Skip to content
Swornim Sanjel
avernek.com/#demo
The Avernek demo: a Messenger window showing an inquiry in Nepali asking the weekend IELTS fee, the assistant's reply offering to hold a seat, and a timeline reading reply sent in 6 seconds, interest captured, qualified, team alerted

Plate 04 · Product · Founder-led sales

The product I built the company around

Businesses running ads collect messages across Facebook, Instagram, WhatsApp, web forms and marketplace listings. Avernek sells the system that answers them. My part is deciding what it does and what it must never do, sitting with the client to work out what the business actually knows, then pricing it and closing it. Our CTO builds it.

Role
Product definition, Client discovery, Pricing and closing, Team lead
Year
2026
Status
Delivered
Stack
Discovery interviews · Knowledge design · Escalation rules · Client delivery

Client names withheld pending permission.

The problem underneath the problem

Every business asked for the same thing, 'reply to messages automatically', and every business meant something different by it. A clinic needs to know which treatment, which branch, and whether it's urgent. A showroom needs to know the model, the budget and whether the person can actually visit. A property business needs the location and whether the enquiry is a buyer or a broker.

A single generic auto-reply is worse than no auto-reply, because it tells the customer they're talking to a wall. The system had to hold a specific business's actual knowledge and behave differently depending on what was being asked.

How a message becomes a qualified lead

Five of these steps are the world the business already lives in. The middle four are the system.

  1. Customer

    Sees an advert or a post and sends a message.

  2. Channel

    Facebook, Instagram, WhatsApp, web form, or a marketplace listing.

  3. Capture

    System

    Every message is recorded the moment it lands, so nothing depends on someone noticing.

  4. Intent

    System

    What is this person actually asking for, and how urgent is it?

  5. Knowledge

    System

    The business's own answers: services, branches, prices and hours, not general knowledge.

  6. Qualify

    System

    Serious buyer or browser. Scored, so the team knows who to call first.

  7. CRM

    One structured pipeline the owner can actually see.

  8. Human

    Escalated to a person with the context already gathered.

  9. Report

    Response speed, lead quality and follow-up status, back to the owner.

Where the product comes from

Our CTO builds the engine. My part is everything in front of it: sitting with the owner, working out what the business actually knows, and turning what's in their head into a structure the system can answer from.

That's most of the delivery time on every build. Which questions have exactly one correct answer. Which change by branch, service or season. And which must never be answered by a machine at all.

That last category is the one I care about most. Anything touching a medical opinion, a firm price commitment or a complaint goes straight to a person. Clients are usually surprised by how much we refuse to automate, and it's the part of the scope I argue hardest for.

Honest limits

This is presented as architecture rather than results. The performance figures belong to the clients who paid for the systems, and they haven't been cleared for publication here. Nothing in this case study contains a client's private workflow, credentials, or data.