Two agents. One AI brain. Built for Dapper.

What the brief asked for.

Three observations from the brief that shape everything below.

1.1 An operating layer, not a chatbot

Dapper's brief is precise: an AI-powered layer that removes friction between physicians, patients, pharmacies, and fulfillment. One coherent surface, met in whichever channel fits the moment.

1.2 Three audiences, three different shapes

1.3 The constraint

A $15-25K prototype budget covers six to ten weeks for a senior pair. Enough for a credible demo across two surfaces and three modalities. Not enough to prototype the physician surface honestly, because the valuable capabilities live behind EHR integration.

What fits the budget.

Two agents, three modalities, one conversational brain. The Physician surface is deferred to Phase 2 on purpose.

2.1 In scope

Two agents on one brain and one knowledge base, served through a Next.js web app with text, voice, and avatar. Patient handles Q&A, onboarding, refill and shipment lookups on mock data, education, and in-app escalation. Front Desk handles SOPs, account walkthroughs, and the same lookups for staff. Same voice, same avatar, same brand across modalities.

2.2 Out of scope, on purpose

2.3 What "prototype" means

A working web app, two agents, three modalities, real RAG over real content, password-protected URL. Credible enough to put in front of physicians, clinics, and investors. Productionization cost is called out item by item above, not folded into the price.

The Patient agent.

Thirty-plus patient capabilities in the brief, narrowed to a focused slice. The agent is the same one whether the patient types, talks, or sees a face.

3.1 The problem

A Dapper patient on a GLP-1 or HRT protocol has dozens of small questions every week. Where is my shipment. What is this side effect. Can I take this with coffee. How do I auto-refill. Each one currently stalls the patient or lands on a human.

3.2 Capabilities in the prototype

3.3 Where it shows up

Text for everyday questions. Voice for hands-free moments, with native barge-in. Avatar for high-trust demos. One brain, one KB, one escalation rule set across all three.

3.4 Out of scope

Payments and PCI. Real insurance, FSA, HSA. CareTeam routing through Slack or Twilio. Adherence tracking on real patient data. Subscription management. All named in the brief, all Phase 2.

The Front Desk agent.

Front desk staff decide whether Dapper feels useful inside a partner clinic. Same brain as the Patient agent, different audience and retrieval filter.

4.1 The problem

A front-desk team fields the same dozen questions every week. Where is this order. How do I create a Dapper account. What's our SOP for X. Answers live in a training manual, in Slack, in someone's head, but rarely where the team can actually use them mid-shift.

4.2 Capabilities in the prototype

4.3 Where it shows up

Same three modalities as Patient, gated to the staff view by URL. Text for desk-bound queries. Voice for hands-busy moments. Avatar as the same brand face the patient sees.

4.4 Out of scope

Real account creation against production systems. Payment collection. Translation. Operational dashboards (Rx queues, incomplete-order monitors, on-call routing). Conversational layer here, data layer in Phase 2.

One brain, three surfaces.

The same conversational agent powers text, voice, and avatar. The same retrieval pipeline grounds every answer. The avatar vendor is a swappable visual layer, not a structural commitment.

5.1 The principle

One ElevenLabs agent across all three modalities. Text and voice are native modes. The avatar consumes the same audio stream. Modality is presentation, not a different brain.

5.2 Layered architecture & request flow

5.3 Why this shape

Consistency. One agent, one persona across modalities. Cost. Two agents on one brain fits the budget. Optionality. Avatar vendor sits at the edge, swappable without re-architecting.

The stack we'd build on.

Eight layers, every choice with a one-line rationale and a documented escape hatch.

Not in the stack on purpose: observability, email/SMS, payments, real auth. Real production needs, real Phase 2 line items.

The team.

A two-person build core with light PM and QA. Tech Lead owns the agent, RAG, and vendor wiring. Fullstack Engineer owns the web app and deploy.

2.1 FTE-equivalent across six weeks. The team that scopes the work is the team that ships it. No staffing churn between discovery and delivery.

Six weeks. One fixed price.

Five milestones, both agents, three modalities. Price fixed at signing. Anything outside the locked scope routes through a change-request.

We've shipped this shape of work before.

Three engagements that map directly onto the Dapper prototype: one near-twin on multi-agent conversational AI, one HIPAA-compliant patient platform at lean-team scale, one AI intake POC delivered in two months.

