Rule 01
Grounded in your material only
It answers from your pages and documents and shows which one it used. It is not a general chatbot with your logo on it, and it will not have an opinion about your competitors.
ServicesCheapest way in
The firm that replies first usually gets the work.
Speed to reply is the cheapest competitive advantage in a service business, and almost nobody uses it. Most small firms answer the next working day. This is one engine doing two jobs: a chatbot for your website that answers from your own material and qualifies, and an AI SDR in your inbox that replies within minutes at any hour. AI for customer service and AI lead generation, without the part where it invents an answer.
What this is worth, and who measured it
21×
A lead contacted within five minutes was 21 times more likely to qualify than one contacted after thirty. Most small firms reply the next working day.
7×
Firms that responded to a web lead within an hour were nearly seven times more likely to qualify it. In the same audit, 23% of firms never responded at all.
These are findings from published research on the wider market, measured by the people named beside each one. They are not results we are promising you, and we do not have client results of our own to show yet, which we would rather say plainly than dress up.
Three sizes, and what moves the number
Small
$2,000 to $4,800
usually two weeks
The inbox half on its own. Every enquiry answered within minutes at any hour, then followed up until someone replies. For a lot of firms this alone pays for the whole project.
Medium
most firms$4,800 to $11,000
usually three to five weeks
Both halves. An assistant on your site answering from your own material and qualifying visitors, with the inbox agent working behind it. Nothing reaches you unqualified and nothing sits unanswered.
Large
from $11,000
quoted after scoping
Answering from a large or messy document set, writing into several systems at once, or agents that take actions rather than only replying. More guardrails, more testing.
What moves the price
- How much material it has to answer from, and what state it is in. Twelve tidy pages is a different job from four hundred PDFs
- How many systems it writes into: an inbox alone is quick, an inbox plus a CRM plus a calendar is not
- How much qualifying you want before a conversation reaches you
- Whether it only replies or actually does things, because an agent that takes actions needs guardrails a chatbot does not
- How many languages, if more than one
If you are not sure which size you are, the free twenty-minute call will place you within about a thousand dollars, and it costs nothing to find out.
From nothing to answering, in two weeks
Days 1 to 3
We collect what it is allowed to know
Your pages, your prices, your policies, your FAQs, and anything else you would happily hand a new starter. That becomes the only material it can answer from, which is what stops it inventing things. We also agree the two or three qualifying questions worth asking before a conversation reaches you.
Days 4 to 8
Built, grounded, and wired to where you work
The assistant gets built against your material and connected to your calendar, your inbox and your CRM if you have one. The follow-up sequence gets written: how soon, how often, how many times, and exactly what makes it stop.
- Grounded means it answers from your documents and links to the page it used. If the answer is not in there, it says so and offers you.
Days 8 to 12
You try to make it say something stupid
You and your team ask it the awkward questions: the edge cases, the price you do not publish, the thing a competitor would ask. Everything it gets wrong gets fixed before a real visitor sees it.
Go live
Live, logged and reviewed
It goes live and everything is logged where you already work. A weekly digest tells you what it answered, what it booked and what it could not handle, and the third column is the useful one because it tells you what to add.
How it is kept honest
A site assistant that makes something up about your prices or your policy is worse than no assistant at all, so most of the design work is deciding what it is not allowed to do.
It answers only from material you gave it, links to the source, and says plainly when it does not know.
Rule 02
"I do not know" is a good answer
Tuned to hand over rather than guess, and every handover is a captured lead rather than a lost one.
Rule 03
Follow-up is finite and polite
A set number of messages on a schedule you choose, stopping the moment someone replies or asks it to stop. It never nags, because a nagging agent costs you more goodwill than the lead was worth.
Rule 04
It lives where you already work
Your inbox, your calendar, your CRM. We are not selling you another dashboard to check, and we will not make you migrate anything to build an AI agent that works with what you have.
What you get
№ 01
The site assistant
Answering from your own material, linking to its source, qualifying and booking.
№ 02
The inbox agent
Replying to enquiries within minutes at any hour, then following up on your schedule until someone answers.
№ 03
The knowledge set
The material it is allowed to answer from, written down and yours to edit. Several firms have found writing it clarified things they had never actually agreed internally.
№ 04
The weekly digest
What it answered, what it booked, and what it could not handle. The last column is the one that tells you what to fix.
What this is not
What it is not:
- Not an advice engine. It answers questions about your firm, not questions about the client’s situation.
- Not a general chatbot. If it is not in the material you gave it, it says so rather than reaching for the open internet.
- Not a cold-outreach tool. It follows up people who contacted you. We will not build you a spam machine.
- Not a replacement for a person on anything sensitive, and it is tuned to notice which those are.
Fair questions
How is this different from the chat widget our CRM already has?
Those usually answer from a canned FAQ list and give up on anything not in it. This one answers from your actual pages and documents, links to the source, qualifies against your criteria and books a real slot. If your existing widget can already do that, keep it, and we will say so.
Will it make things up?
It is grounded in your material and instructed to say when it does not know, and it shows the page it used, which is how you check. No system with a language model in it is perfectly immune, so the higher-value protection is the second one: it is not allowed to give advice or quote outside an agreed range, and anything sensitive goes to a person.
Have you built one of these before?
Not yet for a client, and we will not pretend otherwise. The engineering is squarely what we do and the price reflects that you are early. The first firm to buy one gets our full attention and a written case study out of it if they want one.
Could we build an AI agent ourselves?
You could, and for a simple internal one you probably should, because the tooling to build an AI agent is genuinely good now. What takes the time is not making it work, it is making it stay correct when nobody is watching, which means grounding it in your material, deciding what it may never say, and testing the awkward cases before a client finds them. That is most of what you are paying for here.
Is this an AI sales agent or a support chatbot?
Both, because they are the same engine pointed at different work. Facing outward it does AI lead generation: answering, qualifying and booking. Facing inward it handles the customer-service questions your team answers twenty times a week. Firms usually buy it for one and end up valuing the other more.
Can we start with just the inbox half?
Yes, and for a lot of firms that is the better first move, because a reply within five minutes beats a clever website widget almost every time. Say so on the call and the price comes down accordingly.
Start here
The cheapest way to find out if this works.
Tell us how long an enquiry currently waits before someone replies. That number is usually the whole business case.
John Weeks
The person who replies is the person who does the work.
Next on the path: Workflow Automation (Once the answering is handled, the next thing eating your week is the retyping.)