Ai-mee — Business Overview

Document Purpose: Internal strategic planning foundation for business plan development. Status: MVP complete (Phase 3). Phase 4 in progress. Ready for beta customer onboarding. Last Updated: April 2026

Table of Contents

1. Executive Summary
2. Problem Statement
3. Solution
4. Key Features & Capabilities
5. Target Market
6. Business Model & Pricing Strategy
7. Competitive Landscape
8. Competitive Advantages
9. Technology Overview
10. Current State & Roadmap
11. Key Metrics
12. Risk Factors & Mitigations

1. Executive Summary

Ai-mee is an AI-powered marketing assistant that automates content creation, brand quality control, and multi-channel publishing — operated entirely through natural language conversation on Telegram, with an optional web dashboard for teams that prefer visual management.

Unlike traditional marketing platforms that require users to log in, navigate dashboards, and fill out forms, Ai-mee meets clients where they already are: in chat. A single message — "Create posts about our summer sale for Instagram, LinkedIn, and email" — triggers a behind-the-scenes four-agent AI pipeline that generates platform-optimised content, enforces brand voice rules, and presents a reviewed, approved draft ready for client sign-off — all within minutes.

The core value proposition is threefold:

• Zero friction — No dashboards, no logins, no forms. Just chat.
• Built-in quality control — Every draft is reviewed against the client's own brand voice rules before human eyes see it.
• Proactive, not reactive — Daily briefings, approval reminders, and weekly performance summaries run automatically. Ai-mee is always working, even when the client isn't.

Ai-mee targets small-to-medium businesses (SMBs) and marketing agencies managing multiple client accounts. The platform operates on a freemium SaaS model with paid tiers designed for scale.

2. Problem Statement

2.1 For SMBs

Small and medium businesses — restaurants, retail shops, service providers, local brands — need consistent, high-quality social media and email marketing to compete. But:

• They can't afford copywriters. Freelancers cost £50–200 per post; a full-time content marketer is out of reach for most SMBs.
• They don't have time for dashboards. Marketing software requires logins, onboarding, tutorials, and ongoing management — all friction that kills adoption.
• AI tools produce generic content. Off-the-shelf ChatGPT outputs don't know the brand's voice, audience, or history. Every piece needs manual editing.
• Nothing is proactive. Every existing tool waits for the user to initiate. There's no system that says "Here's what you should post today" without being asked.

2.2 For Marketing Agencies

Agencies managing 10–50+ client accounts face a different version of the same problem:

• Maintaining distinct brand voices at scale is manual and error-prone. Copywriters must context-switch between wildly different client personas constantly.
• Approval workflows are email chains. Back-and-forth approval threads waste hours per client per week.
• Multi-platform publishing is fragmented. Different tools for Instagram, LinkedIn, Facebook, email, and blog mean no single source of truth.
• Reporting is manual. Aggregating performance metrics across platforms and clients is a spreadsheet nightmare.

2.3 The Root Cause

The marketing software industry has built powerful tools that assume users want to use the software. The real need is different: businesses want their marketing done, not a tool to do it with.

3. Solution

Ai-mee is the first AI marketing assistant designed around the principle that the best interface is no interface — or more precisely, the interface the user already uses daily: Telegram.

3.1 How It Works

For the client, it looks like this:

Client sends: "Create posts about our Easter promotion for Instagram and LinkedIn"

Ai-mee replies (minutes later):
  📸 Instagram — [caption, hashtags, image suggestion]
  💼 LinkedIn — [professional post, discussion frame]

  👍 Approve / ✏️ Edit / 🔄 Regenerate / ❌ Reject

Behind the scenes, a four-agent AI pipeline executes invisibly:

The client never sees a rejected draft. Quality control is internal.

3.2 Proactive Automation

Ai-mee does not wait for requests. Three automated cron jobs run without any client action:

3.3 Optional Web Dashboard

For teams preferring visual management, a full-featured Vue 3 web application provides:

• Client management and campaign overview
• Post editor with live preview
• Content calendar with scheduled post timeline
• Integration settings (social account connections)
• Analytics and performance reporting
• Brand voice rule management
• Media asset library

The dashboard is supplementary — the Telegram interface is the primary product.

