TRACE is a prototype. Demonstrations use synthetic data. No real case data has been processed.
Take field notes in any of 44 languages. TRACE writes the case record, flags the risks, and shows you what is repeating across your files. In the field. On any phone. With or without a signal.
* Indicative figures, not yet independently validated.
Scroll through a day in the field.
Chapter 1
Start every shift knowing exactly where to focus. TRACE reads your full caseload and surfaces the cases that need attention today, based on risk signals, follow-up timelines, and pattern changes across every file.
Today
Active Caseload
Last contact: 2 days ago
Last contact: 6 days ago
Last contact: 11 days ago
3 of 50 cases need attention today.
Chapter 2
Across 50 files, TRACE catches what no individual caseworker could see. Escalating trauma signals. Recurring profile gaps. Missed follow-up windows. It flags them before they become crises. Pattern detection is demonstrated on a simulated caseload in the current prototype.
Cross-case
Pattern Detection
Reviewed 50 active files this morning.
Chapter 3
After every interaction, TRACE writes the case note. You review, approve, done. The note is captured and saved with no signal, and structured when you reconnect.
Draft ready
Voice Note
Follow-up visit conducted at the transit site. Client reports stable shelter placement and renewed contact with two family members. Protection concerns unchanged since last visit. Referral to livelihoods partner discussed and consented to. Next check-in scheduled in seven days.
Saved offline.
Chapter 4
TRACE drafts the full case note, referral letter, and risk score, in the format your reporting system expects. The caseworker reads, edits if needed, and approves.
Ready for review
Structured Output
Drafted from the field note.
Chapter 5
A caseworker should not have to find a form in the right language before they can write anything down. TRACE accepts field notes in any of 44 languages, including Hausa, Swahili, Somali, Wolof, Bambara, Amharic, Yoruba, Igbo and Lingala, and returns a structured record in any of six languages: English, French, Spanish, Arabic, Russian or Chinese. The interface itself is multilingual. Voice is transcribed live in eighteen languages. In twenty-six more TRACE records audio and says so on screen, per language, before the microphone is tapped. No language is given a near-neighbour speech model. A wrong one returns a fluent transcript in the wrong language rather than an error.
Detected: Somali
Language Support
Transcribing…
“Waxaan ka baqayaa inaan ku noqdo deegaankii aan ka imid.”
Structured in English
Client reports: Fear of return to origin area. Last contact: 3 months ago.
Today
Active Caseload
Last contact: 2 days ago
Last contact: 6 days ago
Last contact: 11 days ago
3 of 50 cases need attention today.
Features
KoBoToolbox and ODK are good at collecting data. Neither thinks with it. TRACE does.
Every morning, TRACE reads your full caseload and writes a prioritized briefing: what's urgent, what's overdue, what patterns have emerged across cases. You start the day knowing what matters. The brief is generated when the device has a connection.
The same broker appearing in three unconnected files. A new trafficking route. Debt bondage rising in a specific population. TRACE surfaces what no individual caseworker can see alone, and flags it for human review before anyone acts on it. Demonstrated on a simulated caseload in the current prototype.
Intake and case management run with no network connection. Cases are stored on the device. Sign-in is not enabled in this prototype, so no case created in the app reaches TRACE's servers today. The server path is built: when sign-in is enabled, cases will also be saved to TRACE's servers, where others in the caseworker's organization will be able to read them. AI structuring queues and completes when a signal returns; the daily brief and the AI assistant need a live connection.
Caseworkers write field notes in any of 44 languages, and a structured record comes out in six: English, French, Spanish, Arabic, Russian and Chinese. No per-language form setup. No language is given a near-neighbour speech model. A wrong one returns a fluent transcript in the wrong language rather than an error.
Risk factors surfaced for caseworker review against six CTDC/IOM trafficking indicators, showing which indicators drove each flag and what to ask next. The indicator set is drawn from published standards, has not been field-validated, and is demonstrated on a simulated caseload in the current prototype.
