A new way for startups to compete and get their message out…using AI

Ai-led Pecha Kucha Explainer Videos

Theoremlabs.io is still a 24-month-old startup. We are a hybrid of management consulting and an AI lab for Fintechs, dedicated to learning applications of AI via controlled experimentation. We believe we have a unique model (expertise + talent (grey-n-green) + location) and a novel take on how best to apply AI to practical problems—not just for our clients but also for ourselves. Yes, we eat our own AI-dog food and find creative ways to apply AI to solve problems that we face.

Here is one that we feel can inspire others too…

Problem:

As a startup, our founding team is the best “sales & marketing” team to pitch and present our story and offerings. We don’t have and can’t yet afford dedicated marketing or content creators. Outsourced content creators are, of course, great to have, but they lack our business domain and market knowledge. Not to mention, it’s hard to translate our founding team’s passion about what we do to a third-party content creator.

Solution:

We invented AI-generated Pecha Kucha-style Videos -- A new scalable way for the founding team to blog or tell stories about their AI Accelerators & Experiments using new AI tools.

Here is a reference-One of our teams is building an AI-led methodology and utility tools to reverse engineer legacy cores for Banking, Insurance, and Healthcare. They built this video to explain the story of their experimentation which is now an Accelerator Theoremlabs support.

"What is Pecha Kucha?"

A Japanese storytelling format in which a presenter shows 20 slides for 20 seconds per slide. At a PechaKucha Night, individuals gather at a venue to share personal presentations about their work. The PechaKucha format can be used, for example, in business presentations to clients or staff, as well as in education settings."

Important: We are just lending the Pecha Kucha Style of speaking not to slides/pictures but pictures following your speak. We are not following its strict 20-sec rule and our variation is applied to multi-party speaking (unlike traditional Pecha Kucha).        

Theoremlabs's Workflow for generating AI-led Pecha Kucha Style Explainers:

Step 1: Create a 1-Pager for AI Experiment: The AI lead for the experiment and/or a core team engineer creates a 1-pager story about their AI Experiment. For example, one of our teams is building AI-led methodology and utility tools to reverse engineer legacy cores for Banking, Insurance, and Healthcare. They built this original content using any modern word editor.

Here is a the original version of what this team built.

Step 2: Generate Other Perspectives & Dimensions (Synthetically generated with Human in the Loop): Here we use Google’s latest Notebook LM to generate quizzes, summaries, FAQs, or timeline-style synthetic artifacts. Now, from our original 1-pager, we have 3-5 synthetic diverse perspectives in formats that bring the same information but in new, novel ways or make connections that were hidden in the original version.

Step 3: Generate a Deep Dive Podcast: The original authors (humans) review the synthetic content generated from Notebook LM and now use all or selective synthetic content, with or without modification for relevance & accuracy, as additional sources of knowledge for Notebook LM. Then they use the Podcast feature to generate the podcast.

Here is what we generated:

Step 4: Convert Podcast into Pecha Kucha Video: We take the podcast, extract the transcript from it (we used Descript.com), and use Microsoft Clipchamp to bring the podcast and assisting visuals together for our sales story. We also sometimes use Deep.ai to generate intro/outro music (post-edit of music in open source Audacity) as well as images for our Explainer Pecha Kucha's.

Here is our final version: Watch the video at the beginning of the section.