A Grok bot when that is the current best door.

What you leave with

• A production bot on a defined loop, not a sandbox chat.
• A starter fact graph with proposed, accepted, and ask states.
• A written permission model and a runbook.
• One instruction session for the people who will govern it.
• Thirty days of correction after launch.

How the install runs

1. 1.Name the loop. Write the repeating work in one sentence and name the human who can reject the output. If you cannot name the loop, do not buy seats or start an install.
2. 2.Run a Field Survey. Spend two to three weeks mapping the loop, the tools you already pay for, and the facts a bot would need. Helixfield prices this diagnostic at $9,500. It does not include a production agent.
3. 3.Choose the Grok door — if it is the current best fit. Pick Grok Business seats, a Cursor bot computer, the xAI API, or a mix from the loop — not from a demo. If Astra, Claude, or Gemini is the better door, that is the install. Many companies end up mixed.
4. 4.Seed the fact graph. Write the facts, rules, systems, people, and open questions the bot must pack. Label each proposed, accepted, or ask. Chat history is not the graph.
5. 5.Wire only the tools the loop needs. Connect the applications the job actually touches. Write permissions down. A bot that can click every app is a liability.
6. 6.Evaluate on live work. Score the bot on real jobs, not a scripted demo. Missing facts become asks. Invented numbers fail the evaluation.
7. 7.Teach the owner. The owner learns four verbs: pack, act, extract, ingest. They must be able to reject a bad fact and refuse an outbound draft.
8. 8.Accept the handoff. A named human accepts the system before it touches clients. Client-facing output stays drafted until someone says send. Thirty days of correction follow launch.