Multi-agent conversational AI for drug denial & patient education

Client · confidential US healthcare provider · telemedicine, pharmacy benefits

RAG-based voicebot and chatbot, built on ElevenLabs over GCP. Multi-agent flows with PHI-safe orchestration, identity verification, and modular integration with existing EHR and CRM systems. The near-twin to what we are proposing for Dapper.

Challenges

• Patients lacked clarity on drug denial reasons, PGx testing, and insurance coverage.
• Manual, fragmented workflows delayed treatment decisions and risked patient attrition.
• Repetitive Q&A burdened staff, raising operational cost and burnout risk.

Solution

• Multi-agent conversational AI for providers and patients across voice and chat.
• Agentic flows with PHI-safe orchestration and EHR/CRM integration.
• RAG pipelines grounded in PGx reports and SOPs, with tool orchestration for kit requests, appointment scheduling, and PharmD consults.

Impact

• Reduced average handling time for drug denial cases.
• Increased patient clarity on denial reasons and test benefits.
• Improved provider engagement via automated callbacks and multi-contact fallback logic.

ElevenLabsRAGMulti-agentGCPPHI-safe orchestrationEHR / CRM integration

Ollie Pediatrics · concierge HIPAA-compliant platform

Ollie Pediatrics Pediatric primary care · US

A modern, integrated, AI-powered pediatric experience uniting parents, doctors, and technology under one HIPAA-compliant hub. "Clinician Ask AI" answers provider questions about a single patient in scope. Validates the lean-team, fixed-scope shape we are proposing for Dapper.

Scope

• Product discovery, UX/UI, and full-stack development.
• All-in-one hub for child wellness: birth through adolescence.
• Real-time care access: telehealth, clinic, home visits, callbacks.

Delivery

• 4 months agile engagement with continuous delivery.
• 4-member lean team.
• HIPAA-compliant, mobile-first iOS/Android, modular architecture.

Outcome

• "Clinician Ask AI" in production answering provider questions per-patient.
• "We've aligned on the goals and ways of working that suit Ollie, and our partnership is growing stronger."
• Ramses Rivero, Founder, Ollie Pediatrics.

HIPAA-compliantiOS & AndroidTelehealthClinician AI4-month MVP

Ambient AI intake companion · clinician note generation

Client · confidential US venture-backed elderly-care platform

Real-time transcription with speaker separation, converted into schema-aligned JSON with provenance and confidence scoring. Pre-fills intake forms and generates clinician-ready SOAP and HPI summaries. Validates the prototype shape: lean team, fixed scope, demoable MVP in two months.

Scope

• Ambient capture of nurse-patient conversations.
• Schema-aligned intake form pre-fill with provenance and confidence.
• Reviewer-centric queue + side-panel field validation, one-click autofill.

Delivery

• 2 months agile engagement, fixed scope, MVP delivery.
• 3-member lean team.
• HIPAA-conscious POC with documented hardening plan (BAA, encryption, IAM).

Outcome

• Draft SOAP/HPI summaries linked back to source evidence for rapid verification.
• Clinician-ready notes generated inside the existing intake workflow.
• POC accepted as the basis for the productionization roadmap.

Real-time STTSpeaker separationStructured extractionHIPAA-conscious2-month POC

alldayDr · UK telemedicine & online pharmacy platform

alldayDr UK telemedicine + pharmacy · nationwide

Built and scaled a UK-compliant telemedicine and pharmacy platform from concept to nationwide rollout. Demonstrates the scale path the Dapper prototype is designed to graduate into: real regulatory posture, multi-app surface area, multi-year engagement.

Scope

• End-to-end delivery from ideation to execution.
• 8 mobile and web apps developed across patient, provider, and pharmacy surfaces.
• SNOMED and dm+d standards integrated for clinical and prescription data.

Compliance

• Awarded a place on the NHS GPIT framework.
• CQC and ISO 27001 certified for data security.
• UK-healthcare-aware SDLC across the engagement.

Outcome

• 2+ years agile engagement with continuous delivery.
• 25% reduction in time and cost across release cycles.
• "The team quickly grasped and understood the UK Healthcare system and always came up with new ideas." Suhel Ahmed, Founder & CEO.

NHS GPITCQC certifiedISO 27001SNOMEDdm+d8 appsNationwide rollout

Why SoluteLabs, and what's next.

A senior pair as the build core. Same faces from day one through demo day.

SoluteLabs ships text and voice AI agents in production at enterprise scale. The Tech Lead owns ElevenLabs end-to-end. The Fullstack Engineer ships Next.js + shadcn every week. Not a side project for a generalist team.