AI & Data Use
Version 2026-09-25.2
Draft — attorney review required before commercial launch. Bracketed items are still being filled in.
The short version
- Your data never trains AI. Not ours, not Amazon's, not Anthropic's, not anyone's.
- AI runs inside Amazon Web Services (AWS), through a service called Amazon Bedrock, in the United States.
- AI only suggests. People decide. Nothing an AI suggests takes effect until a person reviews and confirms it. AI never decides anyone's pay, job, or performance.
- AI sees only what a feature needs, described feature by feature below.
1. Where AI is used today
| Feature | What it does | What the AI sees | Who confirms |
|---|---|---|---|
| Setup guide | Interviews the owner or office manager and drafts the office setup: roles, what to track, daily minimums, win-the-day rules, sales goals, lead sources, and pay and credit rules | The setup question, the owner's typed answer, and the current setup draft (office name, time zone, role names, activity names, goals, and rules). It doesn't see team member names, pay amounts, sales, customers, or financials unless the owner types them into an answer. | The owner or office manager reviews every step and applies the setup. Nothing changes until they do. |
| Help chat | Answers "how do I…" questions about using the app | The question, the last few chat messages, and our help articles. It never sees agency data. | Answers are information only; the AI can't change anything. |
If AI isn't available, both features fall back to a built-in guide that runs entirely inside the app, and nothing is sent to an AI service.
Future AI features (for example, reading an uploaded commission schedule) will be added to this table before they're turned on, with what they see and who confirms their output.
2. Where the AI runs
- Service: Amazon Bedrock, part of AWS, the same cloud that hosts AgencyHuddleHQ.
- Model: Claude, made by Anthropic, running inside AWS.
- Location: AWS regions in the United States. Requests may be routed between US regions for capacity, but never outside the United States.
- Encryption: every request is encrypted in transit (TLS 1.2 or newer).
3. No training on your data
- We don't use your data to train, fine-tune, or improve any AI model, and we don't let anyone else do so.
- AWS states that Amazon Bedrock doesn't use prompts and responses to train models and doesn't share them with model providers such as Anthropic.
- We keep Bedrock's optional prompt logging turned off.
- We don't build our own AI models from customer data.
This promise is also in our Terms of Service and Data Processing Addendum, so it's a contract commitment, not only a policy.
4. People review every AI suggestion
- Suggestions are drafts. The setup guide shows each proposed change in plain words. The owner or office manager can edit, undo, or ignore it, and nothing is saved to the live setup until they apply it.
- The AI never invents a rate. For pay and commission, it asks, or labels a best guess clearly as a guess for the owner to confirm.
- AI doesn't make decisions about people. It doesn't decide or recommend hiring, firing, discipline, promotion, or anyone's pay. Pay is calculated by rules the owner writes and publishes, and the owner reviews and approves every payroll before anyone is paid.
- AI can be wrong. Check its suggestions against your contracts and records, the same as you would a new employee's work.
5. What's stored
- Setup guide: the conversation is saved with your office's setup so you can pick up where you left off. It's deleted with your agency's data.
- Help chat: we don't store your questions after the answer is sent.
- Amazon Bedrock: doesn't store prompts or responses.
6. Tips for using AI safely
- Don't type customer details, Social Security numbers, or other sensitive information into the setup guide or help chat. They don't need it.
- You can skip the AI entirely: every setup step can be filled in by hand.
7. Automated-decision laws
California's rules on automated decision-making technology (from January 1, 2027) and Colorado's law on automated decision-making in consequential decisions (SB 26-189, from January 1, 2027) apply when automated technology makes, or materially influences, significant decisions about people, such as employment and pay. AgencyHuddleHQ's AI features don't make or materially influence those decisions: a person makes every decision, using rules they set. If we ever add a feature that could, we'll update this page first and put the required notices, opt-outs, explanations, and human review in place before it's turned on.
8. Questions
Email drewtekellcareers@gmail.com or make a privacy request. See also our Privacy Policy.