Referral letters, risk assessments, and reintegration plans. In English, French, Spanish, Arabic, Russian or Chinese. Every document requires caseworker review before it goes anywhere.
Protection, DRR, shelter, health and cash templates aligned to the forms your reporting already uses.
Cases are stored on the device. Sign-in is not enabled in this prototype, so no case created in the app reaches TRACE's servers today. The server path is built: when sign-in is enabled, cases will also be saved to TRACE's servers, where others in the caseworker's organization will be able to read them, access will be scoped to that organization, anonymous access will be refused, every change will be recorded with who made it and when, and a case will be closable but not deletable from inside the app. When a caseworker uses an AI feature, the case content in that request is sent to Anthropic's API to be processed and returned; it is not retained by TRACE. Encryption at rest and structured consent capture are on the roadmap, not in the current prototype.
Learn moreWho it's for
Spend the interview with the person in front of you, not the form. Leave the site with the case file already written.
Consistent, complete records across every team and site. Referrals and follow-ups, traceable end to end.
Donor-ready reporting from real field data, plus a defensible audit trail for protection and safeguarding.
Use cases
Register survivors and at-risk individuals from the caseworker's own dictated notes after the interview. TRACE structures the account, flags protection indicators, and drafts the referral letter before the caseworker leaves the site.
SCENARIO
A protection interview ends with a survivor's full account, but the case file is written hours later from memory and scattered notes. Indicators are missed, timelines blur, and the caseworker begins the next day already behind.
How it works
Most tools create work. TRACE eliminates it.
The caseworker types or dictates in the language the survivor is actually speaking, in the field, with or without a signal. No form has to be built in that language first.
TRACE transcribes, classifies, surfaces risk factors against the CTDC/IOM trafficking indicators used in the anti-trafficking profile, and checks for gaps in the case profile. Risk scoring runs on-device and works with no signal; AI structuring uses a connection when one is available. Nothing is acted on until the caseworker reviews it.
The caseworker reviews everything before any action is taken. TRACE is the thinking partner. The caseworker is the professional.
Compare
TRACE is not a replacement for your data platform. It is the layer between the interview and the record.
| Capability | RecommendedTRACE | KoBoToolbox | ODK | Paper forms |
|---|---|---|---|---|
| AI daily caseload brief | Automatic | No | No | No |
| Cross-case pattern intelligence | Demonstrated on simulated caseload | No | No | No |
| Offline capture and case management | Yes | Yes | Yes | Yes |
| Voice input | Yes, multilingual | No | No | No |
| AI case file generation | Yes, on reconnect | No | No | No |
| Risk flagging | Yes | No | No | No |
| Referral letters | Yes | No | No | Manual |
| Report generation | Yes | Partial | Partial | No |
| Setup time | Minutes | Days | Days to weeks | Immediate, but no data |
The problem, in the field
TRACE is pre-launch. These are the documentation problems we built it to solve, drawn from field practice.
PROTECTION CASEWORK
A caseworker completes an interview in a displacement site and then spends two hours reconstructing the case file from memory and handwritten notes, with detail lost at every step.
MULTI-SITE PROGRAMS
A program manager receives case data from six field sites in four different formats, unable to see patterns across them because nothing is structured consistently.
CONNECTIVITY GAPS
A field team works where there is no signal for days, so every digital tool defaults back to paper, and the paper does not reach the database for weeks.
About
“The constraint in field casework was never funding. It was time, and the cognitive load of documentation while a survivor is sitting in front of you.”
Elke-Esmeralda Dikoume
Founder & CEO, TRACE
Humanitarian specialist with nine years across UNICEF and the World Bank, five of them at the World Bank. She built TRACE for the documentation problem she kept meeting in the field.
TRACE is in active development. We work with field organizations directly. Tell us about your program and we'll figure out together whether it fits.
We reply to every request from a field organization.