Subtitle: Anthropic’s acquisition of Vercept signals that AI operating software directly — without APIs, without developers — is coming faster than most entrepreneurs realize. Here’s the workflow audit framework that gets you ready.
Most entrepreneurs think of AI as a tool that generates text, analyzes data, or answers questions.
That definition is already outdated.
On February 25, Anthropic acquired a company called Vercept specifically to accelerate Claude’s ability to operate software — not through an API, not through a custom integration, but by looking at a screen and clicking, just like a human employee would.
The era of AI that does computer work is arriving. The entrepreneurs who see it coming will have a significant advantage. Here’s what that means and what to do about it right now.
Key Takeaways
- Anthropic acquired Vercept to advance Claude’s “computer use” capabilities — AI that operates software through its user interface, not through APIs.
- This means virtually any software with a screen can be automated, regardless of whether it has an API or technical integration options.
- The biggest barrier to AI automation for small businesses — “our systems don’t have an API” — is being systematically removed.
- Entrepreneurs who map their current repetitive workflows now will be positioned to automate them faster as the capability matures.
- Computer use AI is already functional for tasks like browser navigation, form completion, and multi-system data transfer. It’s improving rapidly.
The Hidden Wall in AI Automation
If you’ve tried to automate something in your business with AI, you’ve probably hit this wall:
“Our software doesn’t have an API.”
API — Application Programming Interface — is the technical pathway that lets two software systems talk to each other. It’s what Zapier, Make, and most automation tools depend on. If your CRM doesn’t have an API, you can’t automate it. If your accounting software doesn’t expose its data, no automation tool can touch it.
For large companies with technical teams, this gets solved with custom development. For most small businesses, it means the automation doesn’t happen.
That’s the wall Anthropic is now building a door through.
What “Computer Use” AI Actually Does
Claude’s computer use capability doesn’t need an API. It sees the screen. It reads the interface. It moves the cursor. It types. It clicks.
Just like a person sitting at the computer would.
In practical terms, this means:
Before computer use: You need an API, a developer, and a custom integration to automate moving data from your CRM into your invoicing software.
With computer use: You describe the task. The AI opens the CRM, finds the information, opens the invoicing software, and enters it. No API. No developer. No custom code.
The Vercept acquisition is specifically designed to make this capability more reliable, faster, and applicable across a wider range of software environments. Including the legacy systems that most small businesses run on.
Why This Changes the Calculation for Small Business
Every business owner I know has at least three tasks on their list that look like this:
- Copy-pasting data between two systems that don’t talk to each other
- Checking a dashboard, recording the number, entering it somewhere else
- Filling out the same type of form with slightly different information each time
- Pulling a report from one tool and reformatting it for another
These tasks are too small to justify a developer building a custom integration. But they’re not too small to steal twenty hours a month.
Computer use AI is designed exactly for these gaps. The tasks that are too repetitive to be interesting and too inconsistent to be reliably automated the old way.
As this technology matures — and it is maturing rapidly — the question shifts from “can we automate this?” to “have we identified what we want to automate?”
That’s the question to answer now.
The Workflow Audit: Mapping Your Business for Computer Use Readiness
Here’s a simple framework for identifying your highest-value computer use automation targets before the capability is fully mature.
Step 1: List every task in your business that involves moving information between systems.
Don’t filter by whether it’s “automatable.” Just list every task where you (or a team member) open one system, look at information, then enter it into another system. These are your primary targets.
Step 2: Rate each task on two dimensions — frequency and cognitive load.
Frequency: How many times per week does this happen?
Cognitive load: Does this require judgment and decision-making, or is it mostly mechanical follow-a-procedure?
High-frequency, low-cognitive-load tasks are your best automation candidates. The AI doesn’t need to decide anything — it just needs to execute the steps reliably.
Step 3: Document the procedure for your top three candidates.
Write out, step by step, what a human employee would do to complete this task. Where do they click? What do they look for? What do they enter where? The more precisely you document this, the easier it will be to turn into an AI workflow when the tool is ready.
Step 4: Test what’s possible today.
Claude’s computer use capabilities are already functional for many tasks. Try giving Claude a task from your list and see how far it gets. You’ll learn where the technology is now and how close it is to solving your specific problem.
Step 5: Build the habit of thinking about AI-executed workflows, not just AI-generated content.
The biggest mindset shift required isn’t technical — it’s perceptual. Stop asking “what can AI write for me?” Start asking “what can AI do for me?” Different question. Much larger answer.
The Timing Question
I want to be direct about something.
Computer use AI is real and it’s functional — but it’s not perfectly reliable for every task yet. You’ll encounter limitations. Some workflows will require more human oversight than others. Some software environments are harder for AI to navigate than others.
None of that changes the strategic reality.
The entrepreneurs who are thinking now about which workflows they want to automate — who are documenting processes, identifying friction points, and testing current capabilities — will implement computer use automation months ahead of those who are waiting for the technology to be “perfect.”
There is no perfect. There’s ready, and there’s behind.
Frequently Asked Questions
Is computer use AI available today, or is this a future capability?
It’s available today through Claude’s API, and it’s functional for many tasks. Anthropic’s acquisition of Vercept is designed to accelerate and improve the capability, not introduce it. You can test it now.
Does computer use AI work with any software, or only specific applications?
It works with any software that has a visual interface — browser-based tools, desktop applications, legacy systems. The more standardized and consistent the interface, the more reliably the AI can navigate it.
Do I need technical skills to set up computer use automation?
For basic tasks, no. For complex multi-step workflows, some technical configuration helps but isn’t always required. The technology is designed to reduce the technical barrier, not eliminate it entirely.
What’s the difference between computer use AI and robotic process automation (RPA)?
Traditional RPA follows rigid, pre-programmed rules and breaks when anything changes. Computer use AI can adapt — it sees the screen and reasons about what to do, rather than following a fixed script. This makes it far more robust for real-world, slightly-inconsistent workflows.
How do I know if a task is ready for computer use automation?
If you can write down every step a human would take to complete it (open this, click here, copy this value, enter it there), it’s a candidate. The documentation exercise is the test — if you can describe it precisely, an AI can likely follow the description.
The Bigger Picture
Anthropic spent real money to acquire Vercept. Not for a feature — for a capability shift.
They’re building AI that works as an operator of technology, not just a generator of content. And they’re doing it specifically for enterprise and business use cases.
Every business owner who thinks their workflows are “too small” or “too specialized” or “don’t have an API” to automate should be paying attention to this.
The excuse is being eliminated. The question is whether you’ll have your workflows mapped when the capability arrives at your doorstep — or whether you’ll be starting the conversation six months late.
Map the work now. The tools are coming.
Jonathan Mast is the founder of White Beard Strategies and has served more than 500,000 entrepreneurs through his AI coaching programs and the Perfect Prompt Framework. He speaks regularly on practical AI implementation for business owners.
