Your personal marketing assistant

Starting as an internal project of Gravitales – the design agency I co-founded in 2017 – our internal experiments with chatbots quickly found a use case - creating marketing briefings. Given the numerous dashboard-based tools used in marketing and content creation, we hypothesized that a chat interface could significantly reduce complexity. Thus, even non-tech-savvy users can easily access and create content for their marketing.

Content creation is becoming increasingly complex

Content for corporate communications is something special. It must entertain, inform, sell. That's why different disciplines have developed among professionals: Copywriters, bloggers, SEO writers, UX writers, press and PR copywriters, and content marketers.

Many business owners, marketers, or communications teams need (budget) or want (strategy) to create their own content. Due to dramatic changes in media consumption in recent years, the reality is usually like this: Today, you have to write a blog article, tomorrow an event invitation, and next Friday the press release is due. In the meantime, you need to fill in social media channels while you go about your day-to-day business. In addition to this consumption of a large amount of content in an increasingly short amount of time, the way many companies and organizations produce content has not adapted.

Every idea, creative concept, still starts on a blank sheet of paper. Therefore, it must be planned, conceived, researched, written, and optimized each time anew.

This is precisely where the chatbot SIGMUND TALKS came in.

A personal marketing assistant to the rescue

Our assistant had a chat interface. So, you could write short text messages with a computer program. With this, we covered the idea and concept phase. Instead of starting on a blank sheet of paper, you talked to the chatbot Sigmund and filled out your creative briefing. Based on the information, your strategy, and external data sources, we then generated a structured and partially pre-formulated blog article, for example. The result you could then finish and publish.

In my role as Chief Technology Officer, I was primarily responsible for the chatbot's product development and technology integration. Below you will find an excerpt about the main components of the product and what we have achieved with the chatbot.

Brand and chatbot personality

The brand name was created in reference to the founder of talk therapy, Sigmund Freud. As a chatbot, Sigmund's digital successor advised entrepreneurs, marketing teams, agencies, HR departments, or sales as a personal assistant to solve their (content) problems in conversation.

Research and testing have shown that when humans interact with machines - in this case, an intelligent chatbot - the issue of personification plays a central role. That's why we gave SIGMUND TALKS a chatbot persona, which uses an illustrative image style and personifies Sigmund as a likable character who can advise and support users with wit and charm. In addition to addressing users with a light "Viennese charm" and typical polite phrases such as "kiss my hand," Sigmund was also available for a joke or a quiz to encourage interaction playfully.

Product design

Users were guided through a "briefing conversation" with Sigmund through text input in natural language and presented with contextual and dynamic information. As a result, marketing briefings were generated for content creation.

In a second step, concepts with ready-made text modules were generated based on the briefings, which users could use for their marketing content. For example, blog articles were produced with a structure and the essential content elements, such as headlines, an introduction, or call-to-action.

Technology

Chatbots were one of the hype topics in 2018. There were isolated projects, frameworks, and platforms but little empirical data on which technologies were running stably or were particularly well suited for the chosen use case.

User Interface

The user interface was developed with React, styled-components, and Apollo Client and deployed via the cloud service Firebase.

Text input processing and chat conversations

The user input was processed using the Natural Language Understanding platform Google Dialogflow. This way, the user input was handled, and process flows within the chatbot were started with corresponding stored intents.

Universal GraphQL API

All server-side tasks (authentication, briefing, concept generation, user management…) were packaged into individual modules or microservices and bundled in a GraphQL API. This was developed using Apollo Server. MongoDB was applied as a database service. The backend was provided via Firebase and Google Cloud.

Export und Zapier integration

To further process the briefings and the generated concepts, users could either download the content or export it directly to any service of their choice via the Zapier web service.

Achievements and milestones

After some initial experimentation in early 2018, we decided to participate in the CONTENTshift competition in the summer of 2018. We soon gained traction in the press and on the user acquisition side. In October 2018, we were awarded as "Content Start-up of the Year" at the Frankfort book fair.

In early 2019 we closed a pre-seed round from the next media accelerator and continued our growth trajectory. We acquired more than 2,500 users, conducted more than 200 interviews with content creatives, content managers, and publishers to improve our product, and committed more than 1,000 times to our GitHub repositories.

In 2020 SIGMUND TALKS was superseded by its successor Storyliner. You can read more about it here.