Skip to content
MillionPlus Labs

Productising a Multi-Agent AI Operation

The AI operation we built to run MillionPlus, packaged as a deployable product for high-value, founder-led businesses. It answers a real phone line in a natural voice, qualifies leads across eight channels, writes and fact-checks its own content, and syncs everything to the CRM. Live in 4 to 8 weeks.

8+
Channels
19k
Contacts synced
30k+
Live listings
24/7
Availability
The Challenge

Everyone Saw the MillionPlus System. Then They Asked for One.

We rebuilt MillionPlus as an AI-native, multi-agent operation: one brain running eight-plus channels for a luxury finance business. The result created its own demand, and a harder question:

01

Founder-led luxury businesses run on relationships. Enquiries arrive at 2am on WhatsApp and go cold by morning.

02

Off-the-shelf chatbots cannot hold a luxury brand’s tone or qualify a seven-figure enquiry.

03

Agencies sell AI transformation through months of workshops and procurement theatre.

04

A bespoke multi-agent build is a serious engineering project. Most businesses cannot hire for it.

The Solution

The Technology Behind MillionPlus, Productised

MillionPlus Labs turns the proven MillionPlus architecture into a deployable product. Start with the strongest proof: the business has a real London landline, and the AI answers it. It holds a natural conversation, qualifies the enquiry, and drops the full call into the dashboard with a transcript. Behind that phone line sits one brain running every channel, phone, WhatsApp, web chat, email, Instagram, YouTube and voice on the website, with the same knowledge, the same tone, the same compliance rules, and one conversation history per contact. Each deployment is customised to the business and lands in 4 to 8 weeks.

Solution
01

The phone, answered by AI. A caller dials a real London landline and the AI picks up in a natural voice. It holds the conversation, qualifies the enquiry, and lands the whole call in the dashboard with a full transcript. If the caller is high value or asks for a person, the system escalates while they are still on the line: the team gets an email briefing and a WhatsApp alert in real time. Under the hood it is a carrier-grade SIP trunk into an ElevenLabs conversational agent, deduplicated and run through the same qualification and CRM pipeline as every other channel. It replaced a human call-answering service.

02

Email as a full channel. Several monitored inboxes. The AI classifies every inbound email, replies to genuine new enquiries in the client’s own voice with the client’s real signature, sent from the client’s own domain, and creates the lead in the CRM. Thread replies and internal mail are correctly left alone.

03

A content engine that checks its own facts. It writes articles overnight on a schedule, and ingests the client’s long-form videos, multi-gigabyte files, turning each one into publish-ready articles and email campaigns. One click repurposes any article into YouTube long-form scripts, Shorts, Instagram, LinkedIn and X posts, and it selects and optimises the images. Every article runs through a search-grounded fact-check that verifies claims against live web sources and corrects wrong or outdated facts in place, recording the before, the after and the source. Anything it cannot verify is flagged for a human, never invented. This is FCA-regulated finance, so compliance is enforced in the pipeline itself: mandated footer wording, banned claims stripped, live market rates pulled from search grounding rather than model memory.

04

Instagram and YouTube as active systems. Instagram handles DMs and auto-replies on campaign posts, with AI intent classification that sorts commenters into interest lanes and routes the commercially interesting ones to the team. Four separate YouTube brand channels are connected and synced, comments harvested into the system. On top of both sits AI-curated outreach: when the client publishes a new video, the system selects the warmest past commenters across platforms and drafts personalised public replies for human review before anything is sent. One recent campaign posted 58 personalised replies across Instagram and YouTube.

Technical Details

Built with

Next.jsReactTypeScriptTailwindSupabaseGeminiElevenLabsTwilioMeta Cloud APIAttioResendStripeVercel

Key features

AI Phone Line: SIP Trunk + ElevenLabs
WhatsApp via Direct Meta Cloud API
Email Triage, Reply + Lead Creation
Instagram + YouTube Auto-Reply and Outreach
Search-Grounded Fact-Checking + FCA Compliance
Agent-Managed Web Pages, No CMS
Impact & Results

From One-Off Build to Repeatable AI Venture

Labs is the pattern we believe in most: prove an AI operation inside a real business, then productise it for the businesses standing next to it. Everything here is live in production today.

