Ash

Memento mori No life without death, no death without life.

AI Automation Consultant | Insurance | Entrepreneur | Traveler ☠️

[Ash @AshOperates]

@AnthropicAI @claudeai has me making sounds I can't make in public right now.

Claude Code sessions accessible from mobile. Memory that actually manages itself. Pop-up widgets asking clarifying questions with selectable choices instead of guessing wrong.

They found every spot that was frustrating and just... handled it. Thoroughly.

I'm not being professional about this anymore... Don't stop...faster!

I'm not saying I'm in a relationship with Claude but if my wife asks, we're "just talking."

[Ash @AshOperates]

Five cases beats a blank SOP.

AI can help interview the team, find the patterns, and turn judgment into training.

Automate later. Understand first.

[Ash @AshOperates]

Do not start your AI project with a blank SOP.

Start with five real cases.

[Ash @AshOperates]

5. The new-hire trap.

What would a smart new person miss?

That is usually the part your best employee forgot they even knew.

[Ash @AshOperates]

Same risk inside companies.

Bad AI creates dependency.
Good AI builds judgment.

The question is not whether your team is faster.

The question is whether they can see more next time.

[Ash @AshOperates]

There is a version of AI adoption that makes people faster and weaker.

That one scares me.

[Ash @AshOperates]

A 2025 PNAS study tested GPT-4 tutoring with nearly 1,000 math students.

During practice, students did better.

When support disappeared, test scores dropped.

The tool helped them finish. It did not always help them learn.

[Ash @AshOperates]

The question is not whether AI makes your best person 10x.

The better question: does it help the rest of the team stop making the same avoidable mistakes?

That is where the floor rises.

[Ash @AshOperates]

The AI study every operator should know:

5,179 customer support agents.

Productivity up 14%.

But the important part was not the average. It was who improved.

[Ash @AshOperates]

One person escalates fast.
One person misses the signal.
One person knows the edge case.
One person guesses.

AI used well can compress that gap.

Not replacement. Capability transfer.

[Ash @AshOperates]

That is the shift.

Old playbook: document the process, then automate.

Better playbook: use AI to discover the process, then decide what deserves automation.

AI won't fix a team that refuses to learn.

But it can make a willing team harder to break.

[Ash @AshOperates]

I used to think AI needed a clean SOP before it could help.

I don't anymore.

If your process is chaos, AI can create faster chaos. True.

But the useful use case is not automation first. It's discovery first.

[Ash @AshOperates]

Ask it to write your SOP and you get generic mush.

Ask it to walk through the last five ugly examples and you start finding rules.

What happened first?
What did the strongest operator notice?
Where did the handoff break?

[Ash @AshOperates]

Switching AI-augmented SaaS isn't switching tools. It's firing your best employee and replacing them with a stranger.

Own your AI capability. Run it on your terms. Or accept you're renting intelligence you'll never be able to leave.

[Ash @AshOperates]

SaaS companies just became drug dealers.

The first hit is free. AI features are subsidized, so the tokens feel cheap. Your team gets superhuman.

A year later, the AI knows your business better than most employees. Now try to leave.

[Ash @AshOperates]

The learning is baked into their model. You paid for the AI. You trained it. You don't get to keep the trained version.

The new system seems broken. Not bad software. It just hasn't had a year to learn your business.