Two AI releases earlier this year are pushing users to hand over control and let autonomous tools complete tasks on their behalf. IT leaders need to be prepared to deal with the consequences. Anthropic released its Claude Cowork in January for macOS and in February for Windows, and usage of its OpenClaw grew rapidly earlier this year after developer Peter Steinberger, now with OpenAI, opened the project up in late 2025.
While most organizations are focused on developing AI that augments human work, there has been a huge surge in interest in autonomous AI since late last year, says Neal Riley, chief innovation officer and former CIO at IT and digital transformation consulting firm The Adaptavist Group.
Many organizations, even traditionally cautious companies in the financial services and healthcare sectors, have begun experimenting with autonomous artificial intelligence as they seek to reshape their workflows, says. Despite concerns about unpredictable outcomes and autonomous agents that operate as shadow AI, early adopters of AI see huge potential for the technology to be a force multiplier that, for example, enables non‑technical people to solve small IT problems without involving the technical team.
OpenClaw and Claude Cowork are at the cutting edge of this upcoming revolution, allowing users to use artificial intelligence to automate workflow on their computers.
The OpenClaw bots integrate with external large language models (LLM) AI, such as the Claude and GPT models from OpenAI, and users access it via a chatbot that runs on a messaging service like WhatsApp, Telegram, or Discord. Users grant Claude Cowork access to their applications and files and then ask the artificial intelligence to complete the tasks.
Cowork can organize files, create spreadsheets, prepare reports and analyze notes, gaining access to files on the user's computer, pulling content from applications such as Slack and browsing the web for more information.
Before Claude takes action, it presents the user with the plan and waits for approval, according to Anthropic. However, some users have given these autonomous agents a high level of control, and there are risks when they are handed over to their computers without strict limits. Meta AI security researcher Summer Yuetweeted in late February that OpenClaw attempted to delete her inbox after a request to clean it.
One of the top responses to Yue's tweet was an image of someone handing a weapon to a chimpanzee. Researchers have also found several security gaps in OpenClaw, including a vulnerability to prompt injection attacks. Here lies the problem: AI experts see huge potential advantages with autonomous AI — with the ability to create massive efficiencies in the workplace — but the risks are significant.
While users have not yet seen autonomous artificial intelligence complete work faster or cheaper than humans — tokens are expensive — the technology has the potential to reshape the nature of work for the better. Autonomous AIs will allow organizations to reallocate their human resources to new tasks, removing much of the tedious work, say supporters.
Μία από τις μεγάλες προκλήσεις είναι να κατανοήσει το AI πώς λειτουργεί μια επιχείρηση. Η ταχύτητα είναι ένα τεράστιο πιθανό πλεονέκτημα των αυτόνομων agents, αλλά είναι επίσης ένα από τα μειονεκτήματα.
Despite the risks, the market is moving quickly toward autonomous AI, with large adoptions coming in the next two years, says Russell Twilligear, head of AI research and development at AI-generated content delivery company BlogBuster.
Despite concerns about security and production, Riley encourages IT leaders to allow employees to experiment with emerging technology due to its imminent adoption. Organizations that invest in AI training and allow employees to play with the technology tend to achieve better outcomes from developments, he notes.