4. Key Features & Capabilities

4.1 Telegram-First Conversational Interface

• Natural language requests — No forms or templates. Clients describe what they need in plain English.
• Guided onboarding — The /start command triggers a four-question brand voice setup conversation. No forms, no dashboard visits required.
• Inline approval workflow — Approve, edit, regenerate, or reject posts directly in chat. No context switching.
• Media uploads — Clients can send product photos, menus, or reference images via Telegram. Ai-mee stores them and surfaces them in future content briefs.
• Account linking — Telegram accounts are paired to client accounts via the dashboard settings page. Subsequent messages are automatically scoped to that client.

4.2 Multi-Platform Content Generation

A single request can generate platform-optimised content for all selected channels simultaneously:

Each platform receives content tailored to its audience and format — not the same text copy-pasted.

4.3 Brand Voice Engine

Every client's brand is defined across four rule dimensions, stored in the database and enforced on every generated draft:

The Critic agent scores every draft. Any draft scoring below threshold on any dimension is rejected and sent back to the Creator agent with specific revision instructions. Clients see only approved content.

4.4 Publishing Integrations

Ai-mee publishes directly to platforms. Client credentials are encrypted per-account (AES-256) and stored in isolation:

The platform adapter pattern makes adding new publishing targets straightforward.

4.5 Scheduling & Automation

• Single and batch scheduling — Schedule individual posts or multiple posts in one operation
• ISO 8601 timestamps — Precise scheduling with timezone support
• Unschedule capability — Revert a scheduled post back to approved state without losing content
• Automated daily generation — A background job creates up to 3 posts per campaign each day at 9 AM, keeping content pipelines full without manual input

4.6 Analytics & Performance Tracking

Metrics tracked per platform:

• Social: likes, shares, comments, impressions, reach, click-through rate, engagement rate
• Email: open rate, click rate, bounce rate, unsubscribe events, delivered count

Performance data feeds into weekly summary briefings and, in Phase 4, into the generation brief to prioritise content types and topics that have performed historically well for that client.

4.7 Web Dashboard

The Vue 3 SPA (deployed to Cloudflare Pages, globally distributed) covers the full management surface:

• Client and campaign management
• Post editing with live preview
• Analytics charts and trend views
• Integration credential management (OAuth flows)
• Brand voice rule editor
• Scheduled posts calendar view
• Media asset library (images, documents, scraped pages, menus, video)
• Admin panel (manual briefing triggers, usage monitoring)

Full dark mode support; responsive across desktop, tablet, and mobile.

4.8 Email Campaign Generation

• AI-generated HTML templates with semantic content structure
• Tagged content injection blocks for future reuse
• Resend integration for sending with open/click tracking
• Full delivery and event logging

5. Target Market

5.1 Primary: Small-to-Medium Businesses (SMBs)

Profile: Local restaurants, retail shops, service providers, health & wellness brands, hospitality businesses. Typically 5–100 employees, 1–3 people managing marketing alongside other responsibilities.

Core pain: Consistent social presence and email marketing is essential but consumes time they don't have and skills they don't possess in-house. They cannot justify a full-time content marketer.

Why Ai-mee wins here: The Telegram interface requires no software adoption. Clients interact naturally, the way they already communicate. There is no onboarding curve — if they can send a text, they can use Ai-mee.

Estimated market size signal: There are approximately 33 million SMBs in the US alone, with the global SMB market estimated at 330+ million businesses. Social media marketing services represent a multi-billion dollar annual spend in this segment.

5.2 Primary: Marketing Agencies

Profile: Boutique and mid-size agencies managing 10–50+ client accounts. Teams of 5–30 people. Often constrained by copywriter capacity and client approval cycle friction.

Core pain: Maintaining distinct, enforced brand voices across many clients simultaneously is the hardest part of the job. Approval workflows are slow and manual. Publishing across multiple platforms per client compounds the workload.

Why Ai-mee wins here: The per-client brand voice engine solves the hardest part of agency content work. The Critic agent effectively acts as a brand standards enforcer for every account, reducing human QA time. Multi-client management in the dashboard gives account managers a single overview. The approval workflow moves entirely into chat, reducing email chains.

Agency tier pricing (per-seat model) creates strong revenue leverage — one agency seat covers 10+ clients.

5.3 Secondary Markets

5.4 Customer Personas

6. Business Model & Pricing Strategy

6.1 Revenue Model

Ai-mee operates a freemium SaaS model with three tiers. Tiers are designed to align cost with the value delivered, with a natural upgrade path as clients scale.

