The work that used to take five specialists

Marketing a trade or retail business used to be a staffing problem. Doing it properly meant people, or an agency that had people, and the cost of that decided how well most businesses did it.

That has changed more in the last eighteen months than in the decade before it, and most businesses have only half noticed. This is an honest account of what actually changed, what has not changed at all, and what it means whether you are an owner-operator with four staff or a retailer with a marketing team and an agency on retainer.

The five people you used to need

Running the growth side of a business properly has always required five distinct skill sets:

• A web developer to build the site and make changes to it
• A media buyer to run the ad accounts and watch them
• A content person to keep something going out consistently
• An appointment setter to answer enquiries and fill the calendar
• Someone on follow-up, which is the job nobody ever actually gets to

There were only ever three ways to cover those five roles, and each has a catch.

Hire them. Below a few million in turnover that is not affordable. Above it, it is affordable and you inherit a different problem: five specialists whose work you are not equipped to check.

Rent them. Commonly $2,000 to $3,500 a month for a smaller business, and a great deal more once you are a serious retailer. It works while you pay. The results stop the day the payments do, you never really learn what worked, and the production you are buying is usually the least valuable thing an agency does.

Do none of it, live on referrals and bought leads, which is a worse trade than it looks.

What actually got better

Three specific things changed, and it is worth being precise, because "AI is getting better" is the least useful sentence in business right now.

The models crossed a practical threshold. The question was never whether AI could write. It was whether checking its work was faster than doing the work yourself. For a large slice of ordinary business writing and building, that flipped. Below that line AI is a toy that costs you time. Above it, it is capacity.

Assistants became operators. The bigger shift is not model quality at all, it is access. A chat window answers questions. A coding agent works on your actual files, systems and accounts, with your permission, and does the job rather than describing it. That is the difference between asking how to add a service page and having the service page added.

Voice got good enough for the phone. An AI that answers at 7pm, sounds like a person, asks sensible questions and books the job is no longer a demo. For anyone whose enquiries arrive by phone, that is the difference between a booking and a missed call nobody ever knew about.

What did not get better

This part gets left out of most articles like this, which is exactly why it is worth reading.

Judgement did not improve. What you charge, which markets are worth chasing, whether your offer is actually good, when to walk away. None of that is a software problem and none of it got automated.

The wiring is still the hard bit. An AI with no access to your business writes you a nice email. An AI connected to the system holding your customers, calendar, phone number and pipeline can run things. Getting from the first to the second involves authentication, permissions and testing against real records, and done carelessly it is either useless or genuinely risky, because you are handing account access to something that takes actions.

Automation does not conjure demand. Converting enquiries and creating them are two different jobs. Automation does the first, advertising does the second. Both are buildable, and we build both, but the advertising half still needs a budget behind it and no amount of clever software changes that.

It will confidently do the wrong thing. Briefed badly, it produces something plausible and wrong, quickly. The skill that matters is not prompting. It is briefing clearly and reviewing honestly, which is the same skill that makes someone good at managing people.

If you already have a marketing team

Most writing about AI and business assumes there is nobody in the building who does marketing. Plenty of the companies we work with are well past that. They have a coordinator or a small team, and an agency on retainer for what the team cannot do.

For them the question is not whether they can afford the five specialists. They are already paying for all five, some in salary and some in retainer, and still waiting three weeks for a landing page.

The opportunity there is different, and honestly larger. A marketing person who can brief an AI properly stops being a coordinator who forwards requests to other people and starts shipping. The landing page goes live the same day. The nine ad variations get built in an afternoon instead of being negotiated into next month's scope. The report gets produced by asking for it.

That does not mean firing your agency, and we would not advise it. Strategy, a genuinely good creative eye and an outside opinion on your market are all worth paying for. What it means is that you stop renting the production. Keep the thinking, own the doing.

What this looks like when it is actually running

We are not theorising about where this is heading. This system runs our own company every day, so here is the specific version rather than the brochure version.

Our CRM is connected directly to Claude Code across more than 600 operations. Not a copy of the data or a nightly export, a live connection. It can send an SMS or email to a customer, read a call transcript, look across every contact, quote and appointment, build forms, custom fields, automations and workflows, and publish to the site and every social account.

Our Meta and Google ad accounts connect the same way, so campaigns, ad sets and creative are built and managed programmatically rather than clicked together. The unglamorous work collapses: nine creative variations instead of two, last week's search terms pulled and the junk added to negatives, checking that a conversion actually fired before anyone scales a budget. We catch our own tracking mistakes this way, which is less flattering and more useful than a case study.

Enquiries get answered in seconds rather than hours, by an AI that texts and calls, qualifies, and books into a real calendar. That is the part people feel first, because it is the part that was quietly costing them jobs.

This page, this article and the ads pointing at them were produced from the same seat.

Owning the system versus renting the outcome

Here is the distinction worth sitting with, because it decides how you spend the next three years.

When you rent an outcome, you buy leads or you buy an agency's labour. It works while you pay. Stop paying and demand stops the same week, you have built no asset, and the knowledge of what worked leaves with the final invoice.

When you own the system, it is slower to start and it compounds. The site, the follow-up, the database, the ad accounts and the AI answering the phone are yours. Nobody can switch them off, raise the price, or sell the same enquiry to two of your competitors.

We are not purists about this. Plenty of businesses sensibly run both for a while. But it is worth being clear with yourself about which one you are funding, because only one of them is still there in three years.

Ten businesses, September

The obvious objection to all of this is that reading about it changes nothing. That is fair, and it is why we built the workshop rather than another guide.

Trades AI-Q is two live days building this on accounts you own, not accounts we hold for you. Website, lead engine, CRM, follow-up and AI receptionist, working by Saturday afternoon, then sixty days of weekly calls to dial it in. Ten seats, Friday 18 and Saturday 19 September 2026, live online across Australia and New Zealand.

It suits an owner who wants to run this themselves, and it suits a marketing manager whose job would change completely if they could ship without waiting on anyone. If that person is not you, send them, or come together.

The founding cohort is $1,997 AUD ex GST, including the program, the template pack and the first two months of the platform. From cohort two it is $3,997. The founding price exists because the first ten businesses give us the feedback that makes the second cohort better, which is a real exchange rather than a countdown timer.

If you do one thing this week

Do not start with your website and do not start with the CRM connection. Take the most repetitive administrative job in your week, give it to an AI, review the output honestly and correct it until it is genuinely good enough to use.

That exercise teaches more than any prompt library, because it builds the only skill that transfers: briefing clearly and judging the result.

Then ask which parts of your business you are currently renting because they used to need a specialist. For most businesses that list is longer than expected, and considerably shorter than it needed to be two years ago.