# Lauren asks until the report holds up. > Lauren is IncidentKit's AI assistant. She asks follow-up questions, fills in the report and drafts contributing factors. Every field she writes stays a draft until a person approves it. Source: https://incidentkit.ai/product/lauren · Updated Oct 5, 2026 **Lauren, the AI assistant** (report): Tell Lauren what happened. She asks, fills the form, and a person signs. ## What it does - **Follow-up questions, not a 14-field form** She asks one or two questions at a time, and stops when a reviewer would be satisfied. - **Fields filled as you talk** Date, time, place, people, harm, near-miss status and type fill in beside the chat. - **Drafts are marked as drafts** Every AI-written field and suggested action says “Lauren · draft” until a person approves it. - **A person signs** Nothing becomes the record until the reporter attests and a reviewer approves. Lauren's write actions need approval too. - **Investigation drafts** Lauren drafts contributing factors, a five-whys chain and corrective actions for the investigator to edit. - **Facility memory and weekly digests** Lauren can remember approved room names, equipment and routing choices, and send leaders a weekly risk digest. - **Voice and more languages** (rolling out) Voice intake and Spanish and other languages are rolling out. ## How it works 1. **Describe what happened** Type it in plain words. 2. **Answer two or three questions** Lauren asks for what is missing: place, harm, equipment, witnesses. 3. **Review and sign** Check the fields, fix any, and attest. 4. **Route and investigate** People are told by severity. Lauren drafts the investigation. ## What Lauren does. What she never does. **What she does** - **Asks like a risk manager.** One or two questions at a time, and only what is missing. - **Fills the report.** Date, time, place, people, harm and type, from what the reporter said. - **Drafts the investigation.** Contributing factors, a five-whys chain and corrective actions to edit. - **Marks her work.** Every field she writes says “Lauren · draft” until a person approves it. - **Remembers your site.** Approved room names, equipment and routing choices. - **Sends a weekly digest.** A short risk summary for the people who lead the site. **What she never does** - **Sign for a person.** The reporter attests. A reviewer approves. - **Change the record alone.** Her write actions need approval too. - **Decide what happened.** Her investigation drafts are suggestions, not conclusions. - **Close an action.** Closing takes a person and proof. - **Train on your data.** On Regulated and Network, her AI provider may not. - **Work in secret.** On Regulated and Network, her inputs and outputs go to the audit trail. ## Four steps from a text to a signed report. 1. **She reads what was said.** The reporter writes it the way they would say it. There is no form to learn. Example, T. Okafor · CNA: “Found Mrs. K on the floor between the bed and the bathroom at 2:47. She's awake and says her hip hurts.” 2. **She asks only what is missing.** She stops when a reviewer would be satisfied, not when a long form is full. Example, Lauren: “Is a nurse with her, and has she been assessed? Was the fall witnessed?” 3. **She drafts, and says so.** She fills fields only from what the reporter said. Each one is marked as a draft. Example, Draft fields: “Type: Fall, unwitnessed. Harm: possible injury, being assessed.” 4. **People review and sign.** Nothing becomes the record until the reporter attests and a reviewer approves. Example, Sign-off: “Attested by T. Okafor. Approved by R. Nguyen.” ## A draft is never the record until someone says so. Edit her words and the record keeps both versions, with the name of who changed what. Lauren's draft: “The bed alarm did not sound.” In the record: “The bed alarm was silenced during room cleaning.” Approved · R. Nguyen · 8:02 am. - 7:53 am: Lauren drafted “Contributing factor” from the report - 8:01 am: R. Nguyen edited the text - 8:02 am: R. Nguyen approved it ## A short path, with a contract on it. Your staff (Report by text or QR code.) to IncidentKit (Hosted in the United States.) to AI provider (Signs a BAA. May not train on your data.). - **Identifiers are minimized first.** Direct identifiers are reduced before text reaches the provider. - **Nothing you enter trains a model.** Her provider is not allowed to use it for training. - **Inputs and outputs are logged.** You can see what she was given and what she wrote. - **Open.** For incidents with no patient information. Patient information is not allowed in reports. - **Regulated and Network.** Patient information is allowed, under a BAA that covers the app and the AI provider. ## Plan for wrong. Plan for down. - **If she is wrong.** Language models can make things up. That is why a person reviews every draft before it counts, and why she fills fields only from what the reporter said. - **If you disagree.** Edit the draft or delete it. The record keeps your version and shows who approved it. - **If the AI service is down.** Reporting keeps working. Staff use the standard form or quick report, and Lauren's drafts pick up again later. Rolling out: Voice reporting, Spanish and other languages, Human-authored RCA templates. ## Frequently asked questions ### Does Lauren write the report, or does a person? Lauren drafts it, and a person signs it. Nothing she drafts becomes the record until someone approves it, and the audit trail shows who did. ### Can Lauren make things up? Language models can, which is why every draft is reviewed. Lauren fills fields only from what the reporter said, and marks drafts. Investigation drafts are suggestions, not conclusions. ### Is patient information sent to the AI? Only on the Regulated plan. There, Lauren's AI provider is under a BAA (a HIPAA contract) and direct identifiers are minimized first. The free Open plan does not allow patient information. See the [HIPAA page](https://incidentkit.ai/hipaa). ### What if the AI service is down? Reporting keeps working. Staff use the standard form or quick report, and Lauren's drafts resume later. ### Can I see what Lauren was given and what she wrote? Yes, on Regulated and Network. Her inputs and outputs go to the audit trail, next to the names of the people who edited and approved each field. ### Does Lauren replace the investigator? No. She drafts contributing factors, a five-whys chain and corrective actions for the investigator to edit. The investigator decides what happened. ## Related - [Incident reporting software: forms, drafts, workflow](https://incidentkit.ai/product/incident-reporting) - [QR code quick report for incidents and near misses](https://incidentkit.ai/product/quick-report) - [Incident investigations and root cause analysis](https://incidentkit.ai/product/investigations) - [Incident audit trail: every change, who and when](https://incidentkit.ai/product/audit-trail) - [AI for incident reporting: what it can and cannot do](https://incidentkit.ai/guides/ai-for-incident-reporting) - [How AI intake changes incident reporting: forms vs chat](https://incidentkit.ai/blog/how-ai-intake-changes-incident-reporting) - [Incident reporting deadlines and obligations](https://incidentkit.ai/compliance/reporting-deadlines)