6.2 Unit Economics

The cost to deliver Ai-mee's core service is low and highly scalable:

Key insight: Variable cost per post is in the £0.01–0.05 range using Claude Haiku, which is deliberately chosen as the most cost-effective capable model tier. This creates strong gross margins at scale, characteristic of a high-leverage SaaS product.

6.3 Monetisation Infrastructure

• Billing: Subscription billing via a standard payment processor (e.g., Stripe) — not yet implemented, planned for Phase 5
• Usage tracking: Posts generated, platforms published to, and API calls are logged and queryable from the admin panel
• Metering for freemium: Post count, platform count, and feature access controlled at the application layer

6.4 Pricing Considerations

• Agency multiplier: Agencies derive 10–50x the value of individual clients by managing multiple accounts. Per-seat agency pricing captures this.
• Add-on opportunities: Premium image search credits, dedicated briefing frequency, custom LLM fine-tuning (longer-term), white-label Telegram bot instance.
• Retention via brand investment: Clients who build brand voice rules and content history have strong switching costs — migrating this context to a competitor is non-trivial.

7. Competitive Landscape

7.1 Direct Competitors

7.2 Adjacent Competitors

7.3 Competitive Positioning Map

                        HIGH AI AUTOMATION
                               ↑
                          [Ai-mee]
                    Brand-enforced, multi-channel,
                     conversational, proactive
                               |
   CHAT/SIMPLE ←──────────────┼──────────────→ DASHBOARD/COMPLEX
                               |
                [Jasper/Copy.ai]    [Buffer/Hootsuite]
                  AI generation,     Scheduling,
                  no publishing,     no generation,
                  form-based         dashboard-heavy
                               |
                               ↓
                        LOW AI AUTOMATION

8. Competitive Advantages

8.1 Durable Advantages (Moat)

1. Telegram-First UX Zero-friction engagement. Clients interact in the messaging app they use daily. There is no onboarding, no login, no dashboard to learn. Competitors cannot replicate this without rebuilding their entire product architecture around a conversational interface.

2. Built-In Quality Control (The Critic Agent) Every generated draft passes through an automated brand review before the client sees it. This is structurally different from competitors that generate content and hand it to humans to check. Ai-mee's internal QA loop reduces client revision requests and increases first-pass approval rates — which is both a product quality metric and a retention driver.

3. Proactive Intelligence Daily briefings, approval reminders, and weekly performance summaries run without any client action. The product creates value on days the client doesn't use it. This drives retention and perceived value in a way that passive SaaS tools cannot.

4. True Multi-Platform Generation One request generates platform-native content for all selected channels simultaneously. Content is not copy-pasted — each platform receives format-appropriate, length-appropriate, tone-adapted output. Competitors either generate generic content or require per-platform inputs.

5. Brand Intelligence Compound Effect Brand voice rules accumulate in the database. Performance data from published posts feeds back into future briefs. The more content a client generates through Ai-mee, the better tuned the output becomes. This creates a data flywheel effect that deepens switching costs over time.

6. Extensible Architecture The publishing adapter pattern means adding a new channel (TikTok, Pinterest, Bluesky) requires implementing a single interface — not a platform rebuild. As new social platforms emerge, Ai-mee can add support faster than less architecturally deliberate competitors.

7. Cost Efficiency Claude Haiku 4.5 via OpenRouter is the most cost-effective capable model tier available. The choice is deliberate and creates structural cost advantages over competitors using GPT-4 class models for equivalent tasks. Per-post variable cost in the £0.01–0.05 range enables high gross margins at any reasonable price point.

8.2 Switching Costs

Clients who actively use Ai-mee accumulate:

• Brand voice rule libraries tuned over time
• Content history informing future generation briefs
• Agent memory (pgvector embeddings) calibrated to their specific brand
• Approval workflow patterns embedded in Telegram conversation history

Migrating this context to a competitor is non-trivial. Switching costs increase with tenure.

9. Technology Overview

This section provides a non-technical summary of the platform architecture, intended for strategic context rather than technical evaluation.

