Custom software

AI Assistants & Chatbots

An assistant that knows your business, answers at 11pm, and hands over to a human when it should.

How it works

Pricing
Build fee plus monthly running cost, quoted after scoping
Timeline
Typically 2–4 weeks to launch

Running cost depends on conversation volume and is billed through at cost where we can. Assistants need review — an unattended bot drifts out of date the first time your hours or prices change.

The short version

Most of the questions your business gets asked are the same twenty questions. Are you open Sunday. Do you service my area. How much roughly. Can you take my dog / my car / my appointment on Thursday. They arrive constantly, a good share of them outside working hours, and every one that goes unanswered for twelve hours is a customer who has already called someone else.

An AI assistant handles that layer — and only that layer. Built properly it answers from your actual content: your service pages, your hours, your coverage area, your pricing rules. It says "I don't know, here's how to reach someone" instead of inventing an answer, and it hands over cleanly the moment a question needs a person or a commitment.

The failure mode we design against is the confident wrong answer. A bot that quotes a price you don't offer or promises a service area you don't cover costs more than it earns. So we constrain it to your material, test it against the questions you actually get, and give you a transcript of every conversation so you can see what it said in your name.

What people search for

  • ai chatbot for small business
  • ai customer service assistant michigan
  • website chatbot that answers questions
  • ai assistant for local business
Is this you?

Signs it's
time.

If none of these sound familiar, you probably don't need this yet — and we'd rather tell you that than sell you something.

The same questions arrive every day

Hours, coverage, pricing, availability. High volume, low complexity, and answering them is nobody's favourite part of the job.

After-hours enquiries go cold

A message sent at 9pm gets a reply at 9am, by which point the customer has moved on. Response speed decides a lot of local jobs.

Your team can't find your own information

Policies, procedures and specs spread across a shared drive, and the fastest route to an answer is asking whoever's been there longest.

Leads arrive unqualified

Every enquiry needs a call just to establish the basics before you know whether it's worth quoting.

What you get

What's actually in the box.

An assistant grounded in your content

Answers drawn from your pages, documents and policies rather than general knowledge, so it says what you'd say.

Explicit boundaries

A written list of what it will and won't answer. Prices it can quote, commitments it can't make, topics that go straight to a human.

Clean human handover

When it hits its limit it collects the details and routes to your phone, inbox or scheduler — with the conversation attached so nobody repeats themselves.

Lead qualification

Where you want it, it gathers the basics — location, job type, timescale — so the enquiries that reach you are worth the call.

Full transcripts

Every conversation reviewable. That's how you spot both the answers it got wrong and the questions you should add to the site.

An internal-facing option

The same approach pointed at your own documentation, so staff can find a policy or a spec by asking for it.

Commonly connects to

  • Your existing website
  • Email and SMS handover
  • Scheduling platforms
  • CRM lead capture
  • Internal document stores

Not an exhaustive list — if you run something else, ask. What matters is whether it exposes a usable API, and we check that before quoting.

How the work runs

No surprises, no eighteen-month projects.

  1. 01

    Collect the real questions

    We start from your inbox, call notes and chat history. The questions you actually get are rarely the ones an FAQ page was written for.

  2. 02

    Draw the boundary

    We agree in writing what the assistant may answer, what it must escalate, and what it must never say. This is the part that decides whether it helps or hurts.

  3. 03

    Build and adversarially test

    We try to make it say something wrong — out-of-area requests, prices you don't offer, questions with false premises — and tighten it until it declines properly.

  4. 04

    Launch and review transcripts

    It goes live, you read the first weeks of real conversations, and we correct what real customers expose.

What it looks like in practice

Same problem, different trades.

Home services

Before

Evening enquiries asking about coverage area and rough pricing sit unanswered until morning.

After

Immediate answers on coverage and ballpark ranges, with qualified jobs booked into the morning callback list.

Clinics & practices

Before

Reception time absorbed by questions about hours, insurance and what to bring.

After

Routine questions handled on the site; the desk keeps the calls that need a person.

Restaurants & hospitality

Before

Repeat questions about allergens, seasonal hours and private bookings across phone and social.

After

One assistant answering from the current menu and hours, with booking requests routed properly.

Trades with a large catalogue

Before

Staff searching a shared drive for spec sheets and warranty terms mid-job.

After

An internal assistant that returns the right document and the relevant passage.

Representative examples of the problems this work solves, not case studies of specific clients.

Questions

Straight
answers.

What stops it from making things up?

Two things. It answers from your material rather than from general knowledge, and it's explicitly instructed to say it doesn't know and hand over rather than guess. We then try to break it before launch — asking about services you don't offer, areas you don't cover, prices that don't exist — and tighten it until it declines cleanly. No system is perfect, which is why you get transcripts.

Will customers know they're talking to a bot?

Yes, and they should. We label it plainly and make the route to a human obvious. Pretending otherwise annoys people and, in some contexts, creates legal exposure you don't want.

What does it cost to run?

There's a build fee and a monthly cost that scales with conversation volume. For most local businesses the running cost is modest — smaller than the phone line it takes pressure off — but it isn't zero, and we'll estimate it from your actual enquiry volume before you commit.

Can it book appointments directly?

It can, if your scheduler supports it — that's an integration on top of the assistant. Some businesses want that; others prefer the assistant to qualify and hand over so a person confirms. We'd generally suggest starting with handover and adding direct booking once you trust it.

How do we keep it current?

It reads from your site and documents, so ordinary updates flow through. Bigger changes — new services, new coverage, changed pricing rules — need a deliberate update, which is part of the monthly plan.

Discovery call · free

Tell us what
isn't working.

Half an hour on the phone about the process that's costing you time. You'll leave knowing whether ai assistants is the right answer — and if something you can buy off the shelf would do the job, we'll point you at it.

What happens next

  1. 01You walk us through the process that hurts
  2. 02We tell you whether custom is even the answer
  3. 03If it is, you get a written scope and a fixed price

Locally owned, working across Mid-Michigan.