Three Weeks That Signal a Direction

Anthropic's product velocity over the last three weeks has been striking. Not because of individual features – but because of the order in which they're appearing.

First came Dispatch – the ability to assign tasks to Claude from your phone and let it continue working asynchronously.

Then came Computer Use – letting Claude click, scroll, type, navigate applications, and interact with software directly.

Taken together, this marks a paradigm shift in the AI product landscape.

From Conversational to Operational

We're moving from:

• AI as a conversational interface (you ask, it answers)

to:

• AI as an operational interface (you delegate, it works)

That's a much bigger change than it sounds.

A chatbot answers questions. An operational interface takes ownership of workflows. The difference isn't gradual – it's fundamental.

The Deliberate Layering

The key insight isn't simply that Claude can now "use a computer." It's the deliberate order in which Anthropic is layering capabilities:

1. Stronger reasoning – the cognitive foundation
2. Persistent task execution – beyond a single message
3. Remote delegation via Dispatch – assign tasks from your phone
4. Direct software interaction via Computer Use – work inside the system

Each layer builds on the previous one. This isn't a feature blitz – it's architecture.

What This Means for Businesses

At Till Freitag, this isn't an abstract future vision. We already work with Claude as a central reasoning layer in our workflows.

But the implications go further:

For project management: An agent that doesn't just suggest tasks but creates, updates, and tracks them in monday.com.

For CRM: An agent that consolidates customer data, schedules follow-ups, and maintains contact histories – without anyone filling out a form.

For development: An agent that doesn't just write code but reviews PRs, monitors CI pipelines, and coordinates deployments.

The Open Questions

Of course, it's still early. The real questions are around:

• Reliability – How dependable are these systems on messy, real-world workflows?
• Safety – What guardrails does an agent need when it's autonomously operating software?
• Permissions – Who grants an AI agent access to which systems?
• Accountability – Who's responsible when the agent makes a mistake?

These questions aren't trivial. But they're solvable – and Anthropic is addressing them more systematically than most competitors.

Our Take

The most valuable AI products won't just answer questions. They'll take ownership of workflows.

And the last three weeks suggest Anthropic understands that very well.

The direction is clearly toward AI as a digital coworker – not a smart chat window. For companies preparing for this now, it will be an enormous competitive advantage.

For everyone else, it will be a race to catch up.