Personalising Healthcare Guidance with Conversational AI

Ask most people how they'd find the right surgeon for a specific condition, and the honest answer is usually "Google it and hope for the best," followed by a string of phone calls, referrals, and guesswork. Healthcare guidance, especially for something as significant as surgical care, still runs on a process that hasn't really changed in decades, manual triage, generic search results, and a patient left to piece together whether a recommendation actually fits their situation.

We set out to build something that could genuinely improve on that, a platform where a patient could describe what's going on, get asked the right follow-up questions automatically, and land on a recommendation, doctor, procedure, or care plan, that's actually relevant to them. Here's how we approached it, what made it hard, and what came out the other end.

Why Healthcare Recommendations Are a Different Kind of Problem

Medical guidance isn't like recommending a product or a service. Get it wrong, and the cost isn't a bad review, it's a patient pursuing the wrong procedure or the wrong specialist entirely. That raised the bar for everything in this build, the accuracy of the assessment logic, the reliability of the data behind every recommendation, and how carefully the system had to handle situations it genuinely couldn't answer.

What We Set Out to Build

The goal was a platform that lets a user select the part of the body they're having an issue with, then walks them through a conversational, AI-driven assessment that asks dynamic follow-up questions based on their specific symptoms, not a generic questionnaire. From there, the system needed to recommend suitable procedures, relevant doctors, medical devices, and a personalised care plan based on the full picture it had gathered.

Under the hood, the platform runs on LangGraph and LLMs for the conversational assessment logic, with Python and FastAPI powering the backend and PostgreSQL securely handling user data, assessment history, and recommendations. React.js drives the interactive front end. Beyond the patient-facing side, the platform also needed doctor registration with NPI-based verification, admin approval workflows, subscription management, and dynamic content management, essentially a full healthcare ecosystem built around Healthcare software development solutions, not just a chatbot bolted onto a website. 

Why Building This Was Harder Than It Looks

A conversational health assessment sounds simple until you actually try to build one that's both flexible and reliable. A handful of genuinely tough problems showed up early on.

Managing a Chatbot That Actually Adapts

Symptom-based conversations aren't linear. A follow-up question for a knee issue looks nothing like one for a shoulder issue, and the logic needs to branch intelligently based on what the user has already said, not follow a fixed script. Getting that conversational flow to feel natural while staying medically coherent required serious workflow orchestration, not just a decision tree.

Keeping Symptoms Tied to the Right Body Part

Left unchecked, users could select symptoms that had nothing to do with the body part they'd chosen, which would quietly corrupt the accuracy of everything downstream. Preventing that mismatch without making the interface feel restrictive or confusing took careful design.

Handling the Cases the System Couldn't Answer

No healthcare database covers every condition or procedure that might come up, and pretending otherwise would have been worse than admitting a gap. The system needed a graceful way to handle conditions or recommendations that simply weren't available yet, without the user hitting a dead end or a broken flow.

A Landing Page With More Moving Parts Than It Looked Like

The platform's landing page needed multiple nested content blocks and sub-sections, all of which had to stay easy to manage and update. What looks like a simple homepage from the outside was, structurally, a genuinely complex content management problem underneath.

Verifying Doctors, Not Just Listing Them

Letting anyone register as a doctor on a healthcare platform isn't an option. The system needed real validation, NPI-based verification and an admin approval layer, so every recommendation pointing to a doctor was pointing to someone genuinely credentialed.

How We Solved It, Piece by Piece

Each of these problems needed its own targeted fix, built to work together as one coherent system rather than isolated patches.

• Dynamic AI chatbot workflow. Using LangGraph-based conversational workflows, the platform manages follow-up questions dynamically based on the selected body part and symptoms, demonstrating how custom AI development services can support personalised healthcare workflows. With separate nodes and prompts built for different medical conditions so each assessment stays personalised and context-aware rather than generic.
• Accurate body part and symptom mapping. Symptom selection is restricted to only what's relevant to the body part a user has chosen, which directly prevents the mismatched, inaccurate assessments that would otherwise throw off every recommendation downstream.
• Unsupported condition handling. When a user's condition or procedure isn't available in the system yet, the query gets automatically escalated to the admin panel with full details attached, keeping the user's flow intact while giving the platform a clear path to expand its coverage over time.
• Dynamic content and UI management. A dedicated admin-controlled content management section handles the landing page's nested blocks and components, turning what could have been a maintenance headache into something the team can update without touching code.
• Doctor verification and approval workflow. Validation rules paired with an NPI verification step and admin approval process ensure every doctor profile on the platform is authentic before it ever reaches a patient's recommendation.

Together, this gave the platform a genuinely intelligent core, one where the AI assessment layer, the data validation, and the admin oversight all work as a single connected system instead of separate features stitched together.

What Changed Once the Platform Was Live

The result is a platform that noticeably shifts how healthcare guidance actually happens, from something manual and generic to something automated, accurate, and personalised.

• Faster, smarter patient assessments. The AI-driven chatbot lets users complete a full medical assessment through a natural, dynamic conversation, cutting down the manual consultation effort it would normally take to reach the same point.
• Improved recommendation accuracy. Body-part-specific symptom mapping combined with intelligent workflow handling means the doctor, procedure, and care plan recommendations users receive are actually relevant to their situation, not generic suggestions.
• A genuinely personalised healthcare experience. Every care plan, doctor match, and procedure suggestion is shaped by the user's actual responses and condition, which makes the entire journey feel considerably more tailored than a standard search-and-guess approach.
• Centralised healthcare management. Every assessment, recommendation, doctor profile, and user interaction flows into a centralised database and admin dashboard, giving the platform real visibility and control rather than fragmented records.
• A more efficient workflow overall. Automating the chatbot assessments, doctor verification, and recommendation logic cuts down manual effort significantly, freeing up time for both users navigating the platform and administrators managing it behind the scenes.

What This Project Demonstrates

This platform is proof that AI-driven healthcare guidance doesn't have to mean sacrificing accuracy or trust for the sake of automation. When the conversational logic, the data validation, and the verification workflows are engineered to work together, the result is a system capable of genuinely smarter, safer healthcare decisions, not just a faster version of the same old guesswork.

If your organisation is exploring how AI could realistically fit into a healthcare workflow without cutting corners on accuracy or compliance, this is exactly the kind of challenge our team likes solving. Get in touch with Dotsquares to talk through what an AI-powered healthcare platform could look like for your business.