Saibbyweb · Proposal v1.0 Confidential

An AI Companion for Energy Doc

Your presence — extended into the gaps between sessions. Trained on your methodology, speaking with your voice, available to every patient at 2 a.m. and on a quiet Sunday.

Prepared forDr. Hammad Sharif — Energy Doc
Prepared bySaibbyweb — hello@saibbyweb.com
DateMay 25, 2026
Versionv1.0 (Confidential)
Engagement range$15,000 – $84,000

1. Introduction

Dr. Sharif,

This document proposes an AI companion built around your clinical methodology — a product that extends your presence into the gaps between sessions. It is trained on the way you work, speaks in your voice, and is available to every one of your patients at 2 a.m. or on a quiet Sunday afternoon.

The product is engineered for three commitments: it is built around your methodology, available 24/7, and engineered for the level of trust mental healthcare demands.

This is a phased proposal — a path, not a leap. We walk through what we'd build, how, and what it costs at each stage. You can stop at any milestone and own everything built up to that point.

2. Engagement structure: two milestones, one honest path

The work is organized around two milestones with a clear decision point between them.

Milestone 1 — Day 40: MVP Checkpoint. A complete, polished iOS app, ready to validate the concept in the real world. What ships at this point: a premium navy/gold iOS app, methodology-trained chat, voice clone, AI-powered journaling with a crisis-detection layer, and in-app subscription. At the end of Day 40 you have a real decision to make — continue, pivot, or stop.

Milestone 2 — Day 120: Full Launch. iOS plus Android, with the full companion experience live. Everything from the MVP carries forward and is deepened. New at this stage: the Android app, a lip-synced video avatar, couples mode (two speakers in one session), pattern tracking with personalized check-ins, and an admin dashboard. This is the public launch, followed by three months of hyper-care.

A note on what "MVP" means here

The MVP at Day 40 is not a 33% slice of the final product. It is a complete, polished iOS app — think of it as a fully built two-bedroom home. The work between Day 40 and Day 120 does not just add new rooms; some of it reinforces the foundation so it can carry the second floor. A few MVP components are deliberately revisited as we expand. That is intentional, not rework, and it is why a phased approach is more honest than promising everything in one shot.

3. The three commercial paths

There are three ways to engage. Each one is self-contained, and Paths B and C can be combined with an optional Patient Web App add-on (+$20,000).

Path A — Build the MVP

$15,000
40 days · 2 payments · iOS only

The most cautious starting point. You get the premium UI, methodology-trained AI, the voice clone, and the AI journaling layer with crisis detection. This is the right path if the goal is to validate the concept before deciding on a larger commitment. (The Patient Web App add-on is not available on Path A.)

Path C — Public MVP + Full Rollout

$63,999
120 days · 5 payments

Everything in Path B, plus a public App Store launch at Day 40. That means we also run live support and bug response, payment-lifecycle operations, and continue the deeper build through Day 120 while real patients are using the app. Higher cost, faster public exposure.

Optional add-on: Patient Web App (+$20,000)

A web app with full feature parity, available with Path B or Path C. The main commercial reason to take it is that direct web billing bypasses the App Store / Play Store platform fee (typically 15–30%), which materially improves margins on every web subscriber. Pricing with the add-on:

Payment schedule (Path B, illustrative)

MilestoneWithout webWith web
Day 1$7,500$11,500
Day 20$7,500$11,500
Day 40$13,000$17,000
Day 67$13,000$17,000
Day 93$12,999$16,999

4. What ships in MVP vs. Full Rollout

Everything in the MVP carries forward to Full Rollout. Full Rollout adds depth — more devices, more interactions, more intelligence.

MVP at Day 40 (8 capabilities, iOS only) delivers everything needed to validate the concept: the premium navy/gold iOS interface, methodology-trained chat, voice clone, AI-powered journaling, crisis detection with 988 routing, the onboarding flow, in-app subscription, and push notifications.

Full Rollout at Day 120 (16 capabilities, iOS + Android + Admin) adds: the Android app, couples mode (two speakers), pattern tracking across sessions, advanced pattern visualizations, personalized check-ins, the lip-synced video avatar, guided multi-day programs, and an admin dashboard.

The optional patient web app — cross-device, no platform fee — is available on Paths B and C.

5. The engineering work, by module

The build is organized into eight engineering modules, each with its own risk profile.

01 · Low risk
The Therapist's Brain — Methodology Encoding

The problem is translating your methodology into a structured system the AI can reason with consistently across thousands of sessions. The reason it's hard: without explicit scaffolding, the AI drifts into generic CBT-flavored responses instead of holding your specific frameworks and refusal patterns. Our approach: structured system prompts derived from your transcripts, few-shot exemplars for canonical interventions, refusal scaffolds for "things you never do," and validation against a held-out evaluation set.

02 · Medium risk
Voice Cloning — Real-Time Synthesis

The problem is producing a voice that genuinely sounds like you — warm, present, in character — at conversational latency, without burning the unit economics. The hard part is sample curation and balancing time-to-first-byte against quality and cost-per-character in streaming synthesis. Our approach: curated voice samples for a professional clone, streaming TTS with sentence-level chunking, SSML cues for pacing and breath, and cached responses for common openings.

03 · High risk
Multi-Session Memory — Pattern Tracking

Patients return weekly for months. The companion must remember, surface patterns ("you mention sleep every Sunday"), and never retraumatize a patient by forgetting. Naïve long-context approaches break at scale, and pattern detection across sessions needs structured extraction, not just similarity search. Our approach is hybrid: structured fact extraction (themes, recurring stressors) combined with semantic recall, patterns surfaced via offline aggregation, and a strict guarantee that memory never crosses patient accounts.

