AI-powered chatbot for marketing briefings

SIGMUND TALKS started as an experiment at Gravitales, the design agency I co-founded. We were building chatbots for clients and realized the same approach could solve a problem we knew firsthand: the blank-page problem in marketing. Instead of filling out forms or starting from scratch, you'd have a conversation with a chatbot named Sigmund – and walk away with a structured creative brief. The idea turned into a product, the product won "Content Start-up of the Year" at CONTENTshift, and we grew it to over 2,500 users.

The problem

Content creation in marketing is fragmented. Blog posts, press releases, event invitations, social media – every piece starts on a blank page. It has to be planned, researched, written, and optimized from scratch. For small teams and business owners doing their own marketing, the workload is relentless and the tools are either too complex or too rigid.

What we built

Sigmund was a chatbot with a chat interface – simple text messages, natural language. Instead of staring at a blank document, you had a conversation. Sigmund asked the right questions, gathered your strategy and context, and generated a structured marketing brief. From there, the system produced partially pre-formulated content – a blog article with headlines, structure, and key elements – that you could finish and publish.

Sigmund's personality

The name was a nod to Sigmund Freud – the founder of talk therapy. We learned early that people engage differently when a chatbot has a personality, so we gave Sigmund one: a likable illustrated character with a light Viennese charm, polite phrases like "kiss my hand," and the occasional joke or quiz to keep the conversation playful. It sounds like a gimmick, but it changed how people interacted with the product – they talked to Sigmund instead of trying to game a form.

How I built it

As CTO, I was responsible for product development and the entire technology stack. Chatbots were a hype topic in 2018 – lots of frameworks, little proven ground. I chose Google Dialogflow for natural language processing because it handled intent recognition well enough for structured conversations without requiring us to train our own models. The backend was a GraphQL API built with Apollo Server, with each capability – authentication, briefing logic, concept generation, user management – packaged as its own microservice. The frontend ran on React, deployed via Firebase. Users could export their briefings directly or push them to any tool via Zapier.

The result

We went from internal experiment to award-winning product in under a year. In summer 2018 we entered the CONTENTshift competition; by October we were named "Content Start-up of the Year" at the Frankfurt Book Fair. In early 2019 we closed a pre-seed round from the next media accelerator. Along the way: 2,500+ users, 200+ interviews with content creators and publishers, and 1,000+ commits to our repositories.

Get in touch if you're planning to build something similar.