Capabilities

What we build, and what working with us looks like.

Most AI conversations start with the technology. This page starts with the work you're already doing, and where a machine can take part of it off your hands.

What we build

Six kinds of work AI handles well in a small business.

These are the patterns worth paying for. Each one runs inside the software you already use — we connect to it rather than replacing it.

01

Customer messages

Incoming email, web enquiries, and text messages read, sorted, and answered — with anything unusual passed to a person alongside a written summary.

Typically for

Businesses where enquiries arrive faster than one person can answer them.

02

Paperwork & data entry

Invoices, forms, receipts, and scanned documents read and entered into your system, with anything ambiguous flagged for review rather than guessed at.

Typically for

Anyone retyping information that arrived as a PDF, a photo, or an attachment.

03

Scheduling & follow-up

Bookings taken against real availability, reminders sent, and the follow-ups nobody gets around to — quotes, reviews, repeat service — handled on time.

Typically for

Appointment and job-based businesses losing work to slow replies.

04

Quotes & proposals

Draft quotes built from your own pricing and past jobs, ready for you to adjust and send — so the estimate goes out the same day instead of next week.

Typically for

Trades and services where quoting is the bottleneck on winning work.

05

Answers from your own files

A place your team can ask plain questions and get answers drawn from your actual manuals, contracts, pricing, and past jobs — with a link to the source.

Typically for

Teams where one or two people are the only ones who know how things work.

06

Reporting that writes itself

The weekly numbers pulled together and summarised in plain language — what moved, what didn't, and what looks worth a second look.

Typically for

Owners rebuilding the same spreadsheet summary every week.

Not on this list? Bring it to the audit anyway. If AI isn't the right answer we'll tell you, and it costs you nothing to find out.

Where it runs

Self-hosted and on-premises AI, for work that can't leave.

Most AI consultancies have one answer: send your data to a big provider and trust the terms of service. That's fine for a lot of work and wrong for some of it. We build for both, and we'll tell you honestly which one you need.

When on-premises is the right call

  • Proprietary designs, tooling, formulas, or process knowledge that took years to develop
  • Client or contract terms that restrict where their material may be processed
  • Bid and pricing data a competitor would pay real money to see
  • Staff already pasting confidential material into consumer AI tools — the risk exists whether or not you sanctioned it

What an on-premises build involves

  • Sizing the hardware to the actual job — usually one workstation-class machine, not a server rack
  • Installing open models locally, chosen for what your work actually demands
  • Connecting it to your internal systems over your own network, with access limited to the people who need it
  • Handover with the maintenance your IT person — or ours — can actually keep up

The honest trade-offs

Running your own models costs more up front, and open models are not always as capable as the largest commercial ones. For document handling, drafting, search across your own files, and classification, the gap rarely matters. For the hardest reasoning work, it sometimes does. We'll show you the difference on your own material before you buy anything.

Usually it's a split

Very few businesses need everything in-house. The common answer is a line drawn through the middle: routine admin runs in the cloud where it's cheap, and the material that matters never leaves the building. Drawing that line correctly is most of the value.

What happens

An engagement, start to finish.

Most projects run two to four weeks. You'll know the price before any work starts, and you can stop after the audit owing nothing.

Step 1

Free · 30–60 min

The audit

We sit down with you and whoever does the work — not just the owner. We ask what a normal week looks like, where things pile up, what software you're already paying for, and which of your material is too sensitive to leave the building. You leave with a short written list of where AI would genuinely help, ranked by what it's worth against what it takes.

You need to

Give us an hour and be candid about what's annoying. No preparation.

Step 2

2–3 days

Scope and fixed quote

We pick one thing to start with — usually the highest nuisance with the lowest risk — and write down exactly what it will do, what it won't, and what it costs. One number, agreed before we begin. If you'd rather stop here and take the list to someone else, that's a fine outcome and you owe us nothing.

You need to

Read the scope and say yes, no, or "not that one".

Step 3

1–3 weeks

Build and connect

We build it against your real accounts and your real data, and we show you working pieces as we go rather than disappearing for a month. Nothing touches live customers until you've watched it handle real cases and agreed it's ready.

You need to

Grant access to the tools involved, and give us 20 minutes twice a week to check we're on track.

Step 4

2–3 days

Go live and train your team

We turn it on with you watching, then sit with the people who'll use it every day until they're comfortable — including how to switch it off, change what it says, and tell when it's got something wrong.

You need to

Put the relevant staff in a room with us for an hour.

Step 5

30 days, included

Settle in

Real use turns up edge cases no plan predicts. For the first month we fix what breaks and tune what's awkward at no extra charge. After that you can call us when you want to, or not at all — there's no retainer and nothing stops working if you don't.

You need to

Tell us when something looks off.

What you own afterwards

Everything.

  • The system itself, running in your own accounts, under your own logins
  • Plain-English documentation of what it does and how to change it
  • A recorded walkthrough for whoever joins the team next year
  • The ability to hand it to any other developer without asking us first

What we won't do

Worth saying out loud.

  • Replace your staff. These systems take work off people, not people off payroll — and we'll say so if that's the ask
  • Let AI make decisions that carry legal, financial, or safety consequences without a person signing off
  • Move you onto a new platform because it suits us. If the fix is a setting in software you already own, we'll tell you that instead
  • Take the job when AI isn't the answer. Sometimes the honest recommendation is a better process, not a better tool

Start with the audit.

Six questions, a real plan, no cost and no obligation.

Prefer to talk? Call (680) 271-4201 or email contact@mkilnai.com