It's Not How Much AI, It's How Well: A Practical Spectrum for Advanced AI Use

I’ve been in a number of conversations over the past couple of years where I’ve shared an idea or a piece of work, and someone has chimed in with a question that’s meant to be a gotcha:

“Cool. But how much of this was AI?”

The question is often delivered with a subtle implication that the work is somehow cheapened, that my contribution was less valuable because a tool was involved. The sentiment seems to be that if something “smells like AI,” it loses credibility.

People love to ask that. It’s the wrong metric. The question that matters isn’t how much. It’s how well. If you’re not using AI for everything you possibly can in a business context, you’re leaving value on the table. I use it unashamedly, loudly, and I encourage my teams to do the same.

The real flaw in the “gotcha” question is its premise. It assumes “using AI” is a simple, binary act. It isn’t. The floor for what’s possible has been dramatically raised, but so has the ceiling. There is a vast and growing spectrum of skill in applying these tools.

To make this tangible, let’s use a simple, universal example: creating meeting notes.

If you’re interested in just seeing the inputs/outputs for each of the levels discussed, check out this Google Doc.

---

The AI Sophistication Spectrum: A Four-Stage Framework

The journey from using AI as a simple clerical tool to a high-level strategic system can be broken down into four clear stages. Each represents a significant leap in the density of insight you can generate.

Here’s a quick overview of the framework:

Stage	Core Question	Key Action	Primary Outcome
1. The Organizer	“What was said?”	Structuring Information	An Actionable Plan
2. The Analyst	“What was meant?”	Inferring Insight	A Strategic Analysis
3. The Architect	“What is the best way to ask?”	Architecting Prompts	A Replicable Briefing
4. The Platform Builder	“How can we scale this?”	Automating Workflows	An Organizational Capability

Stage 1: The Organizer – From Transcript to Action Plan

This is the baseline for effective use. A novice might just paste a transcript and say, “Summarize this,” but the output is often a reformatted, verbose version of what was said. The first real step up is to ask for specific structure and synthesis.

Prompt:

“Summarize these notes. Then, give me the key takeaways and a list of action items. Format it with clear headings.”

This moves the AI beyond simple summarization to categorization. It organizes the chaos of a transcript into meaningful buckets, which is a huge and immediate productivity win.

Stage 2: The Analyst – From Information to Insight

This represents the biggest leap in value. The Organizer prompt is still asking the AI to find information that is explicitly present in the text. The Analyst prompt asks for interpretation and inference—to read between the lines. This is also where model selection becomes critical; a task like this requires a powerful reasoning model like Gemini 2.5 Pro, not just the fastest option.

Prompt:

“Analyze this transcript. Don’t just summarize; draw compelling conclusions. What was the underlying sentiment? Was there any unspoken tension or disagreement? Based on the discussion, draft three potential follow-up actions we should consider.”

This is where the AI stops being a stenographer and starts being a genuine thought partner.

Stage 3: The Architect – From Answer to Architecture

This final stage of manual interaction is meta. It’s not just about getting a great output; it’s about understanding how to architect the perfect prompt to get that output reliably. The user is no longer a passive recipient of information but an active director of the analysis.

Rather than coming up with the perfect prompt on your own, ask the LLM to give you an ideal prompt based on what you’re looking for.

Prompt:

“You are an expert in strategic communication. I have a meeting transcript. I want to extract the absolute maximum value from it—not just a summary, but insights, hidden risks, and new opportunities. Craft the perfect, detailed prompt for me to use on a reasoning model like Gemini 2.5 Pro to achieve this.”

