Jeff Bezos shared this formative moment of feedback from Jeff Wilke during a conversation at the Italian Tech Week 2025.

Too many ideas? And their weight posing an existential risk to a business like Amazon?

There is a lot wisdom in this observation and the this segment of Jeff’s conversation (roughly from 23:20 to about 27:00 in the video linked below). The core idea, as he explains, is that new ideas create work. And when the capacity of an organization is already full, those new ideas simply stack up like “work in progress” and do not add value. Worse, they create distraction and make the organization less productive.

He went on to say, “you have to release the work at the right rate that the organization can accept it”. That has meant holding ideas close, refining them, and being much more deliberate about the time and place they are released. And as a leader, it raised the question, “How do you build an organization that can be ready for more ideas? How do you do more ideas per unit time?”

The concept resonated with me. In a flurry of energy and inspiration ideas come in droves. But rarely do they yield a lot of progress. Often times teams seem bewildered and confused about what to do, and I suspect it is a consequence of the effect Jeff is describing: More backlog, more distraction, less progress.

So how do you capture ideas, nurture them, and carefully release them? What do these patterns look like in practice? I would love to hear ideas and examples.

AI is everywhere. It seems like everyone is “deploying and adopting AI” and that everyone has an “AI strategy”. But what does this mean? Dig deeper and you discover that AI is an umbrella term for many things and more specific language is required to truly understand what is being adopted, developed and deployed.

What follows is my working understanding of AI and the “flavors” that fall under the umbrella.

Note: Comments and suggestions to improve and refine these descriptions are welcome and appreciated.

What is AI?

A Definition of AI

The field of computer science focused on building these systems spans five broad groupings, each with distinct roles and capabilities.

Five AI Groupings

• Symbolic / Rule-Based AI — encodes human expertise into explicit rules and logic to automate decisions in structured domains.

• Machine Learning & Predictive AI — learns patterns from data to predict outcomes, classify inputs, or optimize decisions, ranging from regression and clustering to reinforcement learning and deep neural networks.

• Generative AI — creates new content such as text, images, audio, or code by learning patterns from vast datasets.

• Agentic / Autonomous AI — goes beyond prediction and generation to plan, reason, and take autonomous actions toward goals, often by combining generative and predictive techniques with external tools.

• Embodied & Future AI — extends intelligence into the physical world through robotics, and into speculative domains like Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI).

AI Groupings, Descriptions & Flavors

Within the five groups sit 10 “flavors”. The flavors are distinguishable based on:

• Data type (structured vs unstructured, labeled vs unlabeled, symbolic vs experiential).

• Problem type (predict, discover, generate, act, interact).

• Autonomy (fixed rules → predictive → creative → autonomous → embodied).

• Outputs (forecasts, clusters, content, actions, physical changes).

• Maturity (mature, emerging, experimental, speculative)

Symbolic / Rule-Based AI

Description: The earliest wave of AI — systems built on explicit human-written rules and logic. Great for well-defined problems, brittle for messy real-world data. Distinct for its hand-coded logic

1. Rule-Based / Symbolic AI (GOFAI): Expert systems using “if/then” rules; encodes human expertise into decision logic.

Machine Learning & Predictive AI

Description: The dominant wave of AI today — systems that learn patterns from data to predict, classify, and optimize outcomes. Distinct for its pattern recognition from data.

2. Supervised Learning: Learns from labeled examples (e.g., predicting house prices, spam detection).

3. Unsupervised Learning: Finds hidden patterns in unlabeled data (e.g., clustering customers, anomaly detection).

4. Reinforcement Learning (RL): Learns by trial and error with rewards (e.g., AlphaGo, grid optimization).

5. Deep Learning: Multi-layer neural networks that handle complex, high-dimensional data (vision, speech, text).

Bonus: Explainable AI (XAI)

Description: Tools and techniques to make ML/Deep Learning decisions transparent and interpretable.

Explainable AI (XAI) is a cross-cutting field focused on making decisions from complex models, particularly deep learning, transparent and interpretable. It is not a type of learning itself but a set of tools and techniques applied to other flavors.

Generative AI

Description: The creative wave of AI — models that produce new text, images, audio, or code based on patterns in training data. Distinct for its content creation ability.

6. Generative AI Systems: Uses GANs, diffusion models, and transformers to create content (ChatGPT, Stable Diffusion, Copilot).

