AI & Data Use Notice

1. The short version

• Your agents' actions are decided by deterministic rules, your policies and your people. No AI model makes, scores or influences a decision unless your workspace turns on an optional AI feature
• Every optional AI feature is off until an owner or admin of your workspace turns it on, and one switch turns them all off
• Intended does not send your data to any third-party AI provider. The only third-party AI processing happens when you connect your own provider key
• Intended improves its products using its own models, which it runs in its own cloud environment, on de-identified and aggregated data, within the contribution settings you control
• We do not sell your data, and we do not let AI providers train on it

2. Decisions are deterministic by default

When your agent asks to act, Intended classifies the request and evaluates it against your policies using deterministic rules. The same request, under the same policies and the same rule-set version, gets the same decision. Every decision receipt records which classifier, classifier version and rule-set version made the decision, so you can see exactly what was applied.

Where you turn on an optional AI feature in the decision path, the model can only suggest. Your policies still make the decision.

3. Optional AI features you control

Each feature below is off by default. Only an owner or admin can turn it on, in Settings → Privacy & AI, and every change to these settings is recorded in your workspace's hash-chained audit log.

• AI-assisted classification of plain-English requests: the request text is sent to the AI provider you choose, using your own API key, which Intended stores encrypted for your workspace. Your agreement with that provider also applies. Intended never uses its own key for this; without your key, requests are classified by the deterministic rules
• AI policy drafting: your plain-English description is processed by a model Intended runs in its own cloud environment. A person reviews the suggested rules, and nothing is saved or enforced unless a person saves it. Intended does not store the description or the draft
• Block all AI: one switch turns off every optional AI feature for your workspace, and it overrides every other AI setting
• Per request: a request that carries the X-Intended-Disable-LLM header is handled without any AI model, whatever your workspace settings say

4. No third-party AI provider receives your data from Intended

Intended does not send your data to OpenAI, Anthropic or any other third-party AI provider, for decisions, for product improvement or for any other purpose. Models Intended uses run in Intended's own cloud environment on Amazon Web Services, which is already one of our listed sub-processors. The only exception is the one you choose: when you connect your own provider key for AI-assisted classification, the request text goes to that provider under your agreement with them.

Intended will give at least 90 days' notice, with a right to object, before any change that would send your data to a new recipient or process it in a new location.

5. How Intended improves its products

Intended uses machine learning, including models it trains and runs itself, to make classification more accurate and to build product intelligence. Improvements reach the default decision path only as new, versioned deterministic rule sets, never as an AI model deciding for your agents, and every receipt shows which rule-set version was used. What may be used depends on your workspace's contribution setting:

• Your workspace only: your own records may be used to improve how the Services work for your workspace. This data stays within your workspace
• Contribution to improve Intended for all customers: de-identified structural signals only, with your workspace, people and agents removed, such as Open Intent codes, decisions and outcomes, argument names and timing. Argument values are never stored, so they are never used. This setting is on by default for Free and Platform workspaces and off by default for Enterprise workspaces, and an owner or admin can change it at any time with one switch
• Prompts: never used unless an owner or admin separately turns on including redacted prompts. That setting is off by default on every plan. When it is on, prompts are redacted again before use and are processed only on models Intended runs itself
• Anything beyond this, such as using identifiable examples from your records: only under a separate written agreement with you

6. How long data is kept

• Authority decision data, including any AI classification result, follows your plan: 30 days on Free, 1 year on Platform, and up to 7 years on Enterprise
• Records Intended derives from a decision to improve how the Services work for your workspace are deleted on the same schedule as the decision they came from
• Redacted prompts are kept only if your workspace turns on raw prompt logging, encrypted under a key specific to your workspace, and are deleted with the decision they belong to
• De-identified contributed records are kept for no more than one year. Models, rules and aggregate statistics built from them are not linked to your workspace and may be kept
• If a legal hold is active for your workspace, nothing is deleted while the hold is in place

7. Your choices

• Turn any optional AI feature on or off, or block all AI, in Settings → Privacy & AI
• Turn contribution and prompt inclusion on or off in the same place
• Send X-Intended-Disable-LLM on any request you want handled without AI
• Ask us to export or delete your data as described in our Privacy Policy

8. Contact

Questions about this notice: privacy@intended.so. Questions about our terms: legal@intended.so.