The First Inch
On-Demand · $149
Team Training
Keynote
What It Is
Industries
Resources
Learn
AI Glossary
FAQ
Company
About
Affiliates
Contact
Get The First Inch · $149
Log in
An agent doesn't just answer — it acts. Watch one work, then learn to train your own in five simple steps.
Ask a chatbot for a hotel and it gives you advice. Ask an agent and it opens the site, searches, compares, and lines up the booking — pausing for your okay. It runs a simple loop, over and over, until the goal is met.
See the page → think about the goal → act with a click or some text → check the result, and go again. Your job isn't to do the steps — it's to set the goal and the guardrails, then approve the big moves.
Here's a real task, start to finish. Press play and watch the agent read the page, fill it in, and choose — then stop and wait for you. (This is a guided simulation of the real flow.)
Find your next stay
The same five moves work for any agent and any task. Click through them — each one is a habit that turns a vague wish into something an agent can run.
Tell the agent the outcome you want, not the clicks. Agents work backward from a clear goal — the sharper the goal, the better the result.
I want a short-list of 3 standing desks under $400 with good reviews, in a table with price, rating, and a link.
Name the finish line. "Good" and "some" are fuzzy; "3, under $400, in a table" is a target.
Hand it the facts it can't see and the lines it must not cross. Context makes it accurate; guardrails make it safe.
Budget is firm at $400. Only use retailers that ship to Texas. Never enter my payment details — stop before checkout.
Tell it what's true (budget, location) and what's forbidden (don't buy, don't share data).
One example of "done right" beats a paragraph of instructions. Show the format, the tone, or a sample of the output you expect.
Format each row like this: | Desk name | $349 | 4.6 stars (2,100 reviews) | [link] |
Agents pattern-match. A single clear example removes 90% of the guesswork.
Turn it loose on a real task and watch the first run. The agent reads the page, decides, and acts — you stay in the loop to approve key moves.
Go ahead and search now. Show me the shortlist before doing anything else.
Start with a low-stakes task. Watch the first run end-to-end before trusting bigger ones.
Coach it like a new teammate. Tell it what to keep and what to change, and it gets sharper every round.
Great list. Next time skip anything under 4 stars, and add the delivery estimate as a column.
Every correction is training. Save the prompts that work — that's your agent's playbook.
Anything repetitive, web-based, and rule-following is fair game. Start here:
Read across tabs and hand you one clean summary with sources.
Complete repetitive web forms and applications from your details.
Find options, compare prices and reviews, and shortlist the best.
Find times, check availability, and tee up bookings for your okay.
Pull data off pages into a tidy table or spreadsheet.
Watch a page or listing and tell you the moment something changes.
Read a thread or page and draft the response in your voice.
Click through a flow like a user and flag what breaks.
Describe the outcome and the constraints. Let the agent figure out the clicks.
Tell it where to pause — before paying, sending, or deleting anything.
Prove it on a low-stakes task before handing over anything that matters.
Review the first run and approve big moves. Trust is earned per task.
Every prompt that nails it becomes a reusable recipe for next time.
You stop doing the task. You start training the thing that does the task.
Add Claude for Chrome, give it your first goal, and watch it go.
Open your portal