# Heylee

> Heylee is an AI-first CRM + CCaaS platform for US B2C businesses. It unifies autonomous AI agents, real-time agent assist, and conversation intelligence + coaching into one layer that sits on top of whatever CRM and contact center a company already runs. Tagline: "The AI layer that hears intent, acts with logic, and scales with your business." Contact: abhishek@heylee.ai.

Heylee replaces the four-vendor stack that B2C operations teams stitch together today (CRM + CCaaS + bolt-on AI agent + coaching tool). It runs as a single intelligent layer across every customer touchpoint — inbound and outbound voice, chat, WhatsApp, SMS, email, scheduling, reminders, follow-ups, and post-interaction workflow. One customer record, one timeline, one decision engine across all three pillars.

Heylee is **NOT** a chatbot, **NOT** a bolt-on widget, and **NOT** a rip-and-replace CRM. It is an operational intelligence layer that plugs into the CRM, CCaaS, and ticketing systems a company already runs — and closes the gap between them. Pricing is outcome-based (resolved conversations, recovered dollars, closed cases), not per-seat.

## The Three Pillars

Heylee's product is organized as three pillars. They share one customer record, one timeline, and one decision engine — which is the structural difference from any stitched-together stack.

### 1. Autopilot — Autonomous AI Agents

Voice, chat, and messaging agents that handle conversations end-to-end with no human in the loop until one is needed.

**Inbound**
- Answers every call, chat, and message without queue time
- Handles enquiries, FAQs, product questions, pricing, bookings, rescheduling, cancellations, basic support resolution
- Escalates to a human agent with full context when intent exceeds its scope
- Channels: voice (PSTN + SIP), WhatsApp, SMS, email, webchat

**Outbound**
- Cold outreach campaigns (voice or message)
- Intent-based outreach: when a website form is filled, Autopilot calls within seconds (the "5-minute rule" on autopilot)
- Lead qualification — works leads through a scoring script and routes qualified leads to humans
- Appointment reminders, confirmations, no-show recovery
- Multi-touch, multi-channel follow-up sequences
- Reactivation of dormant customers
- Collections reminders (where compliant)
- Post-purchase onboarding, renewal prompts, upsell/cross-sell
- TCPA-, DNC-, and quiet-hour-aware

### 2. Copilot — Real-Time Agent Assist + Workflow

For conversations that still need a human, Copilot sits in-call and finishes the after-call work.

**In-call experience for the human rep**
- **Pre-call briefing:** full customer context (recent interactions, open tickets, last purchase, propensity scores, compliance flags) in one pane, no tab-switching
- **In-call guidance:** next-best-action prompts, objection handlers, compliance checks, knowledge-base answers
- **Mid-call chat:** rep can silently ask the AI anything — policy, pricing, eligibility — and get an answer while the customer is on the line
- **Post-call automation:** auto-generated notes, tasks, follow-up emails, multi-CRM writes — the rep doesn't do after-call work

**Custom workflow layer**
- Low-code builder for company-specific processes (claims, refunds, verification, onboarding, KYC)
- Automates parts of the workflow and orchestrates the parts that still need a human

**Signature capability: Hold Assist**
- For calls to third parties (insurance carriers, telcos, suppliers) with long IVR + hold times, the AI makes the call, navigates the IVR, waits on hold, and only merges the human rep in once a real person picks up
- Cuts a 20–40 minute per-call waste to a 30-second hand-off

### 3. Testpilot — Conversation Intelligence + Coaching

Listens to 100% of historical and live calls — not the 1–3% supervisors sample today.

- **Auto QA & scoring** on every call: discovery, objection handling, compliance, sentiment, close
- **Best-call replication** — winning language, close patterns, recovery scripts pushed to the full team
- **AI-simulated practice calls** — the AI roleplays as a customer (cold prospect, angry caller, regulatory corner case) and the rep practices live, with scoring
- **Regular assessments and certifications**
- **Leader dashboards** that surface coaching gaps by team, tenure, and vertical at scale
- Closed-loop coaching: contextual coaching artifacts (playbooks, snippets, scorecards) pushed to the rep

