Tuesday, November 17, 2026

Speed of AI Coding Matters, but Human Testing Matters More!

In light of the grounding of flights in the US due to a new AI-built air traffic control system being implemented, it is very clear that the speed of AI coding is not the only thing that matters. There must be the ability for humans to relatively quickly test and validate that what was AI-coded is properly running and does no harm.    

In other words, while AI dramatically accelerates coding time, it is not the same as confidence in the quality of that code.  It does not absolve you of responsibility and judgment for a properly tested product!  What you still need to do is test and verify and these activities must be elevated. Why?  As we have learned, logical gaps surface leading to failures in services, data, infrastructure, and reliability.

Review the code for understanding.  Ask the AI tool to explain to you what this code does, line by line. Ensure it's not vague, and aligns with what you are thinking. Ask what the expected outputs and outcomes would be. Then ask what would break the code. Finally, ask AI why this approach was chosen over alternatives? A useful litmus test is to ask if you understand the code well enough to feel comfortable maintaining it for the next year.

Then test, test, test.  The more mission-critical the application you are building, the more important it is that humans are directly involved in the testing. Testing should include all the important practices that have been around for decades now.  This will include Human-led verification for: Version control; Code Reviews; Unit Testing; Integration Testing; System/Performance Testing; Usability testing; and more! 

The importance of human-led testing and verification becomes more important than ever. There is a fundamental distinction we need to make: While AI can generate code, it cannot independently establish that the code is correct, appropriate, secure, maintainable, and delivering the intended business value. There is a fundamental distinction we need to make: AI can quickly generate code. It cannot independently establish that the code is correct, appropriate, secure, maintainable, and delivers the intended business value.


Wednesday, April 1, 2026

What are 3 big AI Shifts in the midst 2026?

AI is leaving the “wow” phase and entering the “prove it” phase. Many companies are about to realize their strategy was built for demos, not reality. In other words, AI is growing up. It’s time to move on from the hype toward the hard tradeoffs. What are 3 big AI shifts?

  • Magic to Money 
  • Demos to Deployment 
  • Capabilities to Consequences

For the past few years, AI has lived in the realm of spectacle—impressive demos, viral moments, and “did you see what it can do?” reactions. That phase is ending. What’s replacing it is more grounded, more valuable, and a lot less comfortable. 
AI is shifting from magic to money, demos to deployment, and capabilities to consequences. Let’s examine these more fully.
Magic to Money
The novelty is wearing off. Organizations are no longer impressed that AI can generate content, write code, or analyze data. People are asking whether it actually drives revenue, reduces cost, or creates a defensible advantage. This is where many AI initiatives encounter friction: what initially appears magical, struggles when tied to real business metrics, messy data, and existing workflows. A current challenge is proving consistent, measurable ROI, not just isolated wins and talk.
Demos to Deployment
We’ve all seen the polished demos. But deploying AI into production is a different game entirely. Integrations, governance, reliability, edge cases, and user adoption quickly surface. The gap between “it works in a demo” and “it works every day in real life” is where most efforts stall. The winners are no longer the ones with the flashiest models, but rather those who can operationalize them at scale. A current challenge of getting to “real life” is bridging the last mile from prototype to dependable, repeatable execution.
Capabilities to Consequences
As AI capabilities grow, so do concerns about accuracy, bias, job displacement, environmental cost, security, and trust. Leaders are increasingly forced to weigh not just what AI can do, but what it should do and what risks they’re willing to accept. The conversation is shifting from innovation to responsibility, often faster than organizations are prepared for. A current challenge is how to manage risk and accountability without slowing innovation to a crawl.
What shifts are you seeing, and what challenges are you facing in getting to the full usage of AI?






Monday, February 23, 2026

Risk of AI Sameness

Everyone is racing to adopt AI tools.  Most people are using AI to create “speed”.  Faster emails. Faster proposals. Faster code. Faster content. And yes, faster is good.  But here’s the quiet risk no one is talking about: if everyone in your company uses the same AI the same way, you may slowly start sounding exactly alike.