Agentic / Autonomous AI

Description: The emerging wave — Agentic AI goes beyond single-shot prediction or generation by using a feedback loop. These systems autonomously plan, reason, use external tools (like search engines or APIs), and refine their actions based on an internal state and memory, all to achieve a specific goal. Distinct for its autonomous action and orchestration.

7. Agentic AI: Combines reasoning, memory, and tool use to act like a digital co-worker (AutoGPT, LangChain agents).

Embodied & Future AI

Description: Extends intelligence into the physical world through robotics and sensor/actuator systems. This differs from simple automation by using real-time perception and AI reasoning to adapt to new environments and tasks. Distinct for its physical interaction or general/universal cognition.

8. Robotics / Embodied AI: AI integrated with sensors and actuators for physical interaction (Boston Dynamics robots, drones, warehouse automation).

9. Artificial General Intelligence (AGI): Hypothetical human-level AI that can adapt across any domain.

10. Artificial Superintelligence (ASI): Speculative AI surpassing human intelligence across all tasks — science fiction for now.

Summary of AI Flavors

Here are my notes:

Setting Goals

Three categories of goals

1. Personal/Family

2. Business/Career/Financial

3. Self Development

Questions to help identify and refine those goals:

• What would you do if you were financially free? If you could choose how you spend your time?

• What would you do if you had 6 months to live? What would you stop doing?

• What would you dare to dream (and do) if you know you could not fail?

Goals must be in writing. Review them regularly. Refine them. Carry them with you.

Taking Action

Creating an organized plan of action based on your goals is simple: Take a goal a write every step to achieving it. List out the steps. Organize each step to create a plan.

Urgency is the fuel for for success: “Do it now. Do it now. Do it now.”

The magic word of time management is NO.

Batch tasks. Avoid context switching and benefit from the learning curve of repeated action.

Delegate. Anything lower value or lower priority.

To delegate well you need CLARITY:

1. What needs to be done? Write it down.

2. Who should do it? Match time and capability.

3. Make expectations clear. What needs to be done and when.

4. Inspect. Check in along the way.

ABCDE prioritization method:

• A: What you must do.

• B: What you should do.

• C: What would be nice to do.

• D: Delegate

• E: Eliminate.

Applying ABCDE to goals:

• List everything you want to accomplish. Apply ABC to it.

• Take the A goals and rank in order. Take the top 3 (A1, A2, A3).

• Make a list of everything you need to do to accomplish the top 3.

• Apply ABC to the list of tasks.

• Organize and implement.

Follows a 6-step process:

1. Choose goals

2. Set Priorities

3. Choose Activities

4. Set Priorities

5. Schedule

6. Implement.

Setting priorities is about determining what is relevant versus irrelevant. Relevance is based on value.

Test for value: What impact will this have on my future? The greater the impact, the greater the value.

Subscribe now

Actions:

• One project only.

• Block out EVERYTHING else. There are no emergencies.

• Buy blocks of time.

Features:

Manager Time: FILL YOUR CALENDAR

Actions:

• Book it out to the max. A fully booked day is a maximally invested manager.

• Designate days or times for this.

Features:

The People Are the Product. The core of what we provide to the market is our people. It is their talents and capabilities, their ideas (and resulting work), and the engagement with their personalities that allows our customers to develop a relationship with our business and a desire for continued engagement at a high level and at high value.

Work Is the Tool With Which We Craft Our People. While business outcomes are important, the real work is the development of the skills, capabilities, wisdom and trust of our team. And that is created through the work we do together.

Assign Tasks to Pairs. By working in a pair, you get the majority of the benefits of a team, strengthen the relationship between the people working together, and retain an efficient allocation of resources. This combination maximizes development of the team while achieving high-quality output.

AI changes the paradigm:

• AI makes the objective outcomes. For example: Consider a Google search, which provides a list of sources, from which I (the user) distills an answer. This is all process: a faster search gives me more result to sift through to produce my answer. AI flips the script. It produces the answer. And we judge AI on how good the answer is (how thorough, complete and nuanced it is). AI is about outcomes.

• That means “Speed” is the wrong point of emphasis. Results need to be fast enough, but the true test is the quality of the results. Differentiation is based on the things we can do no one else can – why should customers expect us to consistently produce better AI solutions than anyone else?

• The future of software is AI agents orchestrating a set of other agents executing specific tasks in service of a customer outcome. While gaps in execution can be coded on the fly, a large cohort of pre-built agents allow you to go further and do better (they’ve been tensioned by the experience of the past).

