All work
2026
Build the steps once, run them every time
A visual workflow tool I built because transcription kept getting specialist terms wrong. Connect the steps once, and every recording gets the same care.
Own tool
Workflow design
Demonstration
Client
Own project
My role
Product, design and build
Timeline
About 3 months
Year
2026
I built Nodio because transcription kept getting specialist terms wrong, especially in healthcare work: drug names, conditions, trial terms. A transcript that looks finished but misspells the drug isn’t finished. Nodio started as a fix for that, and became a way to build the whole job once and run it the same way every time.
The challenge
Transcription tools are good at everyday speech. Specialist words are where they slip. A drug name becomes a common word, a trial code loses a letter, and a person’s name gets spelled three different ways. The transcript looks finished, and nothing tells you which words to check. Everything after that, like tidying, summarising and translating, usually happens in other tools, by hand, and a little differently each time.
The approach
Nodio is a canvas. Each step is a card: bring in the audio, transcribe it, correct it, summarise it, translate it, export it. You connect the cards into a workflow, run it and save it, and every recording goes through the same steps.
Accuracy is built in at every stage.
Before: describe the recording in plain words, and give the engine your terms. They can come from your own list, from AI pulling them out of a reference document, or from a ready-made pack for fields like oncology, cardiology, clinical trials, law and finance. Custom spellings fix words that are always written a particular way.
During: speakers are labelled, the language is detected, and long recordings are split and joined back together. To transcribe just one section, drag across the waveform.
After: an AI correction step checks specialist terms, names and grammar against your vocabulary and reference documents, and highlights every change it made. Words the engine was unsure of are listed, so a person knows where to look.
More than a transcript. Nodio pulls out what was talked about: the topics, chapters and summaries, and the people, places, organisations and products mentioned. It shows who spoke and for how long, how the conversation passed between people, and how the tone changed over time, drawn as a timeline. It can also flag how people seemed to feel as they spoke, such as frustrated, pleased or upset. That reading comes from what they said, so it’s a guide rather than a verdict.
Privacy controls are part of it. Nodio can redact 40 kinds of personal information, from names and phone numbers to health conditions and account details. It can also make a copy of the audio with those details beeped out. Before any text goes to AI, it scans for personal details such as Medicare and tax file numbers, and asks you first.
Clean at the surface, detailed on demand. That was the design rule. The canvas shows just enough to recognise each step. Click a card and its full settings open beside it, with the cost of each feature shown as you switch it on. Templates show a likely cost range before you start, and each run shows what it actually cost.
Built with Claude Code.
The outcome
I use Nodio in my own work. It runs today as a working demonstration, with templates for interviews, podcasts, research and transcripts in several languages. Accounts and background processing for long batches are built and waiting on launch work. After that come a step for a person to review and approve results, an audit trail, shared team libraries, and turning results into finished reports.
The longer aim is a library of steps. Each card does one job, so adding a card adds a capability, and teams build their own workflows from the cards without touching code. Whether it becomes a product for other teams is still open.
A demonstration build. The privacy controls are there, but it isn’t yet set up for sensitive data.
Next project
Bringing on new sellers without the chasing
Based in the Blue Mountains
Open to contract work
If you’re working out where AI fits in your team, I’d like to hear about it.
Email or LinkedIn is the easiest way to reach me.