Almost every agency says it uses AI now, so the claim alone tells you very little. What matters is where it's used, who checks it, and whether it brings in customers or just more content. Here's exactly where AI does the work in our client systems, where a person still decides, and the questions worth asking any agency that says it uses AI.
- AI drafts, sorts and summarizes. A senior marketer sets each system up, checks the output, and decides what happens next.
- Mostly Claude, from Anthropic, working inside the email platform, CRM and checkout you already pay for.
- The ad platforms' own AI does the bidding. Our job is giving it the right signal: tracking that reports real sales.
- Nothing reaches a customer unread, and AI doesn't decide budgets, offers or targeting.
- Everything is built in your accounts, so it keeps working if you stop working with us.
Where AI does the work, and who checks it
| System | What AI does | What a person does |
|---|---|---|
| Lead follow-up | Drafts the follow-up emails from your FAQs and your voice guide | Approves every email before the sequence goes live |
| Customer emails and content | Drafts welcome, after-purchase and win-back emails, posts and ad copy in your voice | Reads every draft; you approve the first versions |
| Outreach replies | Labels the replies that look human as interested or not | Answers them, and checks the labels |
| Weekly report | Summarizes what moved in the account numbers | Checks the figures and adds the next action |
| Ad bidding | Meta's and Google's own AI bids inside each campaign | Sets the objective, budget, offer and tests |
Lead follow-up: the speed is automation, the words are AI
A lead from a Facebook form is worth most in the first minutes. That's why we built Nurture, our own lead follow-up app. It's plain automation, not AI: the moment someone submits a Meta lead form, it routes them by their answers and sends the first email. Across its first 52 live leads, the median time from form to first email was 10 seconds, and 30 of the 52 (58%) arrived outside weekday office hours, when nobody was at a desk.
AI's part comes earlier. It drafts the sequence those leads receive, from the business's own FAQs and a voice guide built from its writing. A person edits the drafts, the client approves them, and only then does anything send. Speed without the right words just gets ignored faster. For coaches, how fast to follow up a coaching lead covers the timing in detail.
Emails and content in your voice
Most founders have tried ChatGPT or Claude for an email or two, and the drafts read like AI. The fix isn't a better prompt. It's groundwork: a short interview, plus a read of your best emails, posts and sales pages, turned into a voice guide that every draft follows. After that, AI turns a 30-minute conversation into a month of posts, a newsletter and new ad copy to test, and a person reads every piece before it's published.
We don't use AI to invent things. No client names, no reviews and no figures that aren't on the record. If a page needs proof we don't have yet, it waits.
Outreach: rules first, a model second
Our cold email work runs on our own outbound platform. Replies arrive mixed in with bounces, out-of-office notices and receipts. Plain rules sort those first, at no cost. A model reads only the messages that look like a real person replying, and labels them interested or not, so the real replies get answered first. Paying a model to read an out-of-office message is waste.
Some of the most important work there isn't AI at all. Every address is checked against its mail server before it's sent to: the platform has run 551,802 conclusive checks, and it holds back anything it can't confirm. That's just careful engineering, and we'd rather say so than call it AI.
Ads: the platforms' AI bids, we feed it the signal
Meta and Google already use AI to bid, find audiences and choose placements, for every advertiser. An agency can't out-bid that AI by hand. What it can do is give it the right signal. A campaign told to find clicks finds clickers, so the objective has to match the result you want, as link clicks vs a sales objective explains. And it can only learn from the sales it can see, which is why every account we take on gets the Pixel and the Conversions API checked first.
Budgets, offers, objectives and what to test next stay with a senior marketer. Those are business decisions, and they need someone who can be asked why.
Where we don't use AI
- Sending anything to a customer that nobody has read.
- Deciding your budget, your offer or your targeting.
- Writing reviews, testimonials or results. Ours are real or they aren't there.
- Handling customer data a task doesn't need.
- Building on accounts we own. Your systems live in your accounts, so they keep working without us.
Questions to ask any agency that says it uses AI
Where exactly is it used?
Ask for the specific systems: follow-up, emails, reporting. "We use AI across everything" usually means a chat subscription.
Who reads the output before a customer sees it?
A named person, every time, or you'll be the one rewriting drafts that sound like a robot.
Whose accounts is it built in?
If the automation lives in the agency's accounts, you lose it when you leave.
What data goes into the AI tools?
Only what each task needs, set out in writing before the work starts.
Can you show me one running?
A live system, a test lead or a sample report tells you more than a slide about AI.
What it costs
Our AI marketing automation setups are a fixed price, agreed before we start: from $1,200, with most running $1,200-$4,000, and optional monthly running from $500/mo. Software subscriptions and ad spend are separate. How much an AI marketing agency costs compares that with market rates, freelancers and doing it yourself. The systems themselves are on our AI marketing automation page, and our work on showing up in ChatGPT and Google's AI answers is AI visibility.
Questions people ask
How does NCMborz use AI?
For drafting, sorting and summarizing: follow-up and customer emails written from a voice guide, replies to outreach sorted so people see the real ones first, and weekly reports summarized from the account numbers. A senior marketer sets each system up and checks what it produces before anything reaches a customer.
Which AI tools do you use?
Mostly Claude, from Anthropic, for writing, sorting and reporting, working alongside the platforms you already pay for. The ad platforms' own AI does the bidding: Meta's Advantage+ and Google's Smart Bidding. Nothing we build depends on an account in our name.
Does AI write the emails your clients' customers receive?
It drafts them, from a voice guide built from the client's own writing. A person reads every draft, the client approves the first versions, and nothing sends until both have happened.
Does AI decide where the ad budget goes?
No. Meta and Google bid automatically inside each campaign, as they do for every advertiser. Decisions about the budget, the offer, the objective and what to test next are made by a senior marketer.
Is AI marketing cheaper than a regular agency?
Sometimes, because systems that run on their own take fewer hours each month. Our automation setups start at $1,200, most run $1,200-$4,000, and optional monthly running starts at $500/mo. What an AI marketing agency costs elsewhere is in our pricing guide.
What happens to my customer data?
It stays in your own accounts wherever possible, and AI tools get only what a task needs. We set out in writing how your data is handled before we start.
When is this not a good fit?
When the business hasn't made its first sales yet, because there's no customer list to email and no proven offer to automate. We're also not the right fit if you want content volume for its own sake.
Want to see which of your systems AI could take off your plate? Book a strategy call and we'll go through where your week and your enquiries go.