04 · Medium risk
Couples Mode — Relational Session Engine

Two voices, one shared session, separate emotional tracks — the AI navigates two perspectives without taking sides and routes each turn to the right speaker. A single embarrassing failure here breaks trust, so we choose reliability over cleverness at launch. Our approach: manual speaker tagging via the UI ("Partner A speaking"), per-partner private context plus a shared session context, and explicit refusal patterns when one partner tries to triangulate the AI against the other.

05 · Medium risk
Video Avatar — Lip-Synced, Streaming

A lip-synced video version of you that streams in real time and degrades gracefully when the network can't carry it. Cloud-rendered avatars are network-sensitive, cold starts are slow without warm-pool tactics, and stream interruption mid-session is jarring. Our approach: clean replica training footage, warm-pool pre-allocation at common entry points, an audio-only fallback that swaps in transparently when bandwidth drops, and per-tier minute caps to protect margins.

06 · High risk
Crisis Detection — Safety Layer

Catch high-risk moments — ideation, active crisis, abuse disclosure — and route to 988 or other human resources, without false positives that erode the relationship. False negatives are catastrophic; false positives are corrosive. Pure keyword matching misses real cases; pure LLM classification drifts. Our approach is a two-layer system: keyword detection plus a dedicated classifier independent of the conversational model, a conservative threshold with a reviewable audit log, and an explicit "I'm here, are you safe?" hold-and-route flow.

07 · Low risk
Subscriptions & Tiers — Billing, Trials, Tier Logic

Mobile plus optional web billing, trial windows, the $49.99 / $79.99 tier split, restore-purchase, family sharing — all of it wired correctly on the first attempt. Subscription bugs turn directly into refund tickets and App Store rejections, and tier feature-gating must be deterministic and work offline. Our approach: server-side receipt validation, web payment webhooks, locally cached tier state with TTL, and a tested matrix covering new, restore, family, downgrade, lapsed, and refund scenarios.

08 · Low risk
Compliance & Privacy — Data Architecture

HIPAA-aligned data handling, encryption at rest and in transit, vendor BAA chains, patient data residency, and an auditable access trail. A mental health AI without HIPAA-grade architecture is a liability waiting to be invoiced — and the BAAs must be in your name with clear data-flow boundaries. Our approach: AES-256 at rest, TLS in transit, KMS-managed keys, HIPAA-eligible cloud services only, a vendor BAA chain in your name, and per-patient data export and right-to-delete from day one.

Cross-cutting concerns

Premium UI execution, push notification infrastructure, the onboarding flow, the admin dashboard, Android parity, and QA plus App Store submission — these run through all eight modules.

6. Revenue at scale

Once the product is live, the revenue picture is straightforward. The example below uses the higher-volume case (1,000 active patients on the base tier):

Gross monthly revenue (1,000 × $49.99)$49,990
Platform fee (App Store, ~18%)−$8,998
Net monthly (after fees)$40,992
Annual run-rate$491,902

A few things worth noting on the economics:

7. Beyond Day 40: what MVP proves, and how we partner after launch

What Day 40 buys you. The MVP proves technical feasibility — we can ship a polished iOS companion at premium quality, and the engineering path forward is clear. Equally important, Day 40 lets you spend $15K to learn something concrete, rather than $54K to discover it at month four.

After launch, three partnership tiers are available.

Essential Care

$3,200 / month
Maintenance floor

Monitoring and uptime, bug fixes, OS and SDK updates, security patches, infrastructure management, and 5 hours per week of minor tweaks.

Growth Engineering

$15,000 / month
Dedicated mini-team

40 hours per week, weekly standups, quarterly roadmaps, a direct channel, and first-class roadmap input.

8. What we'll need from you

Eight inputs are required to do this well. None are optional, and all are normal — we'll send a structured intake document on Day 1.

  1. Voice recordings. A representative set of clean, present audio — for voice clone training.
  2. Video footage. Clean clips of you on camera — well-lit, neutral background, varied expressions — for video avatar training.
  3. Session transcripts. A handful, anonymized and consented — for methodology system-prompt design.
  4. Methodology documentation. Frameworks, common interventions, refusal patterns.
  5. Brand assets. Logo, colors, fonts — or we build them in collaboration.
  6. Healthcare counsel. For HIPAA and cross-state licensure (separate engagement, recommended early).
  7. Responsive review cycles. A few business days maximum on design and content approvals.
  8. A path decision on Day 1. MVP only, Internal MVP, or Public MVP — with or without the Web add-on.

9. Engagement terms

The commercial mechanics, kept short.

10. Practical notes before we begin

A few items worth flagging upfront, so nothing comes as a surprise.

The video avatar streams from the cloud, so network quality shapes the experience. We build a graceful audio-only fallback that swaps in transparently when bandwidth drops — a bad connection degrades cleanly rather than fails.

Web subscriptions materially improve margins. Direct web billing bypasses platform fees (15–30%). During the build we'll help tune video minutes per tier so the relational tier ($79.99) stays comfortably profitable.

Quoted prices assume the scope agreed at Day 1. Major scope changes mid-project require change orders. Review cycles that slip past 5 business days extend the timeline. Both are normal — we just flag them honestly upfront.

11. Next step

Whichever path you choose, the next step is the same: a 30-minute call to walk through any of these decisions live, and to answer anything this document hasn't already covered.

You can reply to this proposal to proceed, or schedule a call directly. The fastest way to reach me is hello@saibbyweb.com.

Thank you for considering this work.

— Saibbyweb

Proposal v1.0 · Confidential · Updated May 25, 2026