Week 23

*|MC_PREVIEW_TEXT|*

View this email in your browser Hi Humans, It is time for your weekly AI dose! A quick sum-up of today's newsletter: Klarna is saving millions using AI; I collected the best prompt hacks for you; and did you know that Microsoft will be your new sexy weather reporter? Weekly AI Wrap-Up: The fintech company Klarna, cuts agency expenses by 25% using AI. Here is a link to what you need to know about the EU AI act. Microsoft's new AI foundation predicts extreme weather events. The consultant company PwC is set to become OpenAI’s largest chatGPT enterprise customer. Klarna is saving $10 million annually using AI Klarna, the fintech company known for its “Buy now, Pay later” service, has leveraged AI to achieve significant marketing cost savings. Amounting to roughly 37% annually, or about $10 million. How they did it: By creating its marketing visuals using AI tools such as Firefly, Midjourney, and DALL-E., Klarna decreased its spending on external suppliers by 25%. Klarna's CEO, Sebastian Siemiatkowski, says the reduced reliance on external suppliers has allowed their marketing team to work more efficiently. Additionally, Klarna's CMO, David Sandstrom, mentioned that AI is integrated throughout their marketing processes, from ideation to creation. The company is now investing more in traditional media channels like TV, print, and digital video, aiming to maximize the impact of its AI-enhanced marketing spend. Why It Matters Klarna’s success demonstrates the potential of AI as a transformative tool in marketing. As AI technologies evolve, especially in video and graphic production, the efficiency, personalization, and cost-saving benefits will expand further, shaping the future of digital marketing strategies. This development is forcing people working within Marketing, content creation and graphic design to adapt to the new tools of AI. Which then again will create more and better AI tools… Become a Prompt God Prompt engineering is essential in AI, much like learning to use Google effectively when it first came out. It's all about knowing how to ask the right questions. If you search online, you'll find plenty of guides on how to prompt AI. But you deserve better than that! That is why I have spent hours of my very human time summarizing some of the best and easiest-to-use prompting practices shared by a top-notch team of two AI consultants, two AI team leaders, an AI educator, and an AI researcher. It goes like this: Single-Task Prompts: What It Means: Don’t ask the model to do too many things at once. Keep your questions or tasks simple and focused. Example: Instead of asking, "Can you summarize this meeting and also provide suggestions for improvement?" ask, "Can you summarize this meeting?" And then you can ask, “Now, please provide suggestions for improvement.” N-shot Prompts + In-Context Learning: What It Means: Show the model examples of what you want it to do. Example: If you want summaries, provide several different summaries first. Like showing a student a few math problems before asking them to solve one Chain-of-Thought (CoT) Prompting: What It Means: Encourage the model to think out loud and explain its steps before giving the final answer. Example: Instead of asking for the result, say, "First list the main points discussed in the meeting, then summarize them." Context Optimization: What It Means: Give the model only the most important and clear information it needs to complete its task. Example: If you’re summarizing a meeting, provide only relevant notes from that meeting, and don’t add random emails or unrelated documents. Why It Matters Being good at prompting in an AI world is as essential as being good at kicking a football for a football player! By being good at prompting, you maximize the performance of the large language models (LLMs) you work with and, therefore, the output you receive. Final advice: Try different formulations and prompt styles, and evaluate which gives you the output you like the most. Thank you for reading! If you want to contribute with articles, have inputs, or improvements - please let me know :) Until next time // Annemette Møhl This email was made in collaboration with Lukas Guldstøv. Follow Us Copyright (C) *|CURRENT_YEAR|* *|LIST:COMPANY|*. All rights reserved. *|IFNOT:ARCHIVE_PAGE|**|LIST:DESCRIPTION|**|END:IF|* Our mailing address is: *|IFNOT:ARCHIVE_PAGE|**|HTML:LIST_ADDRESS|**|END:IF|* Want to change how you receive these emails? You can update your preferences or unsubscribe