Case study
A single operating system for an entire marketing function — research, strategy, creative, content, and delivery. Designed, built, and used daily by one person: me.
Real screens from the Marketing OS, v2. Each module below shows its own tool.
A working marketer rents an ad-creative tool, a scraping service, a scheduling app, a research assistant, and a report writer — then spends their day copying context between them. None of the tools share data, and every subscription solves a tenth of the job.
I built the alternative: one system where research feeds strategy, strategy feeds creative, and creative feeds a publish queue — with AI doing the repetitive work at every step.
Tom's Marketing OS · built by me
I don’t just use AI tools. I build them. The Marketing OS is a full production app I designed, coded and run by myself: every screen, every AI function, every integration. It’s the proof behind the systems I build for clients: custom assistants, automations, research tools and content pipelines that fit how a business already works.
Want one for your business? See AI systems →
What I built, by myself
Vite SPA with a config-driven sidebar — new tools are added as data, not code
Auth, Postgres with RLS, pgvector embeddings, and file storage
Deno functions that orchestrate every AI call, poll async jobs, and file results
Claude, GPT Image 2, Sora, KIE.AI, OpenAI, Perplexity, Firecrawl, Apify
Creative
The original core of the system: describe a product, upload photos and reference images, pick a style — and the OS generates three strategically distinct ad variations, each with its own hook and copy angle for A/B testing.
Research
Marketing decisions need data, so the OS has its own research infrastructure: general web scrapers and crawlers, plus dedicated scrapers for Instagram, Facebook, TikTok, X, YouTube, Reddit, and LinkedIn.
Strategy
Give it a product and a market, and a stepped AI engine researches the brand and competitors, then writes a complete marketing strategy: situation analysis, competitor analysis, buyer behaviour, SWOT, STP with perceptual maps, and a behavioural-economics lens.
Knowledge
I vectorised a library of marketing and strategy books — transcripts chunked, embedded with OpenAI, stored in pgvector — and built a retrieval bot over it.
Operations
A five-phase project framework — initiation through closure — where each phase knows which documents it requires. Drop any file in and an AI classifier reads it and files it into the right phase automatically.
You send one brief. Research, strategy and creative come out of the same system, so nothing gets lost between tools and you never explain your business twice.
Every ad, post and report starts from real research into your customers and a proven structure, then gets a hand edit. It reads like your brand, not a template.
Your colours, voice and selling angles are loaded once and reused, so the 30th piece is as on-brand as the first.
Because the heavy lifting is automated, you get more variations to test for the same budget, and learn what actually works.
Freelance project or a role you're hiring for — tell me what you need. I respond to all professional enquiries within 24 hours, usually much sooner.