Customer asks for a refund and they're inside your window. Three agents on your team will write three different answers.
Here's the one I'd have everyone send:
Thanks for reaching out. You're within our [X]-day refund window, so I've started your refund to the original payment method. You'll see it in [X–Y] business days. If anything looks off, reply here and I'll fix it.
Fill in the brackets once, agree on it, and stop rewriting it per ticket.
The other 5 are in the pack, link below
The other 5: delayed order, policy change, where's my order, escalation, and a first-week reply for new hires.
All copy-ready, free: spellshelf.ai/academy/suppor…
1
8
Ask 3 support agents how they answer a refund request, and you'll often get 3 different answers.
Nobody's being careless. Each of them has their own copy: an old Slack thread, a doc from onboarding, something saved in ChatGPT. When the policy changes, one copy gets updated and the other two don't. Customers notice.
So here are the replies every support team should agree on once and share:
Refund request
Delayed order
Policy just changed
"Where's my order?"
Escalating to a senior agent
A new hire's first-week replies
Give each one an owner. When the policy changes, the owner updates it once and everyone copies the new version. New hires start from the team's answer, not a guess.
I've made these free to copy as a starter shelf. Test one, then invite 2 teammates. Link in the first reply.
Support leads: which reply does your team argue about most?
Branding photographers are going to either love me or hate me for building this.
I got tired of watching a loved one:
“Post more on Instagram.”
“Make more Reels.”
“Hope someone finds you.”
Heck with that.
There are businesses all around you that need better photography.
So I built 6 EL Studio to go find them.
Put in your city. Tell it you’re a branding photographer. It finds local businesses that may need you, tells you WHY, ranks the opportunities, and helps you figure out what to say to them.
No demo.
No sales call.
No buying some garbage list of “leads.”
Just go try the damn thing.
6elstudio.com
Then come back and tell me if it sucks.
I can take it.
1
35
If a strong performer left tomorrow and their best AI workflows disappeared with them, you don’t have a knowledge management problem.
You have an ownership problem.
Most companies still refuse to admit this.
If your team still treats prompts like personal notes instead of shared infrastructure, the issue isn’t the model.
Agree or disagree?
1
34
Ty East retweeted
Replying to @elonmusk
In working with @salesforce over the years, how well does Grok Build do with development and administration of Salesforce @elonmusk?
SF has not ever been a traditional development environment.
Many friends and colleagues in the industry are starting to realize that adding various AI into thier company's workflow does not reduce the need for people.
It reduces the need for people to do the automated task.
It creates more work for the speed of the results. If you don't plan for the outcome, why do it.
How are you preparing?
@grok @OpenAI @claudeai @Gemini
1
1
41
Ty East retweeted
AI usage spreads quietly inside companies.
Then the team realizes:
No one owns the workflows.
No one knows which prompts work.
No one knows what new hires should use.
That is prompt chaos becoming an operations problem.
spellshelf.ai
Ty East retweeted
The best AI workflow in your company is probably not documented.
It is probably sitting in:
- someone’s ChatGPT history
- a private doc
- a Slack thread
- one teammate’s memory
That is not a prompt problem.
It is a workflow management problem.
spellshelf.ai
Ty East retweeted
SpellShelf has a better experience for teams building shared AI workflows.
Take the prompts that already work and make them easier to organize, share, reuse, and improve.
Prompt chaos is fixable.
Build your shared prompt library: spellshelf.ai
If you are leading AI adoption internally, you are probably seeing the same pattern:
The challenge is no longer convincing people to try AI.
The challenge is making good usage repeatable.
Teams are already finding prompts that work.
The problem is that those prompts are rarely organized, shared, governed, or improved systematically.
So AI adoption stays messy:
•multiple teams reinvent the same workflows
•prompt quality is uneven
•useful knowledge is hard to transfer
•standardization never catches up to experimentation
What many companies need next is not another model.
They need a shared workflow layer for how teams actually use the models they already have.
That is what turns AI from scattered activity into managed capability.
A strong internal AI program should not just increase usage.
It should increase reuse, consistency, and transferability.
That is where prompt management becomes part of company infrastructure.
Spellshelf.ai
1
18
Founders can usually tell when AI is getting used across the company.
What is harder to tell is whether the value is compounding.
In many companies, AI usage looks like this:
•lots of individual experimentation
•good prompts discovered in isolation
•inconsistent adoption across teams
•very little shared infrastructure
That means the company is learning, but not retaining what it learns.
The real opportunity is to turn useful AI behavior into a team asset.
Something reusable.
Something transferable.
Something a new hire can access.
Something leaders can improve over time.
The difference between “people are using AI” and “the company is getting better at AI” is usually systemization.
1
47
Agencies are using AI heavily.
But many still do it in a way that does not compound.
The best workflows often get buried in:
•client docs
•Slack threads
•personal notes
•saved chats
•whoever happened to figure something out first
That creates avoidable drag:
•new team members ramp slower
•delivery quality varies by account lead
•useful processes disappear between projects
•teams rebuild instead of improve
For agencies, the advantage is not just using AI.
It is building reusable AI workflows that survive across clients and team members.
The agencies that get the most from AI will not be the ones with the most tools.
They will be the ones that keep their best workflows organized and transferable.
1
27

