Ask what AI can do for your business and you can make a list a mile long. While that's cool, it also can be a problem.
Capability stopped being the constraint a little while ago. AI can write, summarize, categorize, answer, extract, recommend, and draft. It can do most of these things well enough to be a useful return on investment. So "what can it do" isn't really the question anymore. The question now is which one of those things, pointed at which part of your business or organization, is worth the money and the ongoing upkeep. That's a judgment call, not technology, and it's the key part of the decision now.
Here are some places where I have and continue to see it paying off.
Communicating with your customers/visitors better
Every business answers the same set of questions over and over, and the person answering them can't be available all day every day. The useful version of AI here isn't a chat bubble in the corner of your site — it's tooling wired into the channels your customers already use. Email that gets sorted and drafted from your actual policies. Texts that confirm an appointment or answer "are you open Saturday?" without anyone picking up. Even a phone line that handles the routine calls and takes a real message on the more unique questions. What makes any of it work isn't just the AI. It's that it answers from your business data, prices, and policies, and passes the questions off to you if it doesn't have the answer.
Actually understanding your own business data
This is an underrated one and a great source of usefulness. You've probably got years of orders, jobs, and customers or contacts sitting in a database, and the only way you see any of it is by looking at the raw data or a report somebody built some time ago that's fallen out of date. It's powerful to ask a plain-language question and get an opinion on the best answers. Which services had the best margin last year, which customers bought twice and then stopped, what's the trend for sales or visitors, what improvements can we make based on our data, etc. You can get a real answer from your own numbers it's often worth more than anything customer-facing. It's also low-risk, because you're the one reading and guiding it.
New features your visitors actually notice
Search that understands what somebody meant instead of matching keywords. A configurator that turns "something around 12x16 off the back of the house, cedar if it's not crazy expensive" into a starting point your team confirms. Recommendations that reflect what people actually buy together. These are the features that make your site more useful than your competitor's, which is one of the only comparisons that matters.
Staying on top of SEO
SEO isn't one big project, it's the ongoing management of small consistent work that everybody intends to do and nobody does. This happens to be something that AI is good at: finding the pages with thin or missing descriptions, drafting service and location pages, flagging content that's quietly gone out of date or compliance. It drafts, you approve. It should never publish on its own, you're going to need to be involved with with some augmentation you'll find a powerful partner in the work.
Making your site easier to keep current
Here's a pattern I see constantly. A client stops updating their website — not because their CMS is bad, but because they use it maybe four times a year. That's long enough between visits to forget the login, forget where the page lives, and forget which field controls what. So the site goes stale, and I get a steady trickle of "can you just change this for me" emails.
That problem is as old as the industry, and it's one of the places where AI meaningfully changes what's possible. The honest answer to "how hard does it have to be to keep my own site current?" looks different today than it did two years ago. If that's the thing that's been quietly bugging you about your site, it's worth a conversation and I have something cooking here. Reach out to me today!
So how do you pick?
Three points to consider, in order.
Start where it already hurts — not where it's most impressive. The best first project is usually boring and specific: a task somebody does every week and quietly resents or that you don't get as much out of as you know you should.
Start where being wrong is cheap. Internal tools before customer-facing ones. When an internal tool gets something wrong, it's wrong in front of you and you can catch it. When a customer-facing one does, it's wrong in front of the person you were trying to impress.
Start small enough to change your mind. If the only version of the plan is a six-month build, that's a red flag to me. You should see something actually working early enough that you can back out or pivot before it gets expensive.
One reality check before any of it: none of this works if your information isn't somewhere we can reach. If your pricing lives across three spreadsheets and a guy named Dave (no offense, Dave), no model on earth is going to quote accurately. That's not an AI problem, it's a data problem and it's worth solving first, because it pays off whether or not you ever add AI into the mix.
The honest part
A fair number of the problems people bring me labeled "we need AI for this" turn out to be a well-built database query, or a form that should have existed years ago, or a report nobody ever automated. Those are cheaper, faster, and they don't break in interesting new ways. When that's the answer, that's what I'll tell you — and I'll write more about that soon, because knowing when not to reach for the capabilities of AI is most of the skill.
But when it is the right tool, it isn't magic and it isn't hype. It's one well-chosen feature, built into software that already does something useful, that returns value to you every day, week, or month.
If you're wondering about AI that's right for your business or organization, here's how I approach AI integration — or get in touch and we'll figure out together how AI and I can help you.
