Getting started
September 17, 2026 · 4 min read
Most AI readiness advice is written for companies with a data team. It talks about data maturity, governance frameworks, and platform strategy, and a ten-person business reads it and concludes that it is years away from being ready.
That is the wrong conclusion. For a small business, readiness has almost nothing to do with data infrastructure. It has to do with whether the work is visible, whether someone owns it, and whether it repeats. Here are the signs we actually look for.
Not perfectly. Roughly. A customer asks for something, it goes here, then here, this person checks it, it gets recorded there. If you can narrate that for at least one part of the business, you can find where AI might help. If every explanation ends in "it depends who is in that day," start by writing it down.
Annoyance is a good signal. The tasks people complain about are usually the ones that repeat, and repetition is what automation feeds on. If nobody in the business can name a tedious task, either the business is unusually well run or nobody has looked.
Ownership does not mean a job title. It means that if the task changed tomorrow, one person would notice and would be the one to decide what to do. Automation needs that person to design the exception path and to say when the system is wrong.
In an inbox, a form, a spreadsheet, a system, a document. AI can read, classify, extract, and draft from information that exists. It cannot recover a decision that was never recorded.
If you can say "a wrong answer here would be an invoice coded to the wrong job" or "a wrong answer here would be a reply that promises a delivery date we cannot meet," you can design the review step. If you cannot describe a wrong answer, you cannot check for one.
Every AI system starts under supervision. Someone has to read the drafts, check the extracted fields, and correct the classifications for the first stretch. That takes time from a real person. If nobody can spare it, the project will be unsupervised from day one, which is how it ends up producing confident mistakes nobody catches.
Some tasks should not be automated, and some businesses should wait. The readiness that matters most is the willingness to be told that and to act on it. A business that has decided in advance that AI must be adopted somewhere will adopt it in the wrong place.
"We need an AI strategy" is not a project. "We re-key every invoice by hand and it takes the bookkeeper most of Tuesday" is a project. If the goal has no task attached, the next step is to find one, not to buy anything.
Some work genuinely is. Bespoke consulting, complex negotiations, one-off projects. AI can assist that kind of work — drafting, summarizing, searching — but it cannot automate it, and a project that promises to will disappoint.
If two people describe the same task differently, the first job is to agree on which version is right. Automating a disputed process picks a side without anyone deciding to.
Customer personal information, health records, payment details, anything under a confidentiality agreement. If the task needs that data and you have no written rule for what may go where, write the rule first. It takes an afternoon. We have a short policy template to start from.
In almost every case, "not yet" is a small fix rather than a long wait. Write the process down. Pick an owner. Agree on the version that is right. Write two lines about data. None of that requires technology, and all of it makes the eventual project better. Businesses that do these things in a few weeks are usually readier than larger companies that have spent a year on a platform.
The most useful shift is to stop asking "is my business ready for AI?" and start asking "is this task ready?" A business can be ready to automate invoice intake and nowhere near ready to automate proposal writing. Judge the tasks one at a time, start with the readiest, and let the others catch up.
What AI consulting looks like for a Las Vegas small business, what a first call should cover, and eight questions to ask any consultant before you sign.
September 17, 2026 · 3 min read
What an AI opportunity assessment should produce: an operations map, a ranked register of impact and effort, readiness notes, a sequence — not a tool list.
September 17, 2026 · 3 min read
A practical order of operations for implementing AI in a small business: start with the work, not the tool, and know when a plain rule beats a model.
September 17, 2026 · 5 min read