Domain context
Domain knowledge instead of standard logic
The quality of a result depends on how well the AI understands the domain context: which phrasings are correct, which errors in the dictation must be fixed, how dictation commands are carried out, what belongs in the finished text and what does not.
Tippiti uses rules developed jointly with medico-legal experts and transcription services. These rules are continually refined. This means the output already fits most use cases without further configuration. If you have specific requirements on top of that, you can add your own rules.
- Built with and for transcription services and medico-legal experts
- Medico-legal report language is deliberately preserved
- Dictation commands are recognised and carried out
- Individual rules and text replacements supported
Agentic workflow
Built-in quality control
Tippiti does not work with a single AI pass. In the agentic workflow, several AI models work together, each with its own role in correction[TOOLTIP] Correction means identifying and fixing transcription and dictation errors. [/TOOLTIP], fact-checking[TOOLTIP] The fact-check verifies details that can be checked against a source, including proper names, medication names, ICD codes and place and postcode combinations. Clinical findings and diagnoses come from the clinician or medico-legal expert. [/TOOLTIP] and plausibility checks[TOOLTIP] The plausibility check looks for clear logical errors, further reducing transcription and dictation errors. [/TOOLTIP]. An independent AI validator[TOOLTIP] A different model runs a separate check for defined types of error, including spellings, place and postcode combinations, ICD plausibility and contradictions within the text.
Clinical judgements come from the clinician or medico-legal expert. The validator checks the text against the defined transcription and consistency criteria. [/TOOLTIP] compares the result with the original recording and verifies that the applicable rules were followed and content was carried over correctly. If discrepancies are found, it triggers another correction round until the validation criteria are met.
The workflow checks the spelling of medication names, active ingredients, ICD codes, proper names and place names. It validates codes, matches postcodes to place names and checks dosages for plausibility. When a factual detail requires external verification, Tippiti automatically supplements the check with a web search[TOOLTIP] When a factual detail such as a new medication or rare proper name requires an external source, Tippiti automatically checks it against external references.
The search uses isolated factual terms only, keeping patient names, diagnostic context, dictation content and other sensitive data out of the query (see FAQ). [/TOOLTIP]. It also detects logical contradictions in the text. That goes beyond correcting transcription errors.
If discrepancies are found, Tippiti corrects the affected passages and validates them again. This cycle continues until the result meets the requirements.
With dictations, Tippiti corrects, structures and validates what was dictated and checks the facts. In Assessment interview & case history mode, it writes up the recorded statements as a structured case history and interview section and can also write up a closing summary. Findings, diagnoses and expert opinions come from you. The finished result is available as a Word document, in the integrated editor or via the API. The Privacy page explains how this is covered legally (DPA, German § 203 StGB, no training on your data).
- Correction and validation in separate roles
- Check for rule compliance and factual accuracy
- Fact-check of proper names, medication names, ICD codes and dictated spellings
- Automatic feedback loop until the result passes validation
Customisation
Adaptable where it needs to be
The base rules cover most requirements. Medico-legal experts and transcription services can also have their own conventions for abbreviations, date formats, ICD code notation and formatting.
Tippiti supports instruction groups and text replacements for requirements beyond the base rules. You select the instruction group for each upload, alongside the processing mode. The available modes are Dictation, Conversation & interview and Assessment interview & case history.
- Individual instruction groups per client or speciality
- Text replacements and formatting rules take effect immediately
- Tippsi Chat in the editor for follow-up edits
Engineering
Quality comes from the details
Send an audio file to a model, and you get text. For a usable result, that is often not enough. A reliable result requires many technical decisions: how to optimise the audio, split it at suitable pauses, preserve context between sections, handle long texts and deal with errors in difficult passages.
Each AI model also has its own edge cases and quirks, and Tippiti’s multi-stage workflow is built to handle them. All of these details determine whether you end up with just a result or with a result you can actually work with.
- Audio optimisation and intelligent segmentation
- Reliable processing even with challenging audio quality
- Consistent context even in long texts
- Iterative quality checks across multiple passes
- Automatic web search for fact validation