9.1 Architecture Overview

Ai-mee is composed of three primary services:

┌──────────────────────────────────────────────────────────────┐
│ Clients (Telegram)          Clients (Web Dashboard)          │
└────────────────┬────────────────────────────┬───────────────┘
                 │                            │
       ┌─────────▼──────────┐      ┌──────────▼────────────┐
       │  AI Agent Layer    │      │  Web Application       │
       │  (GoClaw, Docker)  │      │  (Vue 3, Cloudflare)   │
       │  4 Agents          │      │  Client mgmt           │
       │  Telegram webhook  │      │  Post editor           │
       │  Cron scheduler    │      │  Analytics             │
       │  Agent memory      │      │  Integrations          │
       └─────────┬──────────┘      └──────────┬────────────┘
                 │                            │
       ┌─────────▼────────────────────────────▼────────────┐
       │              API Backend (Node.js, Fastify)        │
       │  Content generation   Publishing adapters          │
       │  70+ REST endpoints   MCP tool server (28 tools)   │
       │  OAuth flows          Email sending & tracking      │
       └─────────────────────────┬──────────────────────────┘
                                 │
                    ┌────────────▼──────────────┐
                    │  Supabase (PostgreSQL)     │
                    │  15+ tables               │
                    │  pgvector (agent memory)  │
                    │  JWT authentication        │
                    │  Row-level security        │
                    └───────────────────────────┘

9.2 Technology Choices & Rationale

9.3 Security Posture

• Credential encryption: OAuth tokens and API keys stored encrypted at rest (AES-256) in Supabase
• Per-client data isolation: Row-level security policies ensure no cross-client data access
• Authentication: JWT-based for web sessions; separate API key auth for service-to-service calls
• Service credentials: Supabase service role key (bypasses RLS) used only for bot-layer server operations, never exposed to client requests
• Webhook verification: Inbound webhook signatures (e.g., Resend) verified before processing

9.4 Deployment

• Development: Docker Compose stack covering all services (GoClaw + API + Supabase local + frontend)
• Production: API on Docker or cloud VM (AWS ECS, Fly.io, Railway); frontend on Cloudflare Pages; Supabase cloud; GoClaw as a Docker container
• Scale path: Current single-instance for each service; horizontal scaling via Kubernetes for GoClaw and API when volume demands

10. Current State & Roadmap

10.1 Completed (Phases 0–3) — Production-Ready MVP

AI Agent Pipeline

• ✅ Liaison Agent — full client conversation management, onboarding, delegation, approval workflows
• ✅ Creator Agent — brief building, multi-platform generation, image suggestions
• ✅ Critic Agent — four-dimension brand scoring, revision loops (max 2), verdict logging
• ✅ Briefing Agent — daily status snapshots, approval reminders, weekly summaries via cron

Content Generation & Publishing

• ✅ 6 content channels: Instagram, Twitter/X, LinkedIn, Facebook, Blog, Email
• ✅ 8+ publishing adapter integrations: Twitter, LinkedIn, Facebook, Mailchimp, Ghost, Blogger, WordPress, Outstand
• ✅ Simultaneous multi-platform generation from a single request
• ✅ HTML email template generation with content injection tagging
• ✅ Resend integration with open/click/bounce tracking
• ✅ Scheduled publishing with batch scheduling support
• ✅ Post-publish Telegram notifications

Platform Infrastructure

• ✅ 70+ REST API endpoints (generation, client management, publishing, OAuth, tracking, admin)
• ✅ 28 MCP tools exposed to agent layer (covering all client operations)
• ✅ Supabase database schema (15+ tables, full workflow state machine)
• ✅ JWT + API key authentication; service role for bot operations
• ✅ AES-256 credential encryption for all OAuth/API key credentials
• ✅ Row-level security across all client data tables

Web Dashboard

• ✅ Client and campaign management
• ✅ Post editor with live content editing
• ✅ Analytics views (per-platform metrics)
• ✅ Telegram pairing via Settings page
• ✅ Integration credential management (OAuth flows)
• ✅ Admin panel (manual briefing triggers)
• ✅ Dark mode; Cloudflare Pages deployment

Automation

• ✅ Three production cron jobs (daily briefing, approval reminders, weekly summary)
• ✅ Automated daily post creation pipeline (max 3 posts per campaign)

10.2 In Progress (Phase 4) — Intelligence Layer

10.3 Planned (Phases 5–6) — Scale & Expansion

11. Key Metrics

Metrics are tracked to serve two purposes: product health monitoring and business performance evaluation.

11.1 Product Health Metrics

11.2 Business Performance Metrics

11.3 Operational Targets (Illustrative)

The following are internal planning benchmarks, not externally committed targets:

12. Risk Factors & Mitigations

12.1 Operational Risks

12.2 Product Risks

12.3 Scaling Risks

12.4 Market Risks

This document is a living reference. Update it as the product evolves, pricing is validated through customer conversations, and market data becomes available.