Self-organizing knowledge networks, redesigning work, the evolution of prompting, and more

📖In this issue

• Lesson: Five questions for solid data foundations for AI

• McKinsey on how AI is transforming strategy development

• Self-organizing knowledge networks help open-ended discovery

• AI can’t beat software freelancers… yet

• Helen Lee Kupp on redesigning work, enabling expression, creative constraints, and women defining AI

• The evolution of prompting

💡Lesson: Five questions for solid data foundations for AI

Data is the food that fuels AI. You cannot gain real value from AI in the enterprise without solid data foundations. One brief (5 min) lesson n my LinkedIn Learning course Building an AI Implementation Roadmap offers five questions to help you shape your data foundations.

The course is free for anyone with a premium LinkedIn account.

🧠🤖Humans + AI

McKinsey on how AI is transforming strategy development

A great article in Australian Financial Review shares details of how leading directors and CEOs are using genAI, including a chairman who used Claude to analyze and respond to a 60 page takeover offer before the market opening.

How AI is transforming strategy development
Summary of key insights

Self-organizing knowledge networks help open-ended discovery

“We present an agentic, autonomous graph expansion framework that iteratively structures and refines knowledge… to foster genuinely novel knowledge synthesis, yielding cross-domain ideas that transcend rote summarization and strengthen the framework's potential for open-ended scientific discovery. ”

Agentic Deep Graph Reasoning Yields Self-Organizing Knowledge Networks
Summary of key insights

AI can’t beat software freelancers… yet

“In a new paper, OpenAI researchers detail how they developed an LLM benchmark called SWE-Lancer to test how much foundation models can earn from real-life freelance software engineering tasks. The test found that, while the models can solve bugs, they can’t see why the bug exists and continue to make more mistakes.“

OpenAI’s study highlights limits of LLMs in software engineering

🎙️This week’s podcast episode

Why you should listen

Helen Lee Kupp has long been creating the future of work, earlier as head of strategy and analytics at Slack, now in applying AI and advanced capabilities with her initiative Women Defining AI. I love this conversation in particular for its focus on how we can create the conditions for people to express their fullest selves and potential at work.

💡Reflections

The evolution of prompting

Prompting is changing with more sophisticated models. This prompt structure shared by OpenAI President Greg Brockman offers useful guidance.

Prompt structure and clarity like this has always been useful - see for example my 2 year old prompt structure guide which offers a similar structure and is still highly relevant.

Return format is always useful. I often use an example markdown structure.

Specifying details, such as trail start, end, and distance, will guide good results. Though it can be good sometimes to ask it to give the details it thinks are relevant.

The inclusion of broad context is critical, something I’ve always emphasized, and which newer models are increasingly being able to apply.

The prompt provides step-by-step instructions, which guide the new ‘reasoning’ models more effectively.

One of the most interesting aspects is asking the model to check all findings. This could have been included with earlier models, but it was less likely to actually help. This can now significantly improve results.

What is missing from this single prompt is the follow up and iteration required to hone results in the direction you want. This has always been the single most important prompting attitude and aptitude..