and daily life—something that can take actions on a user’s behalf instead of just answering questions or generating content. Deep integrations with Google’s own services are also reportedly part of the plan.

The timing is noteworthy. Google I/O 2026 is around the corner, and Remy could end up making an appearance on stage. Google has already been laying the groundwork for agentic AI through initiatives such as its Agent2Agent protocol and Gemini integrations across Workspace. If it ships, Remy could become the consumer-facing layer tying all of that together.

OpenClaw may still be early and imperfect, but it has clearly pushed Silicon Valley to rethink what AI assistants could eventually become.