What Can You Do with Generative AI: Insights from the past Year

It's been almost 12 months since I completed the executive program "Harnessing AI" at Stanford. This experience was primarily about understanding how AI could impact our business and that of our clients.

My key takeaway? This technology will be bigger and faster than I anticipated before enrolling. Thanks to the Stanford staff and my fellow participants, I gained the right perspective to put my newly acquired knowledge into context.

AI: The Biggest Game-Changer in Decades

AI, specifically generative AI, will be the most significant game-changer in decades. At Q&A Retail, all four of our pillars - Insights, Consulting, Education, and Inspiration - will experience changes due to gen AI, each to a different extent.

We've invested heavily in time, services, applications, and experimentation over the past year. We've also been developing our service proposition for clients. How did we start? With a gen AI taskforce, experimenting with different models, searching for solutions, and understanding how gen AI can be used in each pillar.

Getting Started with Gen AI

We began by discussing internally what AI usage would mean for us. We experimented with Large Language Models (LLMs) like ChatGPT, Claude, and Gemini, as well as Microsoft Co-pilot.

Key Concepts We Introduced:

1. LLMs as Interns: We introduced the concept of treating LLMs like interns. This metaphor helped us greatly in understanding how to work with AI tools responsibly. Just as you wouldn't hand over highly confidential information to an intern, we realized the importance of being cautious with sensitive data in AI interactions. The intern analogy also helped us understand that AI outputs need to be checked, just as an intern's work would be reviewed. We recognized that AI, like interns, might make mistakes or occasionally "hallucinate" (create fictional information). Importantly, we established that AI-generated work, like an intern's, should never be sent directly to a client without review and refinement.
2. Persistence in Trying: We encouraged our colleagues to keep trying even when initial results were disappointing. Some team members argued that AI wasn't working effectively, but the more optimistic among us had to give kudos to keep others motivated. We reminded everyone to "think of the intern" - AI needs to learn, requires training, and improves over time. Constructs that didn't work initially became almost perfect with persistence. We also noted that while we might not accept 60% or 80% accuracy from AI initially, we often accepted 85% to 90% from human colleagues. The key difference? AI could produce results in minutes, whereas human work took days.
3. Man-Machine Balance: With every step in our AI journey, we constantly evaluate where and how to use AI, considering cost, time, and performance perspectives. We recognize that while much can be automated, it doesn't always make sense to do so. We keep in mind that each aspect of this balance can change over time as technology evolves and our needs shift.
4. RTF Model: When we started, we were initially overwhelmed by AI results, but our satisfaction quickly diminished. We realized our expectations had changed, and our prompting skills weren't improving. To address this, we introduced the Role-Task-Format (RTF) model for effective prompting. This involves: Role: Setting the scene by describing the AI's role Task: Clearly explaining the task and providing context Format: Specifying how you want the output formatted This structured approach, combined with improvements in LLMs, substantially enhanced our AI interactions.

Gen AI in Our Practice

Insights

We use gen AI for various types of research, improving questionnaires, analyzing open questions, and assisting with desk research. While we haven't seen immediate cost savings, our output quality has significantly improved.

Consultancy

We use AI tools to access and analyze client data, our research, and online resources securely. This process, which would typically take days, now takes hours. We've learned that it's not just about providing access to data; the key is in proper modeling, data selection, and decision-making - much like following a good recipe in cooking.

Inspiration

We use LLMs for improving texts, getting ideas for programs, creating presentation concepts, and developing podcast setups. It helps us create concept PowerPoint presentations and even assists in setting up our podcasts.

Education

AI assists in creating new learning programs, developing cases and assignments, and even producing explainer videos in multiple languages. We've explored translating and lip-syncing existing Dutch videos into English and other languages, with impressive results.

AI Services Offered

After using gen AI ourselves and giving several presentations, we wanted to take the next step. I reconnected with my old study friend from Groningen, Maarten Mantje. We share the same fascination for this technology and are convinced that companies can save time, money, and effort by properly using gen AI.

We've developed the following services to help businesses harness the power of AI:

1. AI Design Sprint: This service provides great insight into where gen AI can be used in your organization. We work together with your team to identify where to use AI and for what tasks, up to the first proof of concept (POC). This sprint helps companies understand the potential of AI in their specific context and lays the groundwork for successful implementation.
2. Workshops for Marketers: These specialized workshops focus on how marketers can leverage AI tools to enhance their strategies, improve content creation, and streamline their workflows. We cover topics like AI-assisted market research, personalized content generation, and predictive analytics for marketing campaigns.
3. ChatGPT for Daily Use Workshop: This practical workshop teaches participants how to effectively use ChatGPT in their day-to-day work. We cover prompt engineering, best practices for interacting with the AI, and how to integrate ChatGPT into various business processes safely and efficiently.

A recent Microsoft study revealed that 75% of employees use gen AI regularly, with 85% using it for business purposes, often without their employer's knowledge or approval. This underscores the urgent need for companies to develop comprehensive AI policies and implementation strategies.

The question is no longer if you will be using AI within your company, but how you will be using it. Your co-workers are likely already using it every day; the only question is whether they're using it in the way you want them to. Our services help you find out where to use AI, how to use it, and set priorities for usage with your team. This ensures that you stay ahead of the curve while using AI in a safe manner.

Can You Make a Fully AI-Generated Podcast?

As an experiment, Maarten and I created a concept for an AI-generated podcast. We used ChatGPT to script conversations on various topics, created avatars, and even explored multilingual possibilities. We call it AMPC - AI Made PodCast. Click on the link to check it out.

AMPC Episode 1

While it still needs fine-tuning, it's an exciting start. If you'd like to see more, let us know your desired topics in the comments.

Start Experimenting

Start experimenting with gen AI yourself. If you need help kickstarting this process in a safe, secure, and fun way, don't hesitate to get in touch with Maarten or me. The AI journey is just beginning, and the possibilities are endless. Remember not embracing the power of AI is most definitely a recipe for competitive disadvantage.