Are you ignoring “AI sameness” risk? Without intention, standard AI tools can homogenize your messaging, flattening distinct perspectives into one generic voice. The danger isn’t bad output. It’s average output at scale. AI isn’t going to replace your team. It’s going to standardize them, but not in a good way.

When everyone uses the same tools, trained on the same data, prompted in the same way, you don’t get divergence, you get convergence. Everyone will sound the same, with the same sentence structure and tone, and with the same “polished but generic” voice. Then there is the further danger that over time, unique thinking gets flattened into safe, average, AI-shaped output. Not because your people aren’t smart, but because the tool defaults to the statistical middle.

AI should amplify your edge, not sand it down. AI should sharpen your thinking, make it more opinionated, and more differentiated. If you’re not intentional about how your teams use it, it can lead to this sameness.  What can you do to avoid this sameness?

  • Craft your own perspective before prompting AI.  As it relates to the topic, what do you actually believe? What do most people get wrong about this? What would you argue in a debate?
  • Use AI to provide you with a draft, not the deliverable. Within that context, establish one strong opinion. Provide one specific example from your world. Craft your own voice so that sentences sound unmistakably like you.
  • Add Friction to the output. AI will often be a people pleaser, so challenge your output. Ask what’s missing. Determine if it feels too safe. Consider if it sounds too predictable? If it reads smoothly but doesn’t make you think, that’s a warning sign.

AI naturally drifts toward the statistical middle. Avoiding sameness requires intentionally looking for differences. It must include injecting strong beliefs, specific context, and human judgment layered on top.  And honestly? The companies that figure this out won’t just use AI faster. They’ll use it to amplify their uniqueness.



Saturday, January 31, 2026

AI Coding: Shifting the Developer Role

Coding with AI is producing code at a faster rate than ever and accelerating the release of production increments. The code can be generated in minutes and feels good because of how quickly it is created. This begs the question, what does the software developer do now?  It changes where the developer’s focus goes. 

While AI is generating the code, it doesn’t own the code.  The developer remains accountable for it, which means they must review the code deeply enough to understand how it works, why it works, and where it could fail. Also, they must focus on verification activities surrounding the code. This article is based on some experimentation with AI and ensuring the developer has a good understanding of the code changes. 

Think of AI as a Junior Engineer, and it is your job to raise them up. It can produce a lot of code quickly, and it can be confidently wrong when doing so. It has no sense of risk, context, or consequences.  Think of the verification as the handoff where ownership transfers to a human. It is still your job to ensure a verified and quality outcome. This should take a majority of engineering time or later on, logical gaps leading to failures in services, data, and infrastructure. How does the Developer responsibility shift?

  • Review code written for understanding.  Ask the AI tool to explain to you what this code does, line by line. Ensure it's not vague and be sure it aligns with what you are thinking. Ask what the expected outputs and outcomes would be. Then ask what would break the code. Finally, ask AI why this approach was chosen over alternatives. A useful litmus test is if you wouldn’t feel comfortable maintaining this code for the next year, you don’t understand it well enough
  • Ensure that the code was version-controlled properly and in the correct branch. This includes checking for potential issues before merging it into the main codebase.
  • Step up Code Reviews.  This means to peer-check code for quality and adherence to standards.  The Developer should share coding standards with the AI tool to ensure it aligns with standards. If coding standards are missing, then they must be written and then added to the AI tool’s vector of information.   
  • Spend time sharing responsibilities with Testing to ensure all verification activities are completed.  This should include appropriate testing: Unit Testing (e.g., testing individual components or functions in isolation), Integration Testing (e.g., testing how different components work together), System Testing (e.g., testing the system as a whole for speed, scalability, and stability), and more. 

AI reduces typing time. It does not absolve you of the responsibility and judgment for a product well built! AI changes where time is spent, not whether time is spent. While we will spend less time coding, we are still accountable to spend time verifying and understanding the code that has been generated to ensure it meets the needs of the outcomes we are looking for.  





