Most AI advice starts with the tool. We start with your Tuesday.
You have been told AI will change your business and handed no idea where to start. So we start where the time actually goes: quoting, scheduling, intake, invoicing, the follow-up nobody gets to. We find the two or three places AI removes real hours — and say plainly which parts are not worth automating. You leave with a written plan you can act on with us, with another shop, or by yourself.
The assistant in the corner of this page is one we built and run. Ask it what an engagement would look like for your business — that conversation is the demo.
A worked example: the invoices nobody chased
This is the shape of a first engagement — an illustration of how we work, not a case study of a specific client.
How it runs now
A six-truck HVAC company finishes the month with thirty-odd unpaid invoices. The owner opens the accounting system on a Sunday night, cross-checks who has already paid, remembers which customers pay late but always pay, and writes each email by hand. Roughly three hours, done tired, and the awkward ones get skipped.
What we set up
An assistant that reads the unpaid invoices out of the accounting and payment systems, matches each one against that customer's own history, and drafts a reminder in the right register — gentle for the builder who always pays in week three, direct for the account that has slipped twice. Nothing sends on its own. The owner opens one queue, reads thirty drafts, edits two, and approves.
The Sunday night goes from three hours of writing to about twenty minutes of reviewing, and the awkward ones stop getting skipped — which is usually worth more than the time. Note what did not get automated: the decision to send, and the judgement about which customer gets which tone. Those stay with the owner by design, because a reminder sent to the wrong account costs more than it saves.
We build this kind of assistant on Claude, Anthropic's AI model, using its Agent Skills format — a way of packaging a repeatable business task so the assistant performs it the same way every time instead of improvising. Anthropic publishes a set of these for common small-business work: cash-flow snapshots, invoice follow-up, month-end close preparation, lead triage. We fit them to how your business actually runs, connect them to your real systems, and put an approval step wherever money or a customer relationship is involved.
How an engagement runs
We watch the task once
You show us how an overdue invoice actually gets chased today — which screen you open, what you look up, how you decide the wording for a good customer versus a chronic late payer. Fifteen minutes of watching beats an hour of describing.
We separate judgement from typing
Almost every "AI could do this" task splits into a decision only you should make and a pile of retrieval and drafting around it. The drafting is the part worth automating. The decision stays yours, and the plan says so explicitly.
We build the smallest working version
One task, wired to your real data, run in front of you. If it does not save real time on the first pass, we say so and stop — that is a cheaper outcome than a subscription you quietly stop using in March.
You get the plan either way
The written plan is yours at the end of the engagement whether you build it with us, hand it to someone else, or shelve it. We would rather be the shop that told you the truth than the one that sold you a project.
The boring half nobody sells you
The reason AI projects fail at small companies is rarely the model. It is that nobody wrote down the rules before switching it on. Every engagement includes a one-page policy your staff can actually follow, covering:
- Which systems the assistant may read, and which it must never touch
- What staff are allowed to paste into a chatbot — customer records, card numbers and health information generally are not
- Whether anything you send is retained or used for training, and how to check rather than assume
- Who approves an action that costs money or leaves the building, and how that approval is recorded
If you are in a regulated trade, this is also where HIPAA and client-confidentiality questions get answered before anything goes live, not after. See HIPAA IT support for how we handle that side.
Read it yourself
We would rather you verify this than take our word for it. These are the primary sources, published by Anthropic — not our summaries:
- What Claude isAnthropic's own overview of the model we build on.
- Agent Skills documentationThe technical format behind the repeatable tasks described above.
- Full developer documentationEverything Anthropic publishes for people building on Claude.
- Commercial termsThe terms that govern business use — worth reading before you sign anything.
- Privacy policyWhat happens to data you send. The answer to "is it training on my stuff?"
MAE-AI Solutions is an independent IT consultancy. We are not affiliated with, endorsed by, or a reseller for Anthropic. "Claude" and "Anthropic" are trademarks of Anthropic PBC and are used here only to identify the tools we work with. We build on other vendors' AI tooling too, and we will tell you when something other than Claude is the better fit for your problem. Product capabilities and terms are set by the vendor and change — the links above are authoritative, this page is not.
Thirty minutes, no obligation. Come with the task that annoys you most.