It is a fair question, and I get asked it more often than I did a year ago. If AI can write your ad copy, and Google's own tools can set your bids, and a cheap subscription can generate a month of social posts before lunch, what exactly are you paying an agency for?
The American Marketing Association released its 2026 State of Marketing Careers Report this month, built on a survey of 1,412 marketing professionals. It is the most honest read I have seen on this, partly because it is not flattering to people like me. It says plainly that most of what agencies have historically sold can now be automated.
It also draws a clear line, and the line is the useful part.
What AI actually took over
The AMA scored every marketing skill against Stanford's Human Agency Scale, which runs from AI handles this alone to this cannot function without continuous human involvement.
Email marketing came out fully automatable. Just above it, needing only minimal human input: search engine optimization, paid media, social media, copywriting, graphic design, lead generation and performance analytics.
That is most of a traditional retainer.
At the other end of the same scale sat marketing strategy, brand management, critical thinking and judgment. The AMA sums the shift up in one line: Companies are not hiring less judgment. They are hiring less execution.
I think that is exactly right, and I would rather show you what it means than argue it.
Two problems AI would not have caught
A few years ago I worked with a company that sells home accessibility equipment. Their previous agency had built the campaigns around their address. It was an office, not a showroom. People were turning up at the door expecting to look at products, and staff were spending their days sending away visitors who had driven across the county for nothing.
Nothing about those ads was badly written. The targeting was competent. Nobody had asked one basic question about how the business actually works.
The second one is more recent and more interesting, because AI caused it.
A restoration client kept getting inquiries for mold inspection. They do mold remediation. They are not licensed to do inspections, so every one of those calls was a dead end and a bad experience for the person calling. It took some digging to find the cause. Years earlier, a previous agency had published a blog post and a landing page about mold inspection. That page was still sitting on the site. Google Ads has been folding more AI into how campaigns find and use your content, and the system found that page, saw people responding to it, and did what it is built to do. It promoted it. Then promoted it harder.
The AI was not broken. It was working correctly. It optimized toward engagement on a page that should never have existed, because it had no way of knowing this company cannot legally perform the service that page describes. We traced it back, shut it down, and the wrong-fit calls stopped.
The trap nobody warns you about
People say AI cannot prioritize. That is not quite true. It does prioritize, and it is getting better at it every month. The problem is that its prioritization is only as good as the context it has, and it has far less of yours than you think.
What it is very good at is finding things. It will hand you a long list of genuine, defensible opportunities, all of which look worth doing. And because AI has made each of them fast to act on, the temptation is to act on all of them.
That is how a month disappears. Not through laziness, through speed. You end up with your attention split across eight half-finished efforts instead of two that would have moved the business. Deciding what to ignore has quietly become the most valuable thing a marketer does, and it is the one thing the tools are least equipped to help with, because ignoring something requires knowing what it would have cost you.
What experience brings that a tool cannot
It is easy to say AI lacks judgment and leave it there. It is worth being specific about what that actually means, because some of it is genuinely hard to replace.
A lot of it is pattern recognition built across many businesses. You see something that looks excellent on paper and you know it will not work here, because you watched a version of it not work somewhere else for reasons that never showed up in any dataset.
Some of it is the owner. Plenty of business owners have clear preferences about who they want as customers, and those preferences are legitimate even when they are not profit-maximizing on a spreadsheet. A tool optimizing for volume will happily deliver the customers you have spent years trying not to attract.
Some of it is operational. A campaign can work beautifully and still be the wrong campaign, because the type of customer it brings in cannot actually be served well by how the business is set up. That shows up as unhappy customers and staff, long before it shows up in a report.
And a great deal of it comes out of conversations. Something a client mentions in passing on a call, a frustration, a job they regret taking, a market they are quietly moving toward. None of that gets written down anywhere a tool can read. The person who was in that conversation carries it into every decision they make on the account afterward.
Finally there is the sense you develop for where things are heading, and which approach is going to suit a particular client before the numbers say so. That one is the hardest to defend and often the most valuable.
How we work with AI
I am transparent about this. We use AI at Digital Lotus for the repetitive work, the administrative work and a large share of research. It is the reason a small agency in Westlake Village can give a local business the depth of analysis that used to require a team.
What we do not do is let it decide what matters. Every recommendation gets filtered through what we know about your business, your customers and what you actually told us in the last conversation. The tools do the gathering. The thinking stays where it belongs.
The AMA report notes that most organizations are still stuck in experimentation mode, with individuals using tools ad hoc and no shared standard anywhere. That matches what I see. The gap is almost never the technology.
What to ask before you hire anyone
If you are looking at marketing help in Thousand Oaks, Newbury Park or anywhere in Ventura County right now, the old questions are not worth much. How many blog posts, how many ads, how many posts per month. Volume is the cheapest thing in marketing now, and paying by the unit is how you overpay.
Ask these instead:
- Who is making the judgment calls on my account, and can I talk to that person directly?
- What do you know about my business that would stop you from running the wrong campaign?
- Where exactly does AI do the work, and what does a human check before it goes live?
- How will you show me this produced revenue and not just activity?
The second one is the tell. An agency selling execution has never had a reason to find out.
Where this leaves us
AI did not make agencies obsolete. It made the production part of the work cheap, which means whatever you were paying for on top of production is now the entire value. For us that is context and judgment.
It has also taken us somewhere I did not expect. Clients started asking how we do what we do, and that turned into work of its own. We trained a sales team on using AI for lead research and prospecting, including how to build their own repeatable workflows so nobody was starting from a blank page every morning. For a couple of clients we built reporting dashboards that pull everything into one place, so they stopped assembling spreadsheets by hand every month and could share a live view with their own clients instead.
Our approach at Digital Lotus has always been partnership. Most of the time that means running your marketing, and that is still the work. But this is a strange and fast moving period for everyone, and when a client asks for help with something sitting next to their marketing, we are glad to share what we know. We did not set out to offer this. It came out of the relationships, and it has turned into a real and growing part of what we do.
Those were always the valuable parts of this work. AI just made it impossible to hide behind everything else.
Want a straight read on your own marketing?
If you are not sure whether your marketing is running on judgment or on autopilot, that is exactly the conversation to have. You would speak with me, not a rotating account manager.