When AI Stops Advising and Starts Acting

I’ve always been fascinated by tools that promise to make work a little easier. In high school my teachers called it “looking for shortcuts.” I still prefer to think of it as finding better ways to get things done.

Lately, AI automation with tools like OpenClaw and Microsoft’s Project Opal have been generating a lot of buzz in the tech community. These tools look a lot like an exciting leap towards autonomous agents.

This welcomes a new generation of AI that doesn’t just chat or give advice but can actually perform tasks on your behalf.  

While I’m optimistic about the potential, I also think it’s important to look at them with a healthy dose of caution and practicality.

What Are OpenClaw and Opal?


OpenClaw is an open-source personal AI assistant that runs on your own device. It’s like having a 24/7 digital intern that you can chat with on platforms like WhatsApp or Slack. You give it natural-language instructions like “schedule a meeting for next week” or “clear my inbox” and it will attempt to do just that on your behalf.

Microsoft’s Project Opal, on the other hand, is part of the Microsoft 365 Copilot platform for businesses. It’s designed to automate multi-step office tasks (think: gathering data for reports, updating systems, or handling routine HR processes) by using an AI to navigate software just as a human would, all within a secure, monitored environment.

Both OpenClaw and Opal are examples of a broader shift in AI: moving from simply assisting with information to actually executing tasks. This shift could be transformative. For those of us with less time on our hands than we’d like, these agents offer the promise of liberating us from repetitive “busy work.” Imagine an AI that can handle the drudgery of scheduling meetings, filing expense reports, or collating compliance documents… Freeing your time for higher-value work or creative problem-solving. In early internal trials, Microsoft employees using Opal to automate workflow tasks saved significant hours — demonstrating the tool's real-world productivity potential.

As someone who helps oversee how new tools are introduced in our organisation, I’m cautiously excited. It’s not hard to see the potential. Fewer mundane tasks for people could mean happier employees and greater efficiency. This technology is developing quickly with support from major players (the creator of OpenClaw even joined OpenAI, indicating the growing interest in personal AI agents).

The long-term vision is interesting…

A Cautious Approach to Adoption


However, “cutting-edge” doesn’t always mean “ready for everyone.” These tools are still in their early days. OpenClaw, for example, is a community-driven project that rapidly gained popularity, but it also experienced a high-profile security vulnerability recently. A reminder that new software can have teething problems. Project Opal is currently in a limited preview; Microsoft itself describes it as early and evolving. 

For most organisations, the message is clear: there’s no need to rush head-first into implementation. Instead, you should take thoughtful steps: 

Understand the Use Case: Identify where an AI agent could genuinely solve a problem or save time in your context. Not every task is a reason for automation, and not every team is ready for it.

Start Small: If you’re interested, start with a pilot or experiment. For example, you might trial OpenClaw with a small team to automate simple administrative tasks or explore Microsoft’s Opal in a sandbox environment for specific workflows. Use these trials to gather feedback and measure impact before scaling up.

Provide Training and Oversight: AI agents introduce a new way of working. Teams need guidance on how to delegate tasks, monitor outcomes, and intervene when needed. Building trust in the technology takes time and transparency. Encourage open dialogue where your team can share concerns and successes.

Mind Security & Compliance: Most importantly for enterprise use, you absolutely must ensure that any AI agent operates within the guardrails of your organisation’s security policies. Tools like Opal come with administrative controls for a reason. Use them to whitelist approved actions and protect sensitive data. Even with a personal tool like OpenClaw, discuss data privacy and security implications before rolling it out broadly.

Looking Ahead


The future of AI automation tools is undoubtedly bright. I think that over the next few years we’ll see these autonomous agents become more robust, secure, and deeply integrated into the tools we use every day. They have the potential to transform how we work, maybe one day performing tasks overnight or in the background, so we come into the office with routine work already done.

Yet, with all the excitement, it’s worth remembering that we don’t all have to jump in at once. In technology (as in fashion), timing and fit are crucial. Early adopters will naturally lead the way, discovering best practices and surfacing challenges. The rest of us can learn from their experiences. There’s great value in being informed and ready, but also in being deliberate.

In my role, I remind colleagues and clients that embracing innovation is a journey: explore and experiment but do so pragmatically. By testing the waters and understanding where AI automation adds real value, we can adopt these powerful tools in a way that benefits everyone without unnecessary risk or disruption.

In short, tools like OpenClaw and Opal point to a future where AI moves beyond assistance and into execution.

The question is no longer “Can AI help me think?”
It’s “Can AI help me do?”

Increasingly, the answer appears to be yes. The organisations that benefit most won’t necessarily be the fastest adopters, but the most thoughtful ones — introducing these tools carefully, securely, and where they genuinely add value.