AI Analytics

AI Analytics provides intelligent analysis of your campaign responses, helping you understand patterns, trends, and key insights from participant feedback.

#What is AI Analytics?

AI Analytics uses artificial intelligence to analyze all responses in your campaign and generate a comprehensive report. Instead of manually reading through individual responses, you get:

• Key Findings: The most important insights from your data
• Response Patterns: Trends and distributions in the answers
• Sentiment Analysis: Overall tone of text responses
• Actionable Recommendations: Suggested next steps based on results

#When is AI Analytics Available?

AI Analytics is available for campaigns that meet these criteria:

• Campaign has at least one response
• You are the campaign creator: the request is signed by your account (silently for Google users; wallet users confirm a "sign message" request)

You can generate analysis for both active and ended campaigns. Once generated, the report appears in the campaign's Analytics tab for everyone who is allowed to read the campaign's answers.

#How to Access AI Analytics

1. 1Go to your campaign page
2. 2Click on the "Analytics" tab
3. 3If no analysis exists yet, click "Analyze Responses"
4. 4Read what will be sent, to whom and through whom, and confirm: each participant's answers (without addresses), or totals only. Nothing is sent before you confirm, for any campaign
5. 5Wait a few seconds for the AI to process responses

#What's Included in the Analysis

#Summary

A brief overview of the campaign results highlighting the most important findings.

#Key Insights

Bullet points covering:

• Majority preferences in choice questions
• Common themes in text responses
• Notable patterns or outliers
• Unexpected findings

#Response Breakdown

For each question:

• Distribution analysis (for choice questions)
• Theme extraction (for text questions)
• Notable quotes or responses

#Recommendations

Actionable suggestions based on the data:

• What decisions the data supports
• Areas that might need further investigation
• Follow-up actions to consider

#Saving and Updating Analysis

#Auto-Save

When you generate an analysis, it's automatically saved. You can return to the campaign later and see the same analysis.

The report is stored on Walrus, decentralized storage on Sui, and the campaign records a pointer to it together with the SHA-256 of the file. Only the campaign creator can set that pointer: the smart contract enforces it. Everyone who opens the report gets it verified against the on-chain hash, which is why the page says "Verified" above it. If Walrus cannot take or serve the file, Logbook's cache steps in; it needs no trust, because a file that does not match the hash is rejected.

Publishing takes two confirmations for wallet users (a "sign message" request for the upload and a transaction for the pointer); Google users confirm nothing.

#Regenerating

If new context is needed or you want a fresh perspective:

1. 1Click the "Re-analyze" button
2. 2The AI will create a new analysis
3. 3The old analysis is replaced

#Limitations

• Language: The report is written in the language of the campaign's title and questions
• Response count: Very large campaigns may exceed what the AI can read in one request
• Text responses: Analysis of free-text is based on pattern recognition, not human judgment
• Subjectivity: AI interpretation may differ from human interpretation
• Daily limit: 10 analyses per account and 20 per IP address

#Privacy and Data

#What the AI Sees

• All campaign questions and options
• Vote counts and percentages, and the written answers (text answers and "Other" texts)
• Each participant's answers as a numbered row, without addresses or timestamps, unless you chose totals only. The rows and the written answers are in a shuffled order, not the order of the responses, because that order is public next to each address
• Only answers you can read and that count in the tally (for private campaigns, the ones decrypted in your browser); locked and invalid responses are left out

#What the AI Does NOT See

• Respondent wallet addresses, transactions and the times of responses
• Names from member cards
• Other campaigns or data

A written answer can still name its author, or someone else, by what it says. Read the preview before you send.

#Data Processing

• Analysis is generated on-demand
• Response data is sent in plain text through Logbook's server to a third-party AI provider, DeepSeek (api.deepseek.com, a company in China), only during analysis. Logbook's server does not store it; how long DeepSeek keeps it is DeepSeek's policy, and it is handled under that provider's terms
• Participants are told before they respond that the creator can do this: the notice above the Submit button says so for every campaign
• The finished report is stored on Walrus, with a copy in Logbook's cache, and its hash is recorded on-chain (the file is encrypted for private campaigns)

#Private Campaigns

The saved report of a private campaign is encrypted in your browser before it is stored, for the same readers as the results it summarizes: "Only me" reports only for you, "Voters only" reports for voters, "After the end date" reports for everyone with access once the campaign ends, and reports of "Everyone with access" campaigns for everyone who can open the campaign. On address-list campaigns with "Everyone with access", a report is encrypted for the current access epoch: an address removed from the list cannot read a report generated after its removal. Viewers who may not read the answers see a locked Analytics tab.

One thing encryption cannot change: to produce the report, what you agree to send is decrypted in your browser and passed, in plain text, through Logbook's server to the AI provider. Two parties that the campaign's setting does not name as readers then see those answers. That is why Logbook asks before it sends any answers, and offers totals only. Don't run the analysis if the answers must never leave your browser.

#Best Practices

1. 1Wait for all responses: Generate analysis after the campaign ends to capture all data

1. 2Review critically: AI analysis is a starting point, not a final conclusion

1. 3Cross-reference: Compare AI insights with your own observations

1. 4Context matters: The AI doesn't know background context — interpret findings accordingly

1. 5Iterate if needed: If the first analysis misses something, regenerate with the same data

#Cost

AI Analytics is free during the beta period. After beta, pricing details will be announced.