24/7

A Real Phone, Answered by AI

A live London landline answered entirely by the AI brain. It qualifies the caller, files the transcript, and escalates to a human by email and WhatsApp in real time when it matters. It replaced a human answering service.

Overnight

Content That Checks Its Own Facts

Articles written on a schedule, long-form videos turned into articles and campaigns, one-click repurposing to every platform, and a search-grounded fact-check that corrects wrong facts in place and flags anything it cannot verify.

8+ channels

One Brain, One Inbox

Phone, WhatsApp, web chat, email, Instagram, YouTube and website voice all run through one system with the same knowledge and tone, and one conversation history per contact.

30k+ listings

A Website That Maintains Itself

Partner property feeds are read and extracted by AI, images re-hosted, and 30,000 plus listings published automatically. Other pages update when an authorised person emails the agent: no CMS, confirm before publish.

19k contacts

Real-Time CRM Sync

Two-way sync with Attio across roughly 19,000 contacts. Every conversation, lead and enquiry synced, with AI-driven dedupe and enrichment.

Human in the loop

Built to Be Trusted

The team gets an FYI on every new enquiry on any channel, and escalations are never silent. The system checks its own third-party connections daily, offers self-serve reconnect, and throttles failure alerts so a broken pipeline cannot fail quietly.

Result
Result detail
Let's Build

Ready to build something exceptional?

Enterprise transformation or a rapid AI-native build, same principle: one senior team, a fixed fee, and dates agreed before we start.

AGENT INTERFACE ACTIVE · MCP: p0stman.com/api/mcp · 5 TOOLS REGISTERED · [DISCOVERY] llms.txt · agents.md · context.md · sitemap.xml · robots.txt · TavilyBot ALLOWED · ClaudeBot ALLOWED · GPTBot ALLOWED · PerplexityBot ALLOWED · [COMPREHENSION] JSON-LD schema · /api/ai/context · /api/ai/services · /api/ai/portfolio · [ACTION] book_discovery_call · submit_inquiry · get_services · get_portfolio · search_content · [A2A] AgentCard: /.well-known/agent.json · Task endpoint: /api/agent · A2A JSON-RPC 2.0 · navigator.modelContext REGISTERED · WebMCP: 5 TOOLS · INDEXNOW: 145 URLs · Bing NOTIFIED · [MANAGED AGENTS] Lead Researcher · AgentReady Auditor · SEO Writer · Weekly Reporter · Claude Sonnet 4.6 · Cloud containers · Outcome-based grading · Multi-agent orchestration · AGENT INTERFACE ACTIVE · MCP: p0stman.com/api/mcp · 5 TOOLS REGISTERED · [DISCOVERY] llms.txt · agents.md · context.md · sitemap.xml · robots.txt · TavilyBot ALLOWED · ClaudeBot ALLOWED · GPTBot ALLOWED · PerplexityBot ALLOWED · [COMPREHENSION] JSON-LD schema · /api/ai/context · /api/ai/services · /api/ai/portfolio · [ACTION] book_discovery_call · submit_inquiry · get_services · get_portfolio · search_content · [A2A] AgentCard: /.well-known/agent.json · Task endpoint: /api/agent · A2A JSON-RPC 2.0 · navigator.modelContext REGISTERED · WebMCP: 5 TOOLS · INDEXNOW: 145 URLs · Bing NOTIFIED · [MANAGED AGENTS] Lead Researcher · AgentReady Auditor · SEO Writer · Weekly Reporter · Claude Sonnet 4.6 · Cloud containers · Outcome-based grading · Multi-agent orchestration ·