Mediaboss Marketing

We had no idea what to expect when we launched MarTechBot in April. While factually correct, characterizing our experiment as the first generative AI chatbot designed specifically for marketing technology professionals seemed somewhat grandiose.

Was MarTechBot innovation? Would it be useful? Did anyone care? All we knew for certain was wed trained MarTechBot on MarTech.org content and wed learn a lot by putting the technology into your (and our) hands.

Since April, MarTechBot has engaged in over 600 conversations with marketers. Its answered questions, made recommendations and generated content based on your prompts. Based on the responses its created, MarTechBot has delivered value to the marketers whove used it.

To discern how youre using MarTechBot, we categorized each conversation in two ways.

First, we determined the purpose of each conversation. Was the conversation a question (e.g., What is a CDP?), a request for a recommendation (e.g., What is the best CDP?) or a generative/creation task (e.g., Write a project plan for implementing a CDP.)?

We then categorized each conversation by identifying a central theme. For example, general marketing conversations were tagged as marketing, while conversations specifically about CDPs were tagged as CDP.

This heuristic illustrates how MarTechBot has been used so far. In all cases, we used our judgment to classify each conversation manually (a labor of love!). This methodology is decidedly imperfect but useful.

Completing generative/creation tasks was the purpose of nearly half of all conversations with MarTechBot. This category includes brainstorming sessions, content planning and ideation of new strategies and frameworks.

By leveraging the creative potential of MarTechBot, marketers tapped into its vast knowledge base to generate fresh ideas and discover novel approaches to tackle marketing challenges. The bot was the catalyst for solving problems or generating suggestions about creativity.

Example conversations:

The second largest segment, constituting 43.5% of conversations, was marketers seeking answers to questions, much like theyd use a search engine. MarTechBot was used to clarify concepts, explain best practices and provide expert opinions.

Example conversations:

Recommendation-oriented conversations account for less than 10% of interactions with MarTechBot. In this category, marketers sought guidance, e.g., selecting the right marketing technologies, optimizing campaigns and the best ways to optimize customer experiences.

Example conversations:

Our analysis also revealed the themes that dominate conversations with MarTechBot. Here are the top five:

Rounding out the top themes 10 were email marketing, the catch-all category general, martech, data and AI.

We identified and classified ~50 themes, a diverse collection highlighting modern marketers wide range of opportunities and challenges. Beyond the top themes shown in the graphic, honorable mentions to Google Analytics/GA4, marketing automation and SEO.

The insights gained from analyzing real-world MarTechBot conversations provide valuable guidance to marketers looking to harness the power of this innovative tool and other generative AI platforms. By understanding the categories of recommendations, questions and generative/creation tasks, as well as the primary themes, marketers can leverage MarTechBot to gain insights, make informed decisions and foster creativity in their marketing endeavors.

As MarTechBot continues to evolve and address its limitations, it holds immense potential to transform how marketers strategize, execute and achieve their goals. The power of AI-driven chatbots like MarTechBot lies in their ability to provide personalized recommendations, unravel complex marketing dilemmas and inspire innovative ideas. By embracing this technology, marketers can enhance their capabilities, stay ahead of the curve and unlock new dimensions of success.

Were committed to improving MarTechBot by including more information in its language model.

For example, were addressing issues that result from training MarTechBot on long-form articles, video transcripts, or PDF files. In those cases, the bot has trouble parsing information and may sometimes generate unexpected or inaccurate responses.

In other cases, the bot provides answers that are incomplete or incorrect. We are actively working on refining these aspects and ensuring more accurate responses.

Try MarTechBot now!

Additional reporting and analysis by Karina Sarango.

Get MarTech! Daily. Free. In your inbox.

Read more here:
MarTechBot: Insights from real-world usage (so far) - MarTech