Product

Eight building blocks. One auditable documentation flow.

The product interface connects speech, medical structuring, mission logic and human approval. Planned integrations are clearly marked as target architecture.

Eight building blocks

Eight building blocks. One record.

Every block is designed as assistance — it prepares, but it does not decide.

Live transcription

In development

Hands on the patient — not on the pen.

Whisper-based speech recognition is planned (target architecture) for medical vocabulary and low latency. The on-device vs. cloud-fallback split will be assessed during the pilot.

Medical AI analysis & NER

Assistance

Turning speech into medical entities.

Relevant statements on complaints, vitals, medications and allergies are detected, given context and prepared as suggestions.

Field mapping

Concept

From words to a protocol.

Detected content is sorted into DIVI, MIND or provider-specific fields. ICD/SNOMED references stay hints, never an autonomous diagnosis.

XABCDE & SAMPLER

Product logic

Structure for everyday missions.

Initial assessment and history follow familiar schemes, so teams don't have to learn a new way of thinking.

Measures timeline

Prototype

Whatever happens gets a timestamp.

Measures, medications and changes appear as a time-based sequence and stay traceable for handover and review.

Vitals & ECG context

Target architecture

Readings without retyping.

Vitals and ECG context can feed into the record. Hardware integrations are target architecture and are not claimed as finished.

Live translation

Prototype

Make the history understandable.

Multilingual communication supports history-taking; relevant answers can be carried into the record in a structured way.

Human-in-the-loop review

Principle

Responsibility stays with the team.

No suggestion becomes a record entry without active confirmation. AI assists, humans decide.

Workflow

Six steps from the moment on scene to a structured record.

Each card flips: the moment on scene on the front, the product logic on the back.

  1. Start recording

    One tap starts capture alongside care. The interface stays usable for mobile teams and with gloves.

    • iPad-native operation
    • robust capture
    • offline mode as target architecture
  2. Live transcription

    Spoken information is turned into traceable raw text. The target is latency in the seconds range.

    • medical language
    • time reference
    • visible origin
  3. AI analysis & NER

    Medically relevant content is detected, segmented and prepared with context and confidence.

    • entity recognition
    • context marking
    • confidence hint
  4. Field mapping

    Suggestions are sorted into the right mandatory and provider-specific fields.

    • DIVI / MIND
    • provider-specific
    • no autonomous diagnosis
  5. Human-in-the-loop review

    The team actively reviews, corrects, discards or confirms every suggestion.

    • review
    • correct
    • approve
  6. Export & sync

    Approved data is prepared for handover and existing systems. FHIR/HIS is target architecture.

    • handover summary (ISBAR)
    • data minimisation
    • FHIR/HIS planned

Prototype

See how AESKONdocs works during the mission.

Prototype · fictional mission data · no sound