How AI Can Help You Without Replacing You - Insights from Peter Lee of Microsoft Research

Key Concepts

• Generative AI (Gen AI): Artificial intelligence capable of creating new content, such as text, summaries, and ideas.
• Conversational AI Interaction: A methodology for using AI not by giving direct commands, but by engaging it in a dialogue or brainstorming session to refine and improve work.
• AI as a Sounding Board: The perspective of using AI as a collaborative partner to bounce ideas off of, rather than as a tool to simply generate final answers.
• Agentic Systems / Researcher Agents: Advanced AI systems (like those in Microsoft Copilot, ChatGPT, Gemini) capable of performing complex, multi-step tasks such as conducting research, analyzing information, and synthesizing findings.
• Content Refinement: The process of using AI to improve, shorten, or make existing, human-written content more effective and actionable.

---

Introduction: Dr. Peter Lee's Approach to Using AI

Dr. Peter Lee, head of Microsoft Research, discusses his personal and professional use of Generative AI. He emphasizes a philosophy of using AI as a collaborative partner and a "sounding board" rather than a replacement for human creativity and work. His approach focuses on engaging AI in a conversation to brainstorm, refine, and enhance his own output, providing actionable advice for professionals looking to integrate these tools into their workflow without being engineers.

Personal and Professional Use Cases for Generative AI

Dr. Lee outlines several specific ways he incorporates AI into his daily routine, ranging from personal health management to complex professional tasks.

1. Personal Health Management

• Application: Dr. Lee described using GPT-4 as early as two and a half years ago to analyze his complete blood test results.
• Specific Example: He would ask the AI to interpret the data and provide insights, such as whether he should be concerned about his elevated LDL cholesterol and what other metrics to pay attention to.
• Industry Trend: He notes that this capability is increasingly being integrated directly into health system portals and personal health applications, eliminating the need for users to manually copy and paste sensitive data into a separate AI tool like ChatGPT.

2. Managing Research and Information Overload

• Challenge: As the head of Microsoft Research, which spans 14 laboratories worldwide, Dr. Lee receives a tremendous volume of research reports and papers daily, making it impossible to keep up with everything.
• AI Solution: He uses AI to read and summarize these lengthy research papers. The AI helps by:
  • Providing concise summaries.
  • Suggesting specific sections or areas where he should "zoom in" for a deeper look.
  • Leveraging memory of his context and previous readings to provide smarter, more relevant insights over time.

3. Content Creation and Refinement: A Conversational Methodology Dr. Lee makes a crucial distinction: he does not use AI to write his emails, reports, or memos for him. Instead, he writes the content himself and then uses AI as a partner to improve it.

• Case Study 1: Improving National Academies Report Recommendations

  • Problem: While leading a report-writing effort, he felt the report was strong, but the final recommendations were not "crisp or actionable."
  • Process: Instead of asking the AI to "write better recommendations," he initiated a dialogue. He loaded the document into Microsoft Co-pilot and framed his prompt as a conversation: "I'm not satisfied with the recommendations... I don't feel that they're crisp enough, they're actionable enough. Can we have a conversation and brainstorm about how to improve this?"
  • Outcome: This brainstorming session helped generate new, more effective ideas. He even continued these conversations via his car's Bluetooth during his commute.

• Case Study 2: Shortening a Commencement Speech

  • Problem: He wrote a commencement speech and was later informed he had a strict time limit of under 15 minutes.
  • Process:
    1. He first asked Co-pilot: "How long would the speech take to deliver?" The AI estimated 19 minutes with a normal cadence.
    2. He then explained the 15-minute constraint and engaged the AI in a back-and-forth conversation to get suggestions on "how I might tighten this up."
    3. He selectively accepted and rejected the AI's ideas to refine the speech to the required length without losing his voice.
  • Key Insight: Dr. Lee highlights that AI is a "willing participant for nearly anything," making it an ideal partner for tedious tasks that a human colleague might not find engaging.

The Power of Advanced "Researcher Agents"

Dr. Lee discusses the remarkable capabilities of more advanced AI, which he refers to as "deep research" or "agentic systems."

• Definition: These are AI agents found in platforms like Microsoft Copilot or Gemini that can conduct in-depth research on a given topic.
• Real-World Application: Preparing for a Board of Directors Meeting
  • Task: Dr. Lee was preparing a speech for the board of a major insurance company.
  • Process: He gave his prepared talking points to a researcher agent and tasked it to: "do some research on the salient touch points and issues that this board of directors might be most interested in or might be really most top of mind for them."
  • Outcome: The agent returned a comprehensive, 15-page analysis. Instead of being overwhelmed, Dr. Lee used this document as the basis for a further conversation with the AI to distill the key points. This allowed him to refine his message and ensure his interaction with the board would be highly "pertinent" and "relevant."

Key Arguments and Perspectives

• AI as a Collaborator, Not a Ghostwriter: The central argument is to shift the user's mindset from viewing AI as a tool that produces final work to a partner that helps refine and improve human-created work. The host summarized this perfectly as "prompting as a to be a sounding board versus, hey, give me the answer."
• Maintaining Ownership and Quality: Dr. Lee states, "I feel better about the quality of the end if I've done this." By writing the initial draft himself and using AI for refinement, he maintains his personal voice, ownership, and a sense of meaning and purpose in his work.
• Overcoming a "Lacking in Imagination": Dr. Lee begins by stating, "I think we're all lacking in imagination" when it comes to AI's potential. His examples serve as a practical guide to expand that imagination beyond simple text generation.

Synthesis and Conclusion

Dr. Peter Lee provides a masterclass in leveraging Generative AI for high-level professional work. The main takeaway is the strategic shift from command-based prompting to conversational collaboration. By treating AI as a "sounding board" or a tireless brainstorming partner, users can enhance their productivity, refine their ideas, and produce higher-quality work without sacrificing their own voice or ownership. His specific examples—from analyzing medical data and summarizing research to sharpening report recommendations and tailoring a speech for a specific audience—offer a concrete framework for any professional looking to integrate AI into their workflow in a meaningful and effective way.