Is Your Agency Just Using AI to Do All the Work?
A question we hear a lot lately, and one plenty of business owners are probably thinking without asking, is whether their agency is just using AI to do all the work.
That deserves a straight answer. We use AI tools every day, and they are not doing the work.
What they do is handle the parts of the job that used to eat entire afternoons: pulling data, cross-referencing exports, sorting hundreds of URLs by performance. That work still has to happen. It never required strategic thinking, though, and a strategist spending eight hours on spreadsheet assembly was never a good use of a client’s budget.
What AI hasn’t changed is who decides what to do with the answers. That’s still us, and it’s still the part that determines whether your phone rings.
What AI Actually Changed in Our Workflow
Building a full content performance picture for one client used to mean exporting query and page data from Google Search Console, pulling position and keyword data from SEMrush, exporting engagement and conversion data from GA4, then stitching all three together in a spreadsheet with a stack of VLOOKUPs. On a 200 page site, that was most of a day. On an 800 page site, it was a multi-day project we had to plan around.
Now we hand those same exports to Claude with specific instructions about what to cross-reference and how to rank it, and we have a usable, sorted picture in under an hour.
The work we hand to AI tools looks like this:
- Cross-referencing large data exports from Search Console, GA4, and SEMrush to surface patterns across hundreds of pages at once
- Sorting and ranking pages by opportunity so we know where to look first
- Flagging technical and on-page issues across a full site crawl
- Summarizing what competitors cover and where their content stops
- Research and first-draft outlining against a keyword strategy we’ve already approved
- Turning raw reporting numbers into plain-language explanations clients can actually read
Every one of those is a processing task. High volume, repetitive, and rewarding speed and consistency rather than judgment.
What AI Doesn’t Do
An AI tool doesn’t know that your best technician just quit, that you’re trying to move away from one-off residential calls, or that a commercial account in the next county is worth four times what a homeowner job is worth. It doesn’t know your margins or your close rate.
It also doesn’t know what your business should be trying to win at. That comes from experience with businesses like yours, which means years of watching what happened after the plan went live. Our team has spent that time on pest control, roofing, plumbing, HVAC, and electrical accounts around the country. AI can process a keyword export in ninety seconds. It can’t tell you that a high volume search term is a bad target because the people searching it are price shopping and won’t convert.
The tools don’t decide:
- Which services and locations are worth competing fo
- What digital channels are truly the best use of your investment in a local context
- What a page needs to say to convince a homeowner to call instead of clicking back to Google
- Whether a technical fix is worth the development hours it will cost
- When the data is showing something real versus noise from a seasonal swing or an algorithm update
- What goes live on your website
Nothing reaches a client site without a person reviewing it, editing it, and signing off, whether that’s a paragraph of copy, a redirect, or a title tag. The tool produces a starting point. A strategist decides what happens with it, and our name is on the result either way.
Every agency or individual now has access to the same tools. What gets done with the output is the difference.
get startedThe Process, Start to Finish
Every project on our end runs through the same six steps, and AI shows up in one of them.
- Define the goal before any AI tool opens. Grow a specific service in a specific market, recover rankings after a redesign, improve lead quality instead of lead volume. The goal shapes every question we ask after it.
- Pull the data that answers the question. Search Console, GA4, SEMrush, call tracking, Google Business Profile insights, ad platform reporting, 1st party customer data. We choose the sources that matter for the question instead of dumping everything into one pile.
- Give the AI a defined job. We tell it exactly what to cross-reference, what to rank by, and what format to return. What comes back is a sorted view of information we defined, not a recommendation.
- Review the output ourselves, every time. A sorted list of pages isn’t an answer. It’s a place to start looking, and the analysis happens when a strategist reads it.
- Implement by hand. Redirects, content edits, internal links, technical fixes, page builds. A person guides the work and a person checks it.
- Measure against the goal from step one. Rankings, traffic, conversions, and lead quality, tracked in the marketing dashboard so results stay tied to numbers the client can see.
Step three is the one people picture when they hear “AI marketing,” and it’s the narrowest part of the process.
What This Looked Like on a Real Account
Here’s a recent example. The client is a home services company with a strong primary market and a secondary target city about an hour out.
The secondary city was underperforming, and not because content was missing. The site had built up a fair amount of material about that city over several years, including service pages and a stack of blog posts. The main service page for that location still averaged a position above 8. That’s page one on paper, but position 8 pulls about 2.1% of clicks while position 2 pulls 18.7%, and the top three results take roughly 69% of clicks combined.
We identified the opportunity first. The market had enough search volume to justify the effort, the client had crews who could service it, and the competitors ranking above them weren’t especially strong. That read came from a strategist reviewing the account.
Then the tools went to work:
- Pulled query and page level data from Search Console, position and keyword data from SEMrush, and engagement data from GA4
- Handed all of it to Claude with instructions to aggregate every page and article touching that city and rank them from highest potential impact to lowest
- Got back one sorted list instead of five disconnected exports
Reading that list took about twenty minutes. Finding the same thing by hand would have taken two days. Several of the older blog posts were near duplicates of each other, and a few were competing directly against the stronger service page for the same queries, which split the signals across pages that should have been one page. Keyword cannibalization, built up gradually over years of content production.
Fixing it was human work:
- Identified which pieces were genuinely redundant and which had unique value worth keeping
- Redirected the weak duplicates into the stronger, better-matched pages so their equity consolidated instead of scattering
- Reworked on-page SEO on the surviving pages, including titles, headings, and content structure
- Rebuilt the internal linking around that city so the primary service page was clearly the most important page on the topic
One month later, the primary service page for that city had moved from an average position above 8 to an average position of 2. Using the click-through rates above, that’s roughly nine times the click share from the same search volume.
AI didn’t find the cannibalization. It produced the view that made the cannibalization visible. Human strategy started the project, AI compressed the analysis from days into an hour, human expertise implemented the fix, and the client saw movement in weeks.
Why This Matters When You Hire an Agency
Every agency has access to the same tools. That wasn’t true of the software that used to separate good agencies from bad ones, and it changes what you should be evaluating. The tools aren’t the differentiator anymore. What gets done with the output is.
There are two ways agencies get this wrong.
The first is using AI to skip the work instead of speeding it up: content published without anyone reading it, meta descriptions generated in bulk with no thought about what each page is for, blog posts that repeat your city name a dozen times and tell a customer nothing. That approach is usually easy to identify, and Google has spent years getting better at identifying it too. Thin pages published at volume can pull down the pages that were actually working. We covered more of that pattern in our breakdown of common SEO mistakes small businesses make.
The second is refusing to use the tools at all and billing for the hours anyway. If an agency spends twelve hours a month assembling reports by hand, that’s twelve hours of spreadsheet work being paid for at strategy rates.
What Actually Changed About SEO
The claim that AI kills SEO has the direction backwards. Search behavior is shifting, and AI answers in search results are real and worth planning for, which is why we’ve written separately about getting your business recommended by AI and why SEO is not dead.
What lost value is doing SEO thoughtlessly. When anyone can produce 500 words about water heater repair in a few seconds, producing 500 words about water heater repair isn’t worth much. The work that still moves rankings is the work that was always difficult: knowing which pages matter, understanding what a customer is trying to figure out, fixing the technical foundation, and making decisions you can defend six months later.
Faster tools don’t make that judgment easier to come by.
Want to See How We’d Approach Your Account?
We can walk you through this same process on your own site, including where your content may be competing against itself and which pages have the most room to move.
Call us at (207) 813-4735 or contact us online to get started.
You can also look at full results from accounts like this one in our case studies, or read more about how we work.
