Search results
53 results
Sort by:
Write the Naming Convention Down
Put the naming rule in one file the bot can read every run, then link it from your guides hub.
One Pattern Per Object Type
Give folders, files, tickets, and CRM records each one naming pattern, then leave it alone for 30 days.
Correct Bad Names the Same Day
Rename bad bot names immediately, log the bad example, and keep one canonical path.
Bots Research, You Approve
Split research rights from spend rights: the bot reads and compares, you keep payment and account changes. Put the split in standing rules.
Save the Receipt Path
After you buy, save the confirmation into a known folder with a clear name and tell the bot that path, so returns and warranties are easy.
Checkpoint Notes Every Slice
After each slice of a long bot job, save a short status note with done, next, and blockers, under a boring filename so the next run can resume.
Write a Resume Brief
Close a long bot session with a resume brief, not a goodbye: list the next three actions and any secrets the bot must not invent.
One Status Channel Per Bot
Give each team bot one home channel for status and failures. When alerts scatter across DMs, the outage hides from the team.
Rotate a Human Owner
Rotate a weekly human owner for each team bot so it is never orphaned, and have that person update standing rules after repeat failures.
Pair Each Bot with a Human
Put a named human next to every bot job in the runbook. When quality slips, that person updates the bot's instructions the same day.
Track Hours Returned
Once a week, note the hours a bot gave back versus the hours you spent steering it. If steering wins, shrink the bot's scope.
Sunday Family Brief
Once a week, have the bot draft a short household brief. You edit the tone and anything private before it goes to the family channel.
No Portal Passwords in Chat
A family bot may remind you to check a school or bank portal, but it should not hold those passwords in chat. Keep logins human-operated.
Study from Your Sources
Feed the study bot the notes and materials you are actually assigned, not generic web guesses, and keep source titles next to each summary.
Keep an Error Log
Log the questions you miss in a simple file the study bot can reopen next session, and drill those items before new material.
Check License Before Install
Before the bot adds a dependency, have it report the license and last release. A 30-second check skips mystery packages and later messes.
Keep a Credits Section
Maintain a short credits list of the open source tools your bot stack relies on, kept beside the runbook and updated when dependencies change.
Read the CRM Before Outreach
Have the bot open the contact record before it drafts outreach. If the record is missing or stale, it should ask you first.
Log Fields, Not Blurbs
Require contact, date, channel, and next step on every bot-logged CRM activity. Reject vague blurbs nobody can report on.
Failure Review Twice a Week
For a new team bot's first two weeks, review its failures twice a week. Each repeat failure earns one new standing-rule line.
Shrink Scope Until Green
If a bot is flaky, cut its scope before adding more integrations or tools. A narrow job that works beats an ambitious one that fails.
Keep a QA Checklist the Bot Must Run
Give the bot a short fixed QA checklist: happy path, empty input, bad input, permission denied, and one edge case. Pass or fail every line.
Apply Safe On-Site SEO Fixes
Safe SEO fixes a bot can apply itself: titles that match the page, honest meta descriptions, internal links from hubs, sitemap and robots hygiene.
Project Status Bot
Collect updates and produce status summaries and escalations without chasing people all day.
SOP / Documentation Bot
Turn repeatable work into maintained procedures bots and humans can both follow.