How we actually use AI,
in practice.
Most agencies use AI and hide it. We use it and own it. Here's why, and here's how.
Why we own it
In 2026, pretending not to use AI is a waste of time for everyone. Your prospects know agencies use it. Your competitors use it. And the real quality of a deliverable no longer depends on how many interns worked on it. It depends on the quality of the production method. We'd rather explain how we work openly, than play selective transparency.
Where AI sits in the engine
We use AI at four main stages: research and structuring at the start of a project, generation of first versions of deliverables (copy, visuals, code), quality control (proofreading, inconsistency detection, standards checking), and automation of repetitive tasks (formatting, exports, notifications). At every stage, a human validates and decides. AI never ships directly to a client.
Where AI isn't
Three places we don't use AI: strategic decisions (positioning, tone, angle), client communication (you talk to us, not to an agent), and final QC before delivery (a human reads every word, checks every pixel). We never ship anything we haven't personally reviewed.
What this means for your project
Concretely, you get three benefits. First: short timelines. Where a traditional agency delivers a first draft in two weeks, we deliver in three days. Second: a price that reflects our real production cost, not the cost of an agency with ten salaries to pay. Third: finish quality equivalent to a traditional agency, because finish work stays 100% human.
Our tooling
For full transparency: we mainly use Claude (Anthropic) for copy and analysis, GPT (OpenAI) for certain technical tasks, Google's generative models for visual inspiration (never as final deliverables), n8n for workflow orchestration, and Notion as the project control base. All these tools are configured not to reuse client data in model training.