Anthropic brings agentic computing to life
AND: Gartner and Forrester release 2025 predictions, and Apple Intelligence is finally about to emerge.
Preview from Anthropic shows agentic potential
In short: AI company Anthropic - maker of Claude - has released a preview of “computer use,” a way of having AI “take over” a computer screen to take over a range of tasks.
What's happened: Once again proving that it’s a contender in the generative AI space, Anthropic has released updates to its models but also slipped in something completely different: the unimaginatively-named “computer use”.
We’re also introducing a groundbreaking new capability in public beta: computer use. Available today on the API, developers can direct Claude to use computers the way people do—by looking at a screen, moving a cursor, clicking buttons, and typing text. Claude 3.5 Sonnet is the first frontier AI model to offer computer use in public beta. At this stage, it is still experimental—at times cumbersome and error-prone. We're releasing computer use early for feedback from developers, and expect the capability to improve rapidly over time.
In addition to releasing the API for others to test, the company has already made it available to a range of other companies to experiment with:
Asana, Canva, Cognition, DoorDash, Replit, and The Browser Company have already begun to explore these possibilities, carrying out tasks that require dozens, and sometimes even hundreds, of steps to complete.
Why you should care: In a period where there have been multiple agent-related announcements - from Salesforce, Microsoft and Asana - this particular announcement made the whole concept of agents less abstract, and more real.
In a series of videos, members of the Anthropic team demonstrated how this technology could be used to automate manual tasks. And at a basic level, this is what agents are intended to be: a way of the computer taking over multi-step tasks with minimal, if any, human interaction.
To a casual observer, the videos may appear underwhelming: on the surface, a computer that analyses one screen and copies information to another isn’t as mind-blowing as, say, watching ChatGPT produce a lengthy, structured response for the first time.
But - and this is a big but - this series of videos is probably closer and more accessible for the everyday person who is looking for technology to make their lives easier. And, what’s more, they can visualise how agents may help them - watching a form being filled out is exactly the type of tedious task we want to avoid doing.
The company was careful to state these were early previews of what is possible, and that they are interested in seeing how others use the technology. While other agent-related solutions may be more complex and help perform multiple tasks at scale, one can’t always visualise how these might work, and what they might do. As a way of helping grow awareness and understanding of agentic computing, these worked.
For more: The market is rapidly moving towards agents being part of every company’s marketing pitch, but staying with Anthropic for now, watch their videos on the company’s YouTube channel for yourself.
No surprise: AI continues to dominate 2025 predictions
In short: Both Gartner and Forrester have released their predictions for 2025, and it’s no surprise that AI features heavily in both.
What's happened: Predictions season is upon us as we head into the last quarter of the calendar year, and both Gartner and Forrester have published their best estimates of what the next 12 months will look like.
For Gartner, its top 10 Strategic Technology Trends have a heavy focus on AI, either calling out AI explicitly, or nominating technologies that will either be beneficiaries from advancements in AI, or will help drive greater use of AI.
While Gartner’s coverage is comprehensive, Forrester has taken an even broader approach, offering a range of reports (with supporting blogs) that cover a broad swathe of categories, which cover a range of workplace and industry scenarios.
Why you should care: Remembering that analysts are in the market talking to vendors and customers, predictions are always a useful litmus test for gauging where technology analysts think the market is going.
When it comes to productivity-focused themes, it’s no surprise to see that AI features strongly, and that agents are going to be a topic of interest in the coming year.
Gartner is particularly bullish on agentic AI and its adoption, estimating that we will go from 0% adoption in 2024 to 15% in 2028:
Agentic AI systems autonomously plan and take actions to meet user-defined goals. Agentic AI offers the promise of a virtual workforce that can offload and augment human work. Gartner predicts that by 2028, at least 15% of day-to-day work decisions will be made autonomously through agentic AI, up from 0% in 2024. The goal-driven capabilities of this technology will deliver more adaptable software systems, capable of completing a wide variety of tasks.
Agentic AI has the potential to realize CIOs’ desire to increase productivity across the organization. This motivation is driving both enterprises and vendors to explore, innovate and establish the technology and practices needed to deliver this agency in a robust, secure and trustworthy way.
Forrester looks at broader workplace trends, and perhaps takes a more considered view: it’s reinforcing the message that skills shortages are still impacting many organisations, and that this gap, as well as limitations around training and infrastructure, will impact the success of AI adoption:
The year 2025 will also mark a reckoning in AI adoption. Many companies have jumped headfirst into the generative AI (genAI) craze, but not all will make it out unscathed. One in seven firms will give up entirely, realizing that they lack the skills, training, and infrastructure to make AI work for them. The ones that get it right, however, will embed AI into their workflows at critical moments, driving productivity gains and preparing for an AI-powered future. In the long run, though, AI tools alone won’t save the day. While the number of organizations adopting these tools is set to jump from 35% to 50% in the next year, most users will still find themselves disappointed without a parallel shift in culture and governance to match the promise of the new tools.
Predictions such as these are never likely to be completely accurate. Reflect on past years’ predictions and it’s often the case that the precise prediction hasn’t been realised - or, like most things, will take longer to realise than anyone initially predicts.
That said, such predictions are useful to gauge the proverbial direction of travel - to get a sense of where the industry may shift in the next 12+ months, and this allows anyone with a vested interest in technology - which is, of course, most organisations - to get a sense of what might happen, from teams of people well-placed, and paid, to talk to a wide range of stakeholders to form those views.
For more: Review the predictions for yourself: both Gartner and Forrester have packaged up their trends into guides anyone can download (after registration). And keep an eye out for other firms’ predictions as we get closer to the end of the year.
Links
Apple has released betas of its iOS 18.1 and, more recently, 18.2 operating systems, both of which contain flavours of Apple Intelligence. iOS 18.1 is being released in the next few days, and it’s iOS 18.2 that will contain features that headlined Apple’s WWDC event - including Genmojis, AI image-related features, and integration with ChatGPT for release in late 2024. (The Verge)
OpenAI may release its latest flagship model - codenamed Orion - before the end of the year, although Sam Altman has called this “fake news”. (The Verge)
Anthropic has also released what it calls the analysis tool in Claude, which “enables Claude to write and run JavaScript code. Claude can now process data, conduct analysis, and produce real-time insights.” Tools like Claude appear to be leading the way as the vision of “personal software” becomes more realisable. (Anthropic)
Thanks for reading this newsletter. If you have any feedback on how it could be improved, or what you would like to see, please reach out.
Tim