One consultant, one AI coding assistant, all of the marketing.

Why not do it myself, or hire it out

Work I would never have attempted, done correctly.

I am an AI and machine learning consultant, not a web developer, a hosting administrator or a paid-search specialist. Each of those comes with a learning curve steep enough that I would never have tried them myself, and I could not have come close to doing them correctly. The alternative was hiring professionals for each, at many times the cost, and they would very likely have been using AI to do the work anyway.

With the assistant, the work gets done quickly and gets done right, and I stay close enough to it to know why each choice was made.

The website

A dated site rebuilt, then rewritten around the buyer.

The old site was a WordPress install that had not kept up. The assistant rebuilt it as a fast static site, and since May it has written 78 of the 79 changes made to it. I set the direction and wrote the sentences that matter; it did the building, and a great deal of rewriting. On one day in September it made 23 changes, most of them reworking the copy around what a buyer is trying to decide rather than around what I have done.

It also worked out the hosting, DNS and analytics settings the new site needed, and wired up conversion tracking on the contact form. Along the way it noticed something I had missed: crawlers visiting a mirror copy of the site were being counted as real visitors, which quietly inflated the traffic numbers I was making decisions on. It fixed the counting and checked the fix on the live site.

Being found

For the people already searching for someone like me.

Two channels serve people who are actively looking. Google Ads puts the practice in front of someone searching for an AI consultant. My Google Business Profile, which the assistant wrote and keeps current, is the listing that makes the practice visible in Google Maps and local search.

With Ads, the assistant earns its keep by reading, not by writing ads. I paste the Ads and analytics reports in and it reads them, the job nobody at a one-person firm has time for. It found that 63% of my spend had gone to searches from people who were never going to hire a consultant, most of them let in by Google's "close variant" matching, which quietly widens what a keyword matches and can no longer be switched off. It then wrote the negative-keyword lists and a rebuilt campaign. It has no access to the account: every change is applied by me, in Google's interface, after I have read the reasoning.

Doing the finding

For the people who are not searching, yet.

Many of the firms I can help are not searching for anyone. They have a reason to, a new AI leader, a vendor just chosen, a program announced, and have not gone looking yet. So I built an AI prospecting pipeline, with the assistant, that reads business news and conference agendas for firms with a dated reason to hear from me, checks each firm on its own website, ranks the people there and drafts a first note. Five weeks of it cost $123 in AI model calls, every call logged with its model, tokens and price. It drafts; I edit and send every message myself.

The two sides complement each other. Ads and the Business Profile get me found by people who are looking; the pipeline finds the ones who are not.

The pipeline also grades itself, and the grade is not always flattering: the AI that ranks prospects is measured against a simple formula, and so far the formula is winning. That story, and what it costs to run each step, is written up separately.

Read how the pipeline grades itself · The code

Keeping it consistent

One description of the business, kept true in four places.

What I offer, and the kind of client I want, is written down in four places: the website's pages, the Google Ads keywords and ad copy, the services on the Business Profile, and the configuration that tells the pipeline which firms and people to look for. They drift the moment one of them changes and the others do not, and keeping them in step by hand is exactly the chore a one-person firm lets slide.

Now a change to an offer or to the client I am after goes through all four in one pass. The assistant edits the site and the pipeline's configuration directly, and writes the Ads and Business Profile changes for me to apply, which cuts the work to a fraction of what it would take by hand.

The bigger lesson

Not the typing: the judgment calls nobody has time for.

The most valuable thing AI does here is not writing code or copy. It is taking over the mundane, time-consuming judgment calls: which of hundreds of firms is worth my time this week, who there to write to, and what to say first. I still make the final call on each one, but I start from a short list and a draft instead of a blank page.

If your business has judgment calls like these eating hours every week, automating them is the kind of work I do. See how I work, or get in touch below.