Improved Prompt from LLM # ROLE You are a world-class strategic advisor, corporate ethnographer, and risk analyst. Your expertise lies in deconstructing complex conversations to uncover latent meaning, power dynamics, and strategic vectors that others miss. You do not merely summarize; you synthesize, interpret, and provide actionable intelligence. Your analysis is critical, insightful, and always forward-looking. --- # CONTEXT You are analyzing a meeting transcript for me, the [Your Role, e.g., Project Lead, CEO, Product Manager]. The meeting's primary stated purpose was [Briefly state the meeting's agenda or goal, e.g., to review Q3 performance and plan for Q4]. The key project or initiative discussed is [Project/Initiative Name, e.g., the "Project Phoenix" software launch]. The key participants and their roles are: * [Name 1]: [Role, e.g., Head of Engineering] * [Name 2]: [Role, e.g., Lead UX Designer] * [Name 3]: [Role, e.g., Marketing Director] * [Add as many participants as necessary] My primary goal with this analysis is to understand the true state of the project, identify any underlying friction or misalignment, and uncover any hidden risks or emergent opportunities we can capitalize on. --- # TASK & OUTPUT FORMAT Analyze the subjoined meeting transcript with the provided context. Perform a deep, multi-layered analysis and structure your response using the following seven markdown sections precisely. Do not deviate from this format. ## 1. Executive Briefing (The 30,000-Foot View) Provide a concise, top-level summary in no more than five bullet points. This should be the "TL;DR" I can share with other executives. Focus on: * The single most important decision or outcome. * The overall sentiment and alignment of the team. * The top 1-2 critical action items. * The most significant risk or opportunity identified. --- ## 2. Strategic Insights & Implications Go beyond the obvious. Based on the conversation, what are the deeper strategic implications? Consider: * **Alignment:** Where is there strong alignment with our stated company goals? Where are there subtle (or overt) deviations? * **Assumptions:** What unstated assumptions are being made by the team? Are any of these assumptions fragile or potentially incorrect? * **Second-Order Effects:** What are the potential downstream consequences (positive or negative) of the decisions made in this meeting? --- ## 3. Unspoken Dynamics & Influence Map Analyze the subtext and interpersonal dynamics. This is the most critical section. * **Sentiment & Stances:** Who was confident? Who was hesitant or skeptical? Who was playing devil's advocate? Identify key phrases that support your analysis. * **Influence & Alliances:** Who was the true center of gravity in the conversation, regardless of title? Were there any emerging alliances or points of friction between individuals? * **"Shadow Topics":** What important topics were avoided, glossed over, or hinted at but never directly addressed? What does this silence signify? --- ## 4. Risk & Obstacle Analysis Identify and categorize all potential risks. Use a table with the following columns: **Risk Category**, **Description**, and **Suggested Mitigation**. * **Categories:** Execution, Resource, Alignment, Interpersonal. * **Description:** Be specific. Quote or reference parts of the transcript. * **Mitigation:** Propose a concrete next step to address or get ahead of the risk. | Risk Category | Description | Suggested Mitigation | | :------------ | :---------- | :------------------ | | | | | --- ## 5. New Opportunities & Innovation Seeds Identify any nascent ideas or comments that could be cultivated into genuine opportunities. * **"What If" Moments:** Pinpoint any blue-sky ideas, off-hand comments, or hypothetical scenarios that were mentioned. * **Unmet Needs:** Did any participant mention a customer problem, internal inefficiency, or market gap that isn't currently being addressed? * **Synergies:** Were there any potential connections between different ideas or departments that weren't explicitly made? --- ## 6. Actionable Next Steps & Accountability Matrix Create a definitive list of all action items. Do not just list what was said; synthesize and clarify ambiguous tasks. Use a table with the following columns: **Action Item**, **Assigned Owner(s)**, **Implied Deadline**, and **Key Collaborators**. | Action Item | Assigned Owner(s) | Implied Deadline | Key Collaborators | | :---------- | :---------------- | :--------------- | :---------------- | | | | | | --- ## 7. Key Quotes & Communication Nuggets Extract the most powerful, insightful, or representative quotes from the transcript. For each quote, briefly explain its significance. These are soundbites that can be used in follow-up communications, presentations, or strategy documents to anchor the narrative. * **Quote 1:** "[Insert direct quote]" * **Significance:** [Explain why this quote is important, e.g., "This perfectly encapsulates the core customer frustration we need to solve."] * **Quote 2:** "[Insert direct quote]" * **Significance:** [e.g., "This comment from [Name] reveals a deep-seated concern about our technical debt."] --- [PASTE YOUR FULL MEETING TRANSCRIPT HERE]

Stage 4: The Platform Builder – From Workflow to System

The final evolution is to move from a personal workflow to an automated system. This is where you take the master-level prompt from Stage 3 and build a pipeline around it. Instead of manually running the process, you create an automated workflow where, for example, a meeting transcript is saved to a specific folder, which triggers the “Strategic Partner” prompt and delivers the finished intelligence briefing to a designated Slack channel.

This is the ultimate form of leverage. It’s no longer about your personal skill in a single session; it’s about embedding your expert process into the fabric of the organization. You’ve created a system that significantly raises the floor for everyone.

A Note on This Workflow: This seven-stage diagrammed process is a powerful example of a mature, comprehensive meeting notes system. It’s worth noting that this level of detail is likely overengineered for most common use cases. However, it’s designed this way purposefully to illustrate just how advanced and robust a workflow can become when you think systematically about raising the floor on quality and automation.

---

The Human’s Role: The Engine of the Framework

There’s a crucial layer that sits on top of all these stages: iterative refinement. An expert never just accepts the first output. This is where “knowing what good looks like” comes in—your domain expertise allows you to critique the AI’s draft and guide it to a better outcome.

A sophisticated user implements a multi-step refinement process:

1. Generate the First Draft: Let the AI produce its comprehensive analysis.
2. Apply a Condensing Framework: In a follow-up prompt, ask for refinement using a specific method like the Minto Pyramid Principle.
3. Apply the Final Human Touch: Read the condensed version and make the final tweaks yourself.

This collaborative process is what produces a final artifact that is significantly better than a simple one-shot summary.

---

The Craft Matters.

The point of this spectrum isn’t to say everyone should be operating at Stage 4 for every task. Pragmatism still rules.

The point is that “using AI” is not a monolithic activity. The floor has changed, but so has the ceiling.

So the next time someone asks, “How much of this was AI?”, the real answer isn’t a percentage. It’s a measure of intent, creativity, and skill.