AI & Adoni Terms

1. What Is Adoni

Adoni is Poll City’s built-in AI campaign assistant. He is available from any page in the Campaign OS and is designed to help campaign managers with planning, drafting, analysis, and answering questions about running a campaign.

Adoni is powered by Claude, a large language model developed by Anthropic PBC. When you interact with Adoni, your messages and relevant campaign context are sent to Anthropic’s API for processing. Anthropic’s responses are then returned to you through the Poll City interface.

2. Anthropic as Sub-Processor

Anthropic PBC operates as a sub-processor of Poll City when processing data through the Adoni feature. This means:

• Anthropic processes the data you send through Adoni under Anthropic’s own Privacy Policy and Terms of Service. By using Adoni, you accept that Anthropic’s terms apply to their processing.
• Anthropic does not receive raw personal data about your contacts. See Section 4 for details of what is and is not sent.
• Anthropic may retain prompt data as described in their privacy policy. Poll City does not control Anthropic’s data retention practices.
• Data processed through Anthropic’s API may be processed on servers outside Canada. Anthropic operates primarily in the United States. By using Adoni, you accept this cross-border transfer.

3. What Data Is Sent to the AI

When you use Adoni, the following types of information may be included in the request sent to Anthropic:

• The messages and questions you type in the Adoni chat interface
• General campaign context (campaign name, election type, geographic area)
• Summary statistics (e.g., number of canvassed doors, volunteer count, aggregate poll results)
• Draft content you ask Adoni to review, improve, or respond to

What is not sent to Anthropic (see Section 4 for how this is enforced):

• Individual voter names, emails, or phone numbers from your CRM
• Voter support level ratings for identified individuals
• Donation records for identified donors
• Raw voter list data

4. Data Sanitisation

All campaign contact data that passes through the Adoni feature is sanitised before transmission to Anthropic. This means:

• Names and contact details of individual voters are stripped or replaced with placeholders before being included in any prompt context.
• User-supplied text is screened for personally identifiable information before transmission.
• If you manually type a voter’s personal information into the Adoni chat, that information may be transmitted to Anthropic as part of your message. Poll City recommends not including specific personal data in Adoni conversations.

5. Review Before Use

AI responses are generated automatically and must be reviewed by a human before use in any campaign communications or public-facing materials.

You are responsible for reviewing and approving any AI-generated content before:

• Sending it as a campaign email or SMS message
• Publishing it as campaign social media content
• Using it in any official campaign materials
• Presenting it as factual analysis to the campaign

Poll City is not responsible for the consequences of using AI-generated content without review.

6. No Professional Advice

Adoni does not provide legal advice, electoral law advice, financial advice, accounting advice, or any other professional advice. Responses from Adoni are general-purpose information generated by a language model and should not be relied upon as a substitute for professional advice from a qualified lawyer, accountant, or campaign compliance officer.

In particular: questions about elections finance limits, CASL compliance, campaign spending rules, and electoral law should always be directed to a qualified professional or to the relevant elections authority. Adoni’s answers on these topics may be inaccurate or outdated.

7. Accuracy & Limitations

Large language models can produce inaccurate, outdated, or misleading information — including on factual matters that appear straightforward. Poll City does not guarantee the accuracy, completeness, or fitness for purpose of any AI-generated content.

Known limitations you should be aware of:

• Knowledge cutoff: The model may not know about recent events, by-elections, boundary changes, or new legislation.
• Electoral law variation: Rules differ across municipalities, provinces, and federal elections. The model may give a general answer that does not apply to your specific jurisdiction or race.
• Hallucination: AI models can produce confident-sounding but factually incorrect statements, including fabricated statistics or misquoted laws.

8. Contact

Questions about Adoni, AI data processing, or these terms should be directed to: