SkillCurio Dispatch · Issue #4
AI can do more work. That can make your judgment more valuable.
AI agents are getting more capable. The opportunity is learning how to direct that capability, apply what you already know, and turn AI into leverage.
Get the next Dispatch · The complete issue, every two weeks.
AI keeps getting better at doing the work.
Writing code. Researching. Reviewing documents. Moving information between systems. Running multi-step workflows.
That can sound threatening.
But there is another way to look at it.
The more work AI can perform, the more valuable it becomes to know what work is worth doing, how it should be done, and whether the result is actually useful.
That's good news for people who already understand customers, systems, operations, security, workflows, or an industry.
You don't have to compete with AI at everything it does well.
You can learn how to put it to work.
And several developments this month show what that increasingly looks like.
What changed
1. Anthropic is already seeing a new kind of AI work
Anthropic recently published a detailed look at how AI is changing work inside its own company.
By August, its interactive AI agents were collectively running for roughly six hours for every hour attributed to the employees directing them.
That's a lot of machine labor.
But the humans haven't become irrelevant.
They're increasingly building dashboards, monitoring systems, and workflows to coordinate all that AI-generated activity.
Managers are also placing more emphasis on autonomy, judgment, and the ability to use AI effectively.
Anthropic is an unusually AI-heavy workplace, so we shouldn't assume every company will look like this.
But the direction is useful.
As AI takes on more execution, people can move toward directing, evaluating, and improving the work.
That creates a practical path for learners.
You don't need to become the person who manually performs every step.
You can become the person who understands the process well enough to guide AI through it.
2. Most companies still haven't figured out how to turn AI into real value
Gartner's latest CIO research contains a striking number:
Only 13% of surveyed CIOs said their organizations had achieved significant or transformational value from AI.
That might sound discouraging.
I think it's the opposite.
It means there is still a huge amount of work to be done between having AI and using AI well.
The organizations reporting stronger results were more likely to measure outcomes, stop weak experiments, and redesign work around humans and AI agents.
And only 4% of surveyed CIOs said they had explicit plans to redesign jobs for a human-agent workforce.
Think about what that means.
Companies are buying AI tools.
They still need people who can answer:
Where should we use them?
Which process is actually worth improving?
How do we know whether it worked?
What happens when the AI gets something wrong?
Those aren't questions reserved for AI researchers.
They're questions for business analysts, IT professionals, operations teams, managers, security specialists, and experienced employees who understand the work.
3. AI agents are becoming enterprise users
This is where AI starts becoming very real.
Workday is rolling out infrastructure that allows AI agents to interact directly with HR and finance systems.
Microsoft is expanding security controls specifically for AI agents.
Put those developments together and you can see where things are heading.
An agent may be able to:
• retrieve employee information
• analyze a request
• prepare a recommendation
• update a business system
• trigger another workflow
That sounds sophisticated.
But underneath it are problems many experienced professionals already understand.
Who gets access?
What should they be allowed to change?
Where does approval happen?
What gets logged?
What happens when something behaves unexpectedly?
Those are familiar concepts in IT and security:
least privilege, identity management, audit trails, separation of duties, approval workflows, change control.
AI isn't making that knowledge obsolete.
It's giving it a new place to matter.
If you've spent years working with enterprise systems, permissions, users, workflows, or security controls, you already have part of the foundation.
Now you can start learning how those same principles apply to AI agents.
Explore Workday's Trusted Agent Platform →
Read Microsoft's agent-security research →
Your next learning step
Design an AI agent before you build one
Here's something you can do this week without writing code.
Choose one recurring task you understand well.
Maybe it's reviewing incoming requests, organizing support tickets, checking documentation, preparing a report, or researching information.
Now imagine an AI agent helping with that task.
1. What outcome should improve?
Don't start with “I want to use AI.”
Start with the problem.
2. What information does the agent need?
Only give it what is actually necessary.
3. What actions should it be allowed to take?
Separate reading information from changing it.
4. What still requires your approval?
Decide where human judgment stays in the loop.
5. When should the agent stop and ask for help?
Missing information? Conflicting instructions? Low confidence? Unexpected access?
Your finished map should look like:
Goal → Data → Allowed actions → Human checkpoint → Stop conditions
That's already AI systems thinking.
No coding required.
And if you can do this well for a process you already understand, you're beginning to build a capability companies are going to need.
Success condition: You can explain both what your proposed agent should do and where your own judgment still adds value.
Opportunity Watch
AI may be lowering the barrier between “I have an idea” and “I can try it”
Bill Gross—the entrepreneur behind Idealab—is taking a big swing at AI-assisted entrepreneurship.
His new Idealab.ai platform includes more than 1.2 million AI-generated business ideas searchable by things like skills, geography, capital requirements, and automation potential.
That number isn't the exciting part.
AI can generate ideas all day.
The interesting part is what is happening to the cost of testing one.
Research.
Market analysis.
Prototype development.
Copywriting.
Software creation.
Operations.
Tasks that once required several people can increasingly be attempted by one capable person using AI tools.
That opens a door for experienced professionals.
Someone who has spent 10, 15, or 20 years inside an industry knows things that an idea generator doesn't.
You know which frustrations are real.
You know which processes waste time.
You know what customers complain about.
You know where existing systems break down.
That knowledge can become raw material for building something.
AI may give you more leverage to act on it.
You don't need to become a startup founder tomorrow.
Try something much smaller.
Think of one recurring problem you understand unusually well.
Then ask:
Could I build a small demonstration of a better way?
Maybe the answer is yes now when it wouldn't have been five years ago.
That's exciting.
Read the Fast Company feature →
One thing to remember
AI can produce more.
More code.
More research.
More analysis.
More ideas.
More actions.
That doesn't make your experience less useful.
It can make your ability to direct all that capability more important.
If you understand the work, you can learn where AI helps.
If you understand the risks, you can define the boundaries.
If you understand the customer or business, you can judge whether the result is actually valuable.
And if you can combine those things with practical AI skills, you don't have to watch this transition from the sidelines.
You can participate in it.
You don't need to master everything.
Pick one workflow.
One problem.
One new capability.
Learn how AI can help you do something better than you could before.
Then build from there.
The goal isn't to keep up with AI.
It's to learn how to use AI to move yourself forward.
Originally published in SkillCurio Dispatch on October 22, 2026. This web edition preserves the approved newsletter text and source links.