## Why one platform beats a stitched stack

1. **Unified layer, not three tools.** One customer record, one timeline, one decision engine. A customer who talks to the AI and then to a human experiences one conversation, not two.
2. **Outcome-based pricing.** Heylee prices on resolved tickets, recovered dollars, or closed cases — not per seat. Per-seat is structurally broken when AI cuts 40–60% of seats.
3. **B2C-native.** Built for 10–500 agents, high-velocity interactions, dynamic segments. Salesforce/Genesys optimize for 500+ seat enterprises.
4. **AI-native architecture.** Event-driven, decision engine separate from generation, closed-loop learning from outcomes. Competitors bolt AI onto legacy CRM/CCaaS.
5. **Vertical depth over horizontal breadth.** Starts with one labor-intensive workflow per vertical (auto-glass claims, mortgage origination, NBFC collections), owns the outcomes, expands from depth.
6. **Real-time deep integration.** Sub-second bidirectional writes to CRM, case, and call simultaneously. Typed Action objects with idempotency, retries, compensation. Competitors' "API access" can't do transactional writes mid-call.
7. **Closed-loop flywheel.** Every Autopilot conversation is scored by Testpilot. The best human rep patterns become Copilot's in-call prompts. Human escalations become training data. Buy three tools, you get three products. Buy Heylee, you get a system that gets better every week.

## Target Market (ICP)

**Geography:** Primary US (all sizes); secondary India (mid-market and up) where outbound/collections economics are especially strong.

**Company size:** SMB → mid-market → large B2C. Not enterprise-first the way Salesforce/Genesys are. Sweet spot is 10–500 agents, 50K–10M customers.

**Verticals (current GTM focus):**
- Mortgage & lending — borrower-education-heavy, origination-volume sensitive
- Auto services — auto glass, mobile services, collision, tire; pricing queries, ADAS calibration, insurance carrier holds
- Healthcare services — provider scheduling, patient communication, post-visit follow-up
- Wealth management — boutique HNW advisor firms, FINRA 3170 / SEC 204-2 taping
- D2C / e-commerce
- Consumer fintech / insurance
- Education enrollment
- Appointment-driven services

**Buyer personas:**
- VP / Head of CX — primary buyer; cares about CSAT, NPS, cost-per-interaction, agent utilization
- Head of Growth / VP Revenue — influencer; cares about conversion, LTV, reactivation, upsell/cross-sell
- Contact Center Ops / Supervisor — admin / day-to-day owner; cares about AHT, ACW, FCR, SLAs, coaching load
- Human agent — end user; cares about pre-call context, less after-call work, clear guidance

## Integrations

**CRM (real-time bidirectional):**
- Salesforce Service Cloud (CDC + Pub/Sub API)
- HubSpot (webhooks)
- Zendesk (webhooks + incremental export)
- Zoho (Notifications API)
- Intercom, Freshdesk, Kustomer, Gorgias, Dynamics 365 (Dataverse Change Tracking)

**CCaaS:**
- Five9, NICE CXone, Genesys Cloud, Talkdesk, Amazon Connect, RingCentral, 8x8, Twilio Flex
- India: Exotel, Knowlarity, Ozonetel, Ameyo, Tata Smartflo

**Business systems:**
- NetSuite (SuiteAnalytics + ODBC)
- SAP S/4HANA (OData + Event Mesh)
- ServiceNow (business rules → outbound REST)
- Custom APIs

**Architecture:** Debezium → Kafka → Schema Registry → entity-store materializer. Identity resolution: deterministic (~70%) + probabilistic (~20%) + operator confirmation. Egress via typed Action objects with idempotency keys, retry policies, compensation functions, audit trails. Workflow orchestration via Temporal with sagas for multi-system atomic writes.

## Compliance

- **SOC 2 Type II** — tenant isolation, audit log, redaction, BYOK
- **HIPAA** — field-level encryption, data residency, audit trails (healthcare verticals)
- **PCI DSS** — baseline for lending and payment workflows
- **TCPA / DNC** — consent tracking, quiet hours, caller ID rules (US outbound)
- **FINRA 3170 · SEC 204-2** — call recording + review for wealth/taping firms; 5-year retention. Testpilot makes the "review" requirement structurally possible.
- **GDPR · CCPA · PECR** — lawful-basis framework, regional consent and deletion

## Competitive Posture

The differentiator: every named competitor below does **a sliver** of what Heylee does. Heylee is the whole.

