An AI Sales Agent That Runs a Publishing Business's Inbound Revenue
63 sales inboxes, one AI agent, about $105 a month. It answers every inbound enquiry for a digital publishing business, quotes the right price for the right site, handles objections, collects the content, chases the invoice and makes the upsell, with no human in the loop for a standard deal.
Before: 63 inboxes and a team that could never catch up.
The operator is a digital publishing business with 63 news and blog sites. It sells sponsored editorial placements on them, and every site has its own sales inbox. Every inbox gets enquiries, every day.
For years that meant a team of people doing the same loop by hand: open the email, work out which site it came in on, look up that site's price, reply, negotiate, field the "can you do it cheaper?", collect the article, chase the payment, then remember to offer the next order. Multiply that by 63 storefronts.
The human cost showed up in predictable places:
- Replies went out late. An enquiry that arrived on Friday evening sat until Monday, by which point the buyer had ordered somewhere else.
- Quotes drifted. Different people quoted different numbers for the same site, and discounts crept in to close deals faster.
- Threads got lost. With thousands of past orders spread across dozens of inboxes, nobody could remember what a returning buyer paid last time or whether they still owed for the previous one.
- Growth meant hiring. Every new site added more inbox to cover. Revenue could only grow as fast as headcount, and the team was already costing $9,000–9,500 a month.
The brief was blunt: replace the inbound sales function with one AI agent that represents every site correctly, holds a natural conversation, closes on fixed terms and never loses a thread.
What we built: one brain wearing 63 faces.
One engine, 63 identities
A single conversation engine sits behind every inbox. When an email lands, the system detects which site it arrived at and loads that site's name, pricing tier and rules before a reply is written. The buyer always feels they are talking to the publication they contacted, because, as far as the reply is concerned, they are.
Fixed pricing, so margin is never negotiated away
The biggest risk in sales automation is an agent that gives away money to close. So pricing is fixed and tier based. When a buyer pushes back, the agent answers with value instead of a discount: it adds a complimentary placement from an entry tier. The deal closes and the price holds.
A sales motion broken into stages
Selling was split into swappable stage modules: first contact, objection handling, fulfilment, payment and upsell. The agent always knows where a conversation sits and what it is trying to achieve next. Changing how one stage works means editing one module, not rewriting the agent.
Hub and spoke architecture
Each of the 63 inboxes runs a lightweight listener with one job: catch new mail and pass it to the core. All 63 feed a single main processor that carries the sales logic, the stage flow and the reasoning. Two supporting workflows run beside it:
A memory that never forgets a buyer
Everything the agent knows lives in a structured data layer: sites, clients, live sales, outstanding invoices, bundles, retainers, the stage modules themselves and a full thread and message history. A returning buyer gets an agent that remembers their last order. Accounts flagged as risky are invoiced before fulfilment, not after. The conversation model was chosen for speed and cost at volume, because inbound replies are frequent and need to feel instant.
What changed: about $105 a month doing work that cost $9,000+ a month.
The system runs the standard deal from first email to paid invoice with no human touch, across all 63 sites, day and night. Total running cost is about $105 a month. The sales and admin function it replaced cost $9,000–9,500 a month in wages.
It manages more than 4,300 client records and thousands of historical orders without mixing up threads or misquoting a site. Because prices are locked and objections resolve through added value, margin is protected on every conversation it handles.
Adding a 64th site is now a configuration job: a new listener and a new row of site rules. No new hire.
Day to day now.
A buyer emails one of the sites at 11pm asking about a placement. They get a reply within minutes, in that site's name, with the correct price. They ask for a better deal and are offered an extra placement instead. They send their article, receive an invoice, and if they forget to pay, the chase engine nudges them on schedule. When the payment lands, the placement goes live and the next offer arrives.
The people at the business see the exceptions: unusual requests, custom deals, anything outside the fixed terms. Everything standard is handled before anyone opens a laptop.
The same pattern, in your business.
You may not have 63 inboxes. You probably have one shared inbox that nobody owns, a price list in someone's head, and quotes that go cold because the follow-up never happens. The pattern is the same: one engine that knows your offers and your rules, replies fast, follows up on time and hands your team only the conversations that need a person.
- If you run a clinic or salon, it answers booking and price enquiries out of hours, sends the right treatment information and books the slot.
- If you run a trades or contracting business, it qualifies each enquiry, sends the standard quote for standard jobs, and chases the ones that went quiet after three days.
- If you sell B2B services, it handles first replies, sends the right package, issues the invoice and reminds unpaid accounts without anyone writing the awkward email.
- If you run several brands or locations, one engine can speak as each of them, with its own prices and rules, exactly as it does here.
Not sure which conversations are safe to hand to an agent and which need a human? That is the first question an AI Consultation answers.
AI Automation
This is AI Automation applied to revenue: one system that reads every enquiry, quotes, handles objections, invoices and upsells, while your team handles the deals that need judgement. We build the same inbound engine around your offers, your pricing and your inboxes.