Keyword research. Content briefs. Competitive intelligence. Ranking decay alerts. Reporting. Dozens of repeatable marketing tasks that keep your team busy without making your business any better. I build AI agents that do that work for you — configured to your business, running on your stack, shipping output your team can actually use.
Most companies are being sold AI tools. What they actually need is AI operators — someone to figure out where agents make sense in their marketing function, build them, and keep them running.
That's the job. I draw on a library of 60+ pre-built marketing agents, configure the ones that fit your business, and build custom ones where nothing off-the-shelf works. You end up with a small number of agents doing specific, useful work. Not a dashboard of demos.
Every agent can be tuned to your business, your ICP, your tools, and your decision logic. Below are the categories, with a few named examples in each. Most clients use 3–6 agents — we scope which ones during setup.
No off-the-shelf dashboard. No generic automations. The agents are built around how your business actually works — and they live inside the tools your team already uses.
Agents only work when the inputs are right. Here's what each side is on the hook for.
Zapier connects tools and triggers actions. Agents make decisions. The difference matters: a Zap sends the same email every time a CRM field updates; a competitive-intel agent reads a competitor's blog post, decides whether it's worth flagging, writes the summary, and drops it in Slack only if it meets your criteria.
Zapier is plumbing. Agents are the thinking layer sitting on top of the plumbing. Often I'll use both — a Zap to handle the trigger, an agent to handle the judgment.
Everything gets documented in your drive from day one — the configuration, the prompts, the integrations, the review steps. If I disappear tomorrow, another competent operator can pick it up. The agents live on your stack, not mine.
On models breaking: this is an ongoing ops cost, not a surprise. Providers shift, capabilities change, prompts drift. Part of the monthly fee is me watching that, retuning when something degrades, and swapping underlying models when something better ships.
Before we build anything, we define what "working" looks like for each agent — hours saved, briefs shipped, alerts that actually triggered the right action. That becomes the review metric.
Monthly, you get a one-page report: what ran, what it produced, what it cost in time and dollars, what changed. If an agent isn't earning its keep, we retire it. No agent runs indefinitely just because it was built once.
Each agent has a one-off build fee and a monthly operating fee. No packages, no tiers, no "unlimited agent" plans. You pay for what you run.
Discovery, configuration (or custom build), integration into your stack, and sign-off. Pre-built agents from the 60-agent library come in at the lower end. Custom-built agents, or ones that need complex integrations, scale from there.
Most clients start with 3–6 agents.
Monitoring, maintenance, retuning, model upgrades, monthly performance review. If an agent stops earning its keep, we retire it — you stop paying. No minimum-agent-count fees.
Three-month minimum. 30 days' notice to cancel after that.
Pricing shown is a starting point, not a fixed quote. Final numbers depend on agent complexity, integration scope, and whether custom builds are needed. I'll quote properly once we've talked.
One conversation. We'll look at your marketing function, identify the 3 or 4 agents that would save the most time, and figure out whether this is worth building — or whether one of my other services is closer to what you need.