• Workflows (process) get crafted transparently in response the outcome demanded. Each step is articulated. Each assumption explained. If a user has a perspective on the approach, workflow, step or assumption, adjustments can be made on the fly.

I joined Enverus when they acquired the company I was working for in 2020. Jeff Hughes was the CEO at the time and led the company until middle of last year. His leadership ethos is firmly stamped on the company, much of it captured in 44 principles that he summarized and generously shared.

I have found each to be thought provoking, usually resonating as some sort of “truth”. And like many truths, my understanding grows with time as experience and reflection add details and nuance. In his spirt of sharing, and to preserve them for reference, I have listed them below.

44 Guiding Principles from Jeff:

1. Link teams to the leading indicators of success. Usage leads CSAT which in turn leads retention. Measure usage. Other include white space and funnel.

2. In a software company, the assets are the people, not the products.

3. It's always a leadership problem.

4. Build platforms not products.

5. If you want to get something done, give it to a smaller team.

6. Don't let a crisis go to waste.

7. Product teams need to bias for speed.

8. Hard is hard to follow.

9. The worst of times can be the best for innovation.

10. The most important measures of a SaaS company are growth and retention.

11. Product teams should be linked to outcomes.

12. With products there are value creators and value enablers. There are very few value creators so manage them carefully.

13. Margin expansion starts with differentiation.

14. Create an uncontested marketplace where competition is irrelevant (blue ocean strategy).

15. Leverage (or create) immovable deadlines.

16. Get the right people into the big jobs.

17. Don't be afraid to change, the organization is always more resilient than you give it credit for.

18. Discount to make more money.

19. You get more output out of someone with either motivation or training. We almost always forget about the latter.

20. If you want to innovate, spend more time in the field.

21. The best measure of culture is engagement. Engagement is a leadership problem.

22. Quota margin, not products.

23. If you build stuff for big customers, it rarely works for small ones. The reverse can work.

24. Understand the real available market. It's never the one in the slide.

25. You must be good at the first mile before you get a chance at the second.

26. Work can and should be fun.

27. HR can either be an administrative function or the epicenter for culture.

28. Relationships are built in person.

29. First be a great company to work for.

30. Accelerate what’s going well.

31. The best people are free. They pay for themselves.

32. Small teams of high caliber talent beat large teams every time.

33. Calm is contagious.

34. Winning teams trust each other.

35. Plans are useless but planning is essential (Eisenhower quote).

36. Fresh blood is fresh perspective.

37. You can’t stand for something if you chase after everything.

38. Think big. Start small. Scale fast.

39. Get the company running well before you run it fast (Sandy Ogg).

40. If the answer isn’t “hell yes” its no.

41. Your customers can’t afford your waste.

42. It’s not a company, it’s a product line.

43. Bad gross retention normally means it’s a product problem. Product problems are harder to solve than sales problems.

44. Scalable companies decentralize decision making.

The "Great Resignation" has brought renewed focus to the terms and conditions under which we work. In general we all seek an optimum balance across 4 dimensions of reward when choosing where to work:

1. Financial Compensation

2. Quality of Leadership/Management

3. Quality of Work

4. Quality of Work Environment

As employers we much choose what balance we wish to offer, knowing that it is impractical to try and be above average on all the dimensions. The last two years have substantially damaged many historical strengths. Meanwhile mobile/remote work has altered the rubric upon which we evaluate this equation in our great "reevaluation".

In many cases the "Quality of Work" has deteriorated as we the lose context and significance of our contributions. When we cannot see the "whites of the eyes" of our colleagues and clients, it is much harder to appreciate our own contributions.

We are social creatures (even the most introverted of us appreciate some level of social interaction), and in the sterile Zoom/Teams works, we lose the subtle, perhaps mundane, but important relational moments that create the sense of connection that leaders feed on. Our experience with "Leadership/Management" suffers accordingly.

Working from home (or the beach) does have it's perks. The proximity to other aspects of our life can be life changing for parents, employees with long commutes, etc. But proximity also means the lines between work and life which are challenged by technology also become blurry in our physical environment. Worse, most of us find ourselves working at home in a physical space we never planned to spend our work hours in. For many, this is far from the optimum "Work Environment".

With so many headwinds to our non-financial rewards, it is no wonder many seek better pay, a new role/organization, or simply "take a break". For leaders and organizations, it is a unique opportunity to rethink the reward structure we offer, what compromises we are more willing to make (and NOT make), and to share our philosophy towards employee rewards. The future looks different, but leadership is about embracing our context and making the most of it for the people around us.