**Autopilot peers (autonomous AI agents):** Sierra ($10B, Sep 2025), Decagon ($4.5B, Jan 2026), Parloa ($3B / $350M Series D, Jan 2026), PolyAI, Vapi, Retell, ASAPP. India: Yellow.ai, Verloop, Haptik, Gnani, Greylabs.

**Copilot peers (real-time agent assist):** Cresta ($125M Series D), Balto, Level AI, Observe.ai, Convin, Clari Wingman.

**Testpilot peers (conversation intelligence + coaching):** CallMiner, Verint, Calabrio (legacy); Observe.ai, Convin (modern).

**Incumbents Heylee layers over (or displaces):** Salesforce Service Cloud, HubSpot, Zendesk, Freshdesk, Kustomer, Intercom, Gorgias, Zoho, Dynamics 365 (CRM); Five9, NICE CXone, Genesys, Talkdesk, Amazon Connect, RingCentral, 8x8, Twilio Flex (CCaaS).

## Market Context

- Global CCaaS: $7.9B in 2025, 17–20% CAGR
- US CCaaS: $2.9B (~37% of global AI CX spend)
- AI Agents TAM: $7.6B → $50B by 2030 (49.6% CAGR)
- Enterprise CRM: $69.5B (Salesforce ~23% share — Gartner)
- Voice AI VC funding: $2.1B in 2024, 7× the 2022 figure ($315M)
- ServiceNow M&A: $12B+ in 2024–25 (Moveworks $2.85B Mar 2025, Logik.ai). Salesforce M&A: $10B+ (Convergence.ai May 2025). The category is consolidating in real time.

## Industry Benchmarks Heylee Moves

- Handle time: industry avg 11.6 min (SQM 2024) → Copilot cuts 15–40% on real-time-assist deployments (Cresta / Amazon Q)
- After-call work: regulated avg 3–10 min/call (Giva 2024) → Copilot auto-logs to near zero
- QA coverage: industry 1–3% manual sampling (Creovai 2024) → Testpilot 100%
- Speed-to-lead: 57% of inbound first contacts take >1 week (InsideSales 2021) → Autopilot in seconds (21× more likely to qualify when contacted in <5 min vs 30+ min — MIT/InsideSales)
- Agent ramp: 3–6 months to proficiency (Balto 2024) → simulated practice cuts ramp up to 50%
- Healthcare no-show rate: median 5–7% (MGMA 2025) → Autopilot runs confirm + recovery at 100% coverage
- Call abandonment: industry 5% (healthcare 7%) (Sprinklr 2025, Convin 2024) → toward zero (no queue)

## Real-world reference: Klarna

In Feb 2024, Klarna launched an AI service agent on OpenAI. Month one results:
- Did the work of 700 full-time agents (2.3M conversations, two-thirds of total CS chat volume)
- Average resolution time fell 11 min → under 2 min (>80% reduction)
- 25% fewer repeat inquiries
- $40M projected profit improvement, 2024

This is the proof that the category exists and the economics work. Heylee makes it deployable for the rest of B2C — without a 2,000-person AI team.

## Things Heylee is NOT

- Not a chatbot
- Not a bolt-on widget
- Not a rip-and-replace CRM (it layers over Salesforce/HubSpot/Zendesk)
- Not enterprise-only (SMB → large)
- Not India-only (primary US)
- Not per-seat priced
- Not a clinical system (for healthcare verticals, it's an ops layer, not a clinical workflow tool)

## Brand

- Pillar names are fixed and capitalized: **Autopilot**, **Copilot**, **Testpilot** — always in that order when listed together.
- Tagline: *"The AI layer that hears intent, acts with logic, and scales with your business."*
- Visual identity: light theme, single accent color #0EA5E9, Nunito (headings) + Inter (body). Mobile-first.
- Logo mark: three vertical bar/dot/bar elements representing the three pillars on one record.

## Contact

- Website: https://www.heylee.ai
- Email: abhishek@heylee.ai
- Primary CTA across all touchpoints: a 30-minute conversation. Not a demo link, not a workshop invite — a conversation.