How I use AI agents to run a remote agency, and why I put one in every client package

There are two ways to talk about AI in a marketing agency. One is the pitch: AI will do everything faster and cheaper. The other is the daily reality: some tasks get much easier, some don't change at all, and a few get worse if you hand them over without care.

I use Claude every day to run growth and operations at Etlala, a remote agency serving Saudi e-commerce brands. This is how I use it, and why we decided to give every subscribed client their own AI agent.

Where AI agents help us internally

Research and first drafts

Market studies, competitor analyses and 90-day marketing plans all start with a lot of gathering and structuring. An agent can pull that into a first draft quickly. The judgment, what matters for this client in this market, stays with a person. I treat the draft as raw material, never as the deliverable.

Turning messy inputs into systems

A lot of operations work is converting a conversation into a structure: a meeting into a task list, a policy into a checklist, a pricing idea into a package table. Agents are very good at this kind of translation, and it saves hours each week.

Building and fixing the stack

I designed and deployed our operating stack largely on my own: self-hosted CRM, booking, client portal, website and integrations. Using an AI agent as a technical partner made that realistic for one person without a dedicated developer.

Scripts, briefs and training material

Sales scripts, ad scripts in Saudi dialect, onboarding documents for new team members. The agent drafts. The team edits for voice and accuracy.

Where we keep humans in charge

Client relationships. Account managers own the conversation. Clients pay for judgment and accountability, not for text.
Final creative decisions. AI can suggest. A person decides what represents the brand.
Numbers that go to clients. Every figure in a report is checked against the source.
Anything that sounds certain about the market. If a claim can't be verified, it doesn't go in.

An AI agent in every client package

The bigger decision was external. Our clients are store owners. Much of their week goes to small, repetitive work: updating products, checking what sold, pulling a quick report before a decision.

So we made it part of the offer: every store that subscribes gets an AI agent connected to the store, which manages products and reports through chat. The owner asks in plain language and gets an answer or a change, without opening a dashboard.

We chose not to sell it as a separate add-on, for three reasons:

It changes what the subscription means. The client isn't only buying posts and campaigns. They're getting a working tool that saves them time every week.
It keeps the packages simple. Fewer add-ons, fewer decisions at the sales stage.
It's hard to copy as a bundle. Other agencies can match a price. It's harder to match a system.

We followed the same logic with other zero-cost additions across the packages: Looker Studio dashboards, Microsoft Clarity for behaviour insights, and AI voice-over through ElevenLabs. Each adds visible value without adding much cost.

What I'd tell another operator

Start with a task you already do weekly. Frequent, repeated work is where agents pay off first.
Keep a human checkpoint before anything leaves the company.
Write down what worked as a process, so the gain belongs to the team and not to one person's prompts.
Look for the client-facing version. If AI saves your team time, there is probably a version that saves your clients time too, and that can become part of your offer.

AI hasn't made our team smaller. It has made a small, remote team able to run like a larger one, and it has given our clients a reason to stay that has nothing to do with the next campaign.