A conversational interface can make a consultation easier to begin, but it is not the workflow on its own. The useful work starts when answers are captured in a consistent shape, calculations are made visible, follow-up has an owner and the client receives an outcome that can be checked.
What the published project demonstrates
The Digital Nutritionist for Happy Healthy Nutrition digitised one clinical nutritionist's consultation method into a conversational assistant and companion Recipe Hub. The published case describes custom interface extensions, a calculation engine for BMR, TDEE and macros, lifecycle automation, personalised report generation and a Recipe Hub with dynamic serving adjustment and shopping lists.
Its documented stack includes a Voiceflow conversational assistant, custom JavaScript functions, Make automations, Airtable and a custom web application. The case reports that manual work per client fell from up to four hours to fifteen to twenty minutes. That is evidence about this implementation and practice, not a result to assume for another consultation process.
Separate the conversation from the decision logic
A proposed build should first distinguish what the person is asked, what the system calculates and what a practitioner must decide. That separation prevents a friendly conversation from quietly becoming an unreviewed decision-maker. It also gives each layer an owner when the process changes.
- 01Map inputs before screens. List each answer, its allowed format, whether it is optional and where it will be used. Include how missing or contradictory answers are handled.
- 02Make calculation rules inspectable. Define the inputs, formulae, rounding and exceptional cases with the qualified person responsible for the method. Test representative scenarios before connecting downstream automation.
- 03Choose an explicit outcome. Decide whether the workflow creates a draft report, a task, an update to a record or something else. The recipient and review point should be clear before it is sent.
Design the integration hand-offs
Voiceflow, Make and Airtable can be useful parts of a workflow, but naming tools does not resolve data ownership. In discovery, agree which system holds the current client record, which events trigger automation, what data is written back and how duplicates are prevented. Treat credentials, permissions and error notifications as part of the operating design rather than an implementation detail.
For a consultation workflow, test the ordinary path and the awkward ones: a partially completed conversation, a changed answer after a calculation, an unavailable downstream service and a report that requires professional review. A practical release checklist also names who can pause the flow, correct a record and respond to an exception.
Keep professional judgement visible
The published nutrition project digitised a specific practitioner's method. A different clinic or advice process needs its own qualified review of the questions, calculations, language and approvals. Automation can prepare consistent inputs and drafts, while the accountable professional remains responsible for advice and decisions.
A consultation workflow earns trust when every hand-off is clear: what the person said, what the system calculated, what was automated and what still requires professional judgement.
Start with the workflow you can inspect
If your consultation process is spread across notes, spreadsheets and follow-up messages, start by tracing one complete client journey. We can help identify a bounded first workflow, the existing systems it must connect to and the review points that should remain with your team.
