The tools are the same. The results aren’t.
You can open the same AI tools an agency uses this afternoon. Ask for a homepage, a brand message, or a content plan, and something plausible appears within minutes. That is genuinely useful.
It is also where the comparison usually stops. Two people can use the same tool and get very different outcomes, for the same reason two people with the same camera don’t take the same photographs. The tool removes effort. It doesn’t supply judgment.
This isn’t an argument against using AI yourself. It’s an explanation of why the results differ, so you can decide where your time is best spent.
A good answer starts with a better question.
AI responds to what you give it. Ask for “a homepage for my plumbing business” and you’ll get the average plumbing homepage: friendly, competent, and indistinguishable from the last ten.
The useful questions come earlier. Which customers are worth more to you? What do they compare you against? What makes them hesitate? What do you want to be known for, and what are you willing to stop saying? Those answers rarely fit in a prompt. They come out of conversations, records, and a fair amount of disagreement.
Much of the difference shows up here: in knowing which questions matter for this business, and noticing when an answer is missing.
Plausible isn’t the same as right.
AI is very good at producing work that looks finished. The headline reads well. The page has sections. The plan has phases. Everything sounds reasonable, which is exactly what makes the problems hard to spot.
A claim you can’t support. A message aimed at the customers you’d rather not attract. A form that works but sends inquiries nowhere useful. A structure that will be awkward to change in six months. None of these announce themselves.
Knowing what to reject matters as much as knowing what to ask for. That comes from having seen what happens after launch, more than once.
Speed helps when you know where you’re going.
The tools make each step faster. They don’t tell you which steps to take, or in what order.
Without a plan, faster usually means more: more drafts, more pages, more options to choose between. It is easy to spend a week producing material and finish with nothing decided. With a clear direction, the same speed turns into fewer rounds and better decisions.
A good process is largely a record of where projects tend to go wrong, and what to settle before that happens. AI can make that process quicker. It doesn’t replace it.
Someone still owns the outcome.
When something goes wrong, the tool doesn’t take the call. Someone has to notice the drop in inquiries, work out why, and decide what to change.
Hiring a professional is partly hiring that responsibility: checking the details, testing the important journeys, connecting the website to the rest of the business, and staying accountable when the evidence says something isn’t working.
You might not see most of that work. Done well, it mostly looks like nothing went wrong.
Use AI. Be honest about the job.
If you have the time, enjoy the work, and the stakes are modest, doing it yourself with AI can be a sensible choice. Start small, test what you publish, and keep notes on what customers actually respond to.
If the website, brand, or marketing carries real weight in how the business wins work, ask a different question. Not “can I produce this?” but “can I tell whether it’s right?”
That is usually the difference people are paying for, whether they realize it or not.
Put the thinking to work
Check one piece of AI work.
Pick something AI helped you write or build. Note who it is aimed at, what it claims, and what a customer should do next. Mark anything you can’t confirm or wouldn’t say to a customer in person. That list shows where judgment is still needed.
See how we approach the work
Written by
Dan Reed
Founder of Multicomma. I design and build websites, ecommerce experiences, and custom tools, with a focus on what the business needs them to do.
More about working together
