Crowd Almanac

AI Product Use

Crowd has optional AI features. This page states exactly what data they use, when, and where it goes — including when they run on real, live event data. Last updated 13 June 2026.

It's optional and off by default

AI features are disabled unless an administrator turns them on for your organization. There is a single master switch. When it's off, no data is ever sent to any AI service and nothing related to AI appears anywhere in Crowd. When it's on, the behavior below applies. An administrator can turn it back off at any time, which stops all of it immediately.

What the AI does

One thing: it suggests how to triage an incident. From a short piece of text describing a situation, it proposes a category (medical, theft, and so on) and a priority (P1–P4), and flags whether it looks like an emergency. That's the entire feature. It does not write reports, summarize records, answer questions, or generate other content.

Every suggestion is advisory. A dispatcher reviews it and decides what to do. The AI never changes an incident on its own — it doesn't set or change priority or status, doesn't create or close incidents, and doesn't send any messages.

Exactly what data is sent, and when

There are two moments when text may be sent to the AI service, and only these two:

  1. When a dispatcher asks for a suggestion. On the "new incident" form, a dispatcher can press Suggest. At that moment Crowd sends the text that dispatcher has typed into the title and description fields, plus the location field if it's filled in. Nothing is sent unless the button is pressed.
  2. When an attendee texts in. If an attendee texts your published number, Crowd may send the text of that message to check whether it looks like an emergency, so it can alert dispatch to a message that needs a fast look. This is used only to raise that alert; it does not create or change any record.

In both cases the data is short free-text. Because it's free-text, it contains whatever the person wrote — which can include personal details if they chose to type them. The AI receives only that text (and, in the first case, the location field). It does not receive your account, the rest of the incident, or anything else.

What is never sent

Where the data goes

Suggestions are generated by Amazon Bedrock, a managed AI service from Amazon Web Services, using the model your administrator has configured (by default a small Amazon model; optionally another model offered through Bedrock). The text is sent to that service only to produce the suggestion and is returned right away.

Inputs sent to Bedrock are not used to train the underlying AI models, and the model provider does not keep them to improve its models. Crowd does not retain a separate copy of the text for AI purposes — the incident you create is stored the same way it would be without AI; the suggestion itself is just shown to the dispatcher.

Running on production data

What "in production" means here There is no separate sandbox or test mode for these features. When an administrator enables AI, it runs on your real, live event data — the actual text dispatchers and attendees enter during the event. Enabling the feature means that text flows to the AI service as described above. If you don't want production data sent to an AI service, leave the feature off.

Accuracy and limits

AI suggestions can be wrong. They are a convenience to speed up triage, not a decision. Always confirm the category and priority yourself before acting, and never rely on an AI suggestion for a clinical or medical judgment. See the emergency-service notice in our Terms of Use.

Turning it on or off

Administrators control the feature under Platform Settings (see Administration). It can be turned on or off for your organization at any time. How personal information is handled generally is covered in our Privacy Policy.

Questions

Email hello@oncrowd.net.