Available for consulting

I build marketing systems that run themselves.

Not the strategy for them. Not the tooling to support them. The systems that actually do the work, running every day without anyone driving them.

30+
Scheduled AI agents running in production, across three runtimes
7,000
Newsletter subscribers on a publication written and published by agents
5,460
Competitor ads analysed for one original research index
3
Businesses built and operated end to end, all with the same playbook

Four systems, built and running.

Every one of these was a decision before it was a build. They run live today across businesses I own and operate.

Automation

A fleet of agents that doesn't rot

More than 30 scheduled agents doing research, drafting, publishing, monitoring and reporting across three separate runtimes, reporting into one dashboard rebuilt twice a day. The hard part was never spawning agents. It was building the check that proves they still do what the documentation claims.

  • Claude Code
  • launchd
  • Python
  • PostHog
  • Slack
Read the case study →
AI search & AEO

One dataset, four surfaces

A free competitor ad tool, programmatic brand and vertical pages, a citable research index covering 91 UK brands and 5,460 ads, and an MCP server so AI assistants can query the data at answer time. The same dataset earns links, captures emails, and puts the brand inside AI answers.

  • Meta Ad Library API
  • MCP
  • Vercel
  • JSON-LD
  • Gemini
Read the case study →
Content systems

A publication run by agents

A marketing and technology publication with 7,000 subscribers on a DR 63 domain, staffed by one writer and one assistant. Agents handle research, drafting, hero images, publishing, SEO and cross-site linking, working from a maintained knowledge base so the output carries a consistent point of view.

  • Beehiiv
  • Webflow API
  • Search Console
  • Schema
Read the case study →
Speed

Zero to a live portal in a day

A property portal taken from decision to a live, fully templated site inside one day. Four CMS collections, every page template, structured data verified in the response body, all built against the API rather than by hand. Agency supply is won by automated outbound running under a written autonomy ladder.

  • Webflow MCP
  • AgentMail
  • JSON-LD
  • Autonomy ladder
Read the case study →

Supervised, and getting better on their own.

The two questions everyone asks about handing work to agents. Both are answered in how the systems are built, not in a policy document.

Under supervision

Agents propose. People approve.

Every system runs against a written autonomy ladder that says exactly what an agent may do on its own, what it must draft for review, and what it must never touch. Routine work runs unattended. Anything that spends money, sends to a customer, or changes a live page waits for a human. You get the leverage without handing over the keys.

Self improving

Every run makes the next one better.

When a system hits an error, needs a clarification it should not have needed, or solves something the hard way, the fix is written back into its own instructions before the job closes. The same problem does not happen twice. The systems get more capable while you are not looking, which is the opposite of how most automation ages.


Five things, all the same idea.

Look at a manual process, see a system, then build it so it runs without you.

01

AI agent systems

Fleets of scheduled agents that do the recurring work: research, drafting, publishing, reporting, monitoring. Including the unglamorous half, which is making sure they keep working after everyone has stopped watching them.

02

AI search and AEO

Visibility tracking across ChatGPT, Claude, Perplexity and Google AI Mode. Structured data architecture. MCP endpoints so AI assistants can call your data directly when they answer a question about your category.

03

Marketing data and APIs

Meta Ads and Ad Library, Search Console, product analytics, billing. Pipelines that end in one dashboard your team actually reads instead of six tabs nobody opens.

04

Content and SEO systems

Agent-run publishing pipelines, programmatic page builds, internal linking, structured data and author schema. Content operations that scale with compute rather than headcount.

05

Email infrastructure

Deliverability, inbox warmup, lifecycle automation and cold outbound that lands. The boring layer that decides whether any of the clever work above ever reaches a person.


What I build with.

Chosen for what they do, not for what they cost. Most of this runs on free tiers or something already being paid for.

Agents and orchestration

Claude CodeClaude Agent SDKHermesMCPlaunchdMakePlaywright

Build and hosting

VercelNext.jsWebflowTailwindPythonGitHub

Marketing data

Meta Ads APIMeta Ad Library APISearch ConsoleGA4PostHogClarityScrapling

Email and lifecycle

AgentMailLoopsBeehiiv

Operations

SlackClickUpClerkPolarNotionGranola

Three shapes.

Shape one

Systems audit

Short and fixed scope. I map what is manual, what should be automated, and what it would take. You keep the written plan whether or not you hire me for the build.

Shape two

Build sprint

One system, built and live. The agent fleet, the AEO layer, the outbound machine, the data pipeline. Defined start, defined end, working software at the end of it.

Shape three

Ongoing

Retained. I operate and extend what has been built, and add the next system when the current one is stable. For teams who want the machine to keep growing.

What are you still doing by hand?

If you can describe the process, it can probably run itself. Tell me what it is and I will tell you whether it is worth automating.

kole@tripledouble.marketing