What we build

AI Assistants and Chat

We build website chat and private team assistants that answer from your own services, prices and documents, capture leads at any hour and hand real opportunities to a person.

Why it pays

Every visitor gets an answer. Your team gets their day back.

We start with the questions your business actually gets, where they arrive, and what a good answer requires before anyone trusts it. Then we choose the model, ground it in your services and policies, connect your scheduling or CRM, and set handoff rules. Day to day, routine questions resolve themselves, and your team only sees the conversations that need a person.

What to expect

What to expect from an AI assistant for your business

Expect routine questions answered at any hour and visitors captured as named leads. Staff find answers without asking around, and a person steps in on anything uncertain.

99.9 percent

Uptime target on hosted builds, with monthly plans for updates and maintenance

Every chat logged

Transcripts you can review, with unanswered questions listed for you

40 chats

Example: 40 chats a week on a small business site, assumed, handled without staff time

Figures labelled Example are worked from assumptions, not client results.

Answers only from approved content

The assistant draws only on the services, prices, policies and FAQs you approved. When the answer is not there, it says so and offers a person.

Website chat that captures leads

A chat window on your site answers visitors at any hour. It collects name, number and the job in mind, then sends the lead to your CRM.

Books appointments in chat

Visitors pick a time from your real calendar inside the conversation. A confirmation text or email follows, so interest becomes a booking right then.

Private assistant for your team

Staff ask in plain language and get answers from your procedures, price books and past jobs. Each answer links to its source document.

Handoff with the full chat

Pricing disputes, complaints, urgent jobs and anything the assistant is unsure about go to a person by text or email. The whole conversation comes attached.

Easy-to-edit knowledge base

Update a price or policy in one place and every answer follows. We show you which questions went unanswered so you know what to add.

Scenarios

Four hypothetical scenarios

These are hypothetical, not client results. Each one shows the shape of the work in a different trade.

01

Auto repair scenario

Drop-off and status questions answered on the shop's website

An independent auto shop adds a chat window where customers ask about services, price ranges and wait times. The assistant answers from the shop's approved sheet and books a drop-off slot. Anything about a repair in progress goes to the service advisor.

Assumed in this scenario

  • About 25 chats a week handled
  • Drop-off slots booked overnight

02

Med spa scenario

Treatment, pricing and aftercare questions answered, consults booked

A med spa's chat widget explains each treatment, price range and downtime from copy the owner approved. It books consultations and collects each client's goals and contact details. Anything about medical history goes straight to a staff member.

Assumed in this scenario

  • Consults booked outside office hours
  • Price ranges quoted from approved copy

03

Construction scenario

Project managers ask questions of the contract and spec book

A mid-size builder loads each project's contract, specifications and submittal log into a private assistant. A project manager asks who approves a substitution or what the retainage terms are. The answer comes back with the page and clause cited.

Assumed in this scenario

  • Answers cite the page and clause
  • About 15 questions a day handled

04

Accounting firm scenario

Staff find the firm's own procedures without interrupting a partner

A six-person firm uploads its checklists, engagement templates and how-we-do-it notes. New staff ask which form a client needs, how a recurring task is handled or who owns a step. Each answer arrives with the source document linked.

Assumed in this scenario

  • About 10 questions a day answered
  • Every answer links its source

How it is built

How the assistant is built, from your content to the chat

Six layers sit between your documents and the answer a visitor reads. Each one narrows what the assistant can say until only approved facts get through.

Your approved content

We collect your services, prices, policies and FAQs. For the team assistant, we add procedures and past jobs. Then we clean everything into short, clearly labeled entries.

Searchable knowledge index

Entries are stored in a searchable index. Each question pulls only the few passages that matter, and the answer can point to its source.

Tone and answer rules

A short instruction set fixes the tone and the topics it may cover. It also lists what the assistant must never promise and when to say it does not know.

The language model

A language model writes the reply from the retrieved passages and the rules. We choose one that fits your volume and budget, and we can change it later.

Channels and actions

The assistant lives in a chat window on your site, a private page for staff, or both. It can book, capture a lead or send a handoff.

Handoff and review

Uncertain or sensitive chats pass to a person with the transcript. Every conversation is saved, and a review list shows questions it could not answer.

Built for operations

Guardrails that keep the assistant accurate and on topic

An assistant that talks to customers needs rules before it needs polish. These limits go in first, and every conversation stays on record for you to review.

You sign off on every fact

Services, prices and policies enter the knowledge base only after you approve them. Later edits follow the same sign-off before they go live.

Topics it will not discuss

Refund disputes, legal or medical questions and anything you list stay off limits. The assistant offers a handoff the moment a visitor goes there.

Role-based access for staff

The private assistant sits behind a login. Each person sees only the documents their role allows, so payroll files and client records stay with the people who should have them.

Transcripts you can review

Every chat is stored with its time, the passages used and the outcome. A weekly list shows unanswered questions, so gaps in the knowledge base get fixed.

Fallback if the model is down

If the language model is unavailable, the chat window switches to your phone number and a contact form. The outage is logged.

Knowledge base you keep

Your knowledge base, instructions and chat settings are documented and handed over, and the statement of work states what you own. A monthly plan covers updates and maintenance.

Measure it

What you see in the numbers and the transcripts

What you track

  • Questions answered without staff time, compared with the questions your team handled before launch.
  • Leads captured after hours, with name, number and job, and how many turned into bookings.
  • Handoffs to a person, grouped by reason, so you see which topics deserve a better answer.
  • Staff questions to the team assistant, a rough count of the interruptions it absorbs each week.

What the system keeps

  • Every conversation, with timestamps, the passages the answer drew on and the page it started from.
  • Questions with no approved answer, collected into a list of gaps to fill in the knowledge base.
  • Each lead captured and each booking made, with the contact details and where they were sent.
  • Every change to the knowledge base, with who approved it and when it went live.

Worked example

Twenty routine questions a day add up to an hour of staff time

Say your office fields twenty routine questions a day at three minutes each. That is an hour a day, or twenty hours across a twenty-day month. A chat that resolves half of them gives back ten hours.

10 hrs

Worked example from the stated assumptions, not a client result

Before any work

A free walkthrough and a written quote come first

We start with a free consultation and a look at the questions your business gets. You receive a written quote and statement of work before any work begins, then three rounds of revisions after the first build.

Written quote

In writing, before any work begins

Questions

The questions we hear first

What can the assistant actually cover?

Start with whatever eats the most time, usually website chat and the questions your team repeats internally. The assistant answers from your approved services, pricing rules, policies and documents, and it can book appointments and capture leads. Anything outside that scope gets handed to a person rather than guessed at.

Is this worth it for a small local business?

Usually, yes. If your team repeats the same answers daily, or after-hours visitors leave unanswered, there is real time and revenue on the table. If you only get a handful of inquiries a week, a simpler fix may come first, and the free walkthrough will say so.

How do security, control and cost work?

You stay in control. The assistant answers only from approved content, every conversation is logged, and actions like booking follow rules you set. Builds typically run $1,000 to $5,000 on a written quote, or $50 an hour invoiced weekly. A monthly plan covers updates, maintenance and a 99.9 percent uptime target on hosted builds.

First move

List the questions your team answers every day

Describe the job you want off your plate. You get back what to build first, what it costs and what it should change, in writing.