Winning by Design · AI for GTM

Build one thing you will actually use this week.

Pick a lane, copy the prompt into whatever assistant your company pays for, and stop at whichever level gets hard. Every lane produces the same five part shape, which is what makes it something you can hand to a teammate.

Step 1 · Pick your lane

Choose the one attached to real work on your calendar.

Step 2 · Copy the base prompt

Your daily brief

Base prompt

  
Standing guardrail, already included: never invent facts. If a source is missing, say so and tell me how to find it. Keep outputs concise and decision oriented.
Step 3 · Go as deep as you can

The same prompt, run three ways.

The shape

Whichever lane you picked, you are building these five parts.

1
Trigger
What makes this run. Every morning, before a call, after a call.
2
Inputs
What it needs from you or your connected sources.
3
Steps
What it does, in order, so the output is consistent.
4
Output
The exact format you want back. Be specific.
5
Guardrails
What it must never do. Start with: never invent facts.
Any assistant

Same concept, four different words for it.

What we meanClaudeChatGPTCopilotGemini
A saved workspaceProjectProjectNotebookGem
A named thing you invokeSkillCustom GPTAgentGem
A wired sourceConnectorConnectorGraph connectorExtension
Long running researchResearchDeep researchResearcherDeep research
Nothing on this page requires one particular vendor. If your company runs Copilot, everything here works, the buttons just have different names.
Two habits that fix most bad output

AI slop is an input problem.

Habit oneShow it what good looks like
Paste in your own writing from before you had AI. Real emails, real proposals, real call notes. It cannot sound like you if it has never read you.
Habit twoMake it interview you first
Add this line to any prompt: Ask me one question at a time until you have what you need, then draft. One instruction, and the quality of everything downstream changes.