By Stef Sy and Karen Tay

Me: “ChatGPT, help me improve this draft? Now try again. Give me a few more options.”

Also me: “AHHHH! Too many options, too much information, wrong tone! Wait, what am I trying to say, anyway?” slams computer shut and takes out a piece of paper to think.

Most people, like me (Karen) interact with Large Language Models (LLMs) this way, treating them like a slot machine – we drop in our coins (input), pull the lever, out comes a “better” output. Voila!

Let’s not deny that this approach brings huge productivity gains. But we also experience frustration: information overload without clarity, nice-sounding phrases lacking precision.

On the other side of the fence are technical people like me (Stef), who are experiencing something magical in how LLMs help my personal growth and productivity – but having a really hard time explaining this mental shift. “Oh no, what are you doing? Just telling it to “try again”? You have to relate to your LLM like a hyper intelligent staff who keeps growing…not a slot machine!”

The growing divide

At Davos, Chris Lehane from OpenAI referred to this divide as “the growing capability overhang… a widening divide between those who are able to use advanced AI tools deeply and productively, and those who aren’t.”

This gap impacts organizations, too. In our work with organizational leaders, we notice growing frustration with current paradigms of using LLMs. While work may “get done faster”, thinking might be less clear, details get missed and judgments are being escalated upwards.

When the workforce treats LLMs as a “slot machine” for productivity, it can unintentionally produce cognitive laziness and uniformity – traits which don’t serve an organization’s long term capability.

Developing a norm for using LLMs

In November last year, we (Stef and Karen), came to the topic from different places but landed on the same question: Is there another way? Can we develop a different norm in using LLMs – one which sharpens and strengthens humans?

As we developed this idea, we agreed on two ways LLMs should be used as a norm:

First, we should treat LLMs as long-term collaborators, not slot machines. If you think about your best working relationships, they were probably not one-way transactions. They were mutually sharpening, dynamic, and sometimes surprising in how they challenged you. The same should be for your relationship with LLMs.

Second, we should treat LLMs as amplifiers of our unique identity – our second brains. While the debate is ongoing, we do not believe there are globally correct judgment calls for most knowledge work, especially within organizations. Individual judgment, along with immediate context that no model can capture, will remain crucial. Humans should use LLMs to clarify and amplify their unique judgment, values and voice. Diversity, not uniformity, is the goal.

Getting to work on your AI Second Brain

Now, the question is: how do we build this new norm? And can we close the technical vs non-technical divide in the process?

Below, we cover three practical paradigm shifts that will sharpen and strengthen you as you use LLMs. These are your first steps towards building an “AI second brain”.

Don’t just query AI, get AI to interview you

First, imagine an AI second brain which makes your best decision-making and communication patterns transparent to you. It helps you clarify your unique voice, style and judgment as you use it.

How? This is the first paradigm shift: most of us “interview” AI for the knowledge we need. Instead, get AI to interview you for the knowledge it needs to do its job of extending your influence.

AI can be a dynamic interviewer, much like a skilled Chief of Staff or Executive Coach, who helps you extract your best work and decisions and identifies the patterns that make you the leader you are. We have out-of-the box “preference interviewer” prompts which you can feed to AI, to help it extract your decision-making, tone and voice.

Don’t just seek “improvements”, train AI to channel your unique identity

Second, imagine an AI second brain which channels your unique identity in its decision-making considerations and communications.

How? This is the second paradigm shift: most of us ask AI to help us “improve” generically, or intuitively. As AI is built on general standards of what “good” looks like, it will always fall short. Instead, train your AI system to be an extension of your unique identity, helping it get more and more precise at channeling your decision-making and voice.

One leader we know has eight “second brain*” AI systems for different domains: business strategy, sales and finance and operations – each with his specific guidance on how to problem-solve, which sources to go to, and how to handle sticky situations.

(*Eight specialized AI systems, each trained on domain-specific decision patterns.)

Don’t stick to one-off interactions, create a continuous feedback loop

Third, imagine a personal AI system which sharpens your thinking and meta-cognition each time you use it. It can observe you (with your permission), highlight blind-spots in your approaches, and give you feedback based on your growth goals… helping you improve as you use it over time.

How? This is the third paradigm shift: most of us use AI in one-off interactions. Instead, use it as an observer: you decide what you want to improve. You decide what AI observes such as your leadership meeting transcripts or personal communications. You decide how it pushes back on you, highlighting your blind-spots and nudging you towards improvement.

For example, one CEO wanted to improve her direct communication skills. With consent, she records her leadership meetings and pumps transcripts into AI. Her personal AI system gives her feedback on her improvements over time, giving nudges on how she could have re-phrased her comments. Feedback goes both ways: give your AI Second Brain feedback and opportunities to reflect on what it has learned – its precision improves as you tweak its context.

Getting started

Just one or two years ago, creating an AI Second Brain required coding knowledge. Today it is easily accessible to non-technical people – the only barrier is understanding, and of course, choice.

If this intrigues you, we have learned that the foundations of an AI Second Brain can be built within two-and-a-half hours for non-technical non-coders: not prompting tips, but an actual system for using AI to channel your unique identity, sharpen your thinking and close your blind spots.

Come design your second brain with us – join our waitlist here and we’ll notify you about upcoming sessions (both online and in-person)!

---

About the authors

Stef Sy is founder of Thinking Machines Data Science, a Philippines and Singapore consulting team that helps organizations design and build AI apps. Thinking Machines is an OpenAI Partner, helping organizations adopt and transform with GenAI.

Karen Tay is Founder of Inherent, a global coaching and learning consultancy, which helps leaders and organizations navigate transition with empathy and strategy. Drawing on research-backed methods and experience across Silicon Valley and Singapore, Inherent supports leaders through corporate programs, CEO advisory, and coaching for an AI-driven economy.