Wednesday, April 30, 2025

Beware of AI BS (aka, Hallucinations)!

User beware! Did you know that your AI query is subject to hallucinations?  What is an AI hallucination? When AI inadvertently generates false or misleading information that seems plausible but is not rooted in reality. It is trying to give you an answer. These are errors in AI outputs that arise from flawed reasoning or inaccurate training data, typically not from malicious intent. 

For example, a language model like ChatGPT might generate an article with fake references or make up scientific facts because it is just predicting what should come next based on patterns in data. In fact, when I asked “what was the duck wearing when it won the Boston Marathon?”, it said that “the duck was wearing a quacking pair of sneakers and a feather-light singlet when it flapped its way to victory!”

This should not be confused with people deliberately using AI tools to create misinformation, typically to manipulate public opinion or cause harm. AI itself may be used to generate highly realistic but fake content, such as fabricated news articles, doctored images, or videos. For example, AI may be used to create Deepfake videos to manipulate someone's face and voice to make them appear to say something they never did. 

Turning back to actual AI hallucinations, what are the risks where it inadvertently poses several serious dangers? Generally, creating and sharing hallucination misinformation can spread quickly, particularly in news, health, legal, or political contexts.  Users who trust AI outputs may unknowingly share false information, amplifying its reach. What are more specific dangers?

  • Generating legal and medical judgments or diagnoses. AI-generated hallucinations in legal documents, medical advice, or financial reports can lead to harmful or even illegal outcomes. This can damage reputations or result in malpractice.
  • Misinterpreting security and safety threats. In cybersecurity or military applications, a hallucinated misinterpretation of data in critical systems (e.g., aviation or nuclear control) could trigger wrong decisions with high-stakes consequences.
  • Spreading stereotypes and reinforcing bias. Hallucinated outputs might reflect or invent stereotypes or discriminatory patterns that reinforce social biases. This can be especially harmful in generative content involving race, gender, religion, or culture.
  • Damaging reputations and polluting research. Fake references or fabricated studies can pollute scientific research, especially if unnoticed in peer review or student submissions. AI hallucinations in education can mislead learners or promote academic dishonesty.

After enough hallucinations are shared and spread, repeated exposure to hallucinated content undermines trust in AI tools and technology in general. Ultimately, if enough misinformation occurs, there will be an erosion of trust and hesitancy to adopt AI. The important thing is be aware that AI tools will inadvertently generate false or misleading information. Don’t accept answers at first blush. Instead, verify the answers, verify the references, fact-check the outputs, and ask AI to double-check its results.


Monday, March 31, 2025

The benefits of Divergent and Convergent Thinking

Divergent thinking embraces an approach that promotes collaboration to generate ideas and numerous solutions with no censoring of ideas.  Once an allocated timebox to share and discuss ideas and options is concluded, divergent thinking is paired with convergent thinking which are approaches to more quickly come to consensus. This is illustrated below.

A traditional divergent technique is brainstorming which focuses you on generating ideas in an unstructured manner.  But due to the fast pace of most companies, we are quick to shutdown brainstorming by criticizing the options of others (including our own). Because of this, I recommend 15 minutes of silence for all brainstorming activities to ensure all ideas get a chance to be revealed prior to any discussion,

People generate their ideas with no discussion so no one’s ideas are dismissed by those more assertive or intimidating. As you may guess, often the quiet team members (aka, introverts) have some of the best ideas and a divergent “quiet” period allows the talented introverts to get their ideas on the table.  The key to divergent thinking is that there is absolutely no censorship of the ideas or solutions coming forth.  All ideas are valuable. 

Once the divergent time-box is concluded, it is time to commence with convergent thinking.  Convergent thinking is an approach to systematically limiting your options and focus on one direction. From divergent thinking, there are often more ideas and possible solutions than can be handled. 

In convergence, various techniques can be applied to bring together ideas.  This may include looking for affinities where common themes may arise from some of the ideas.  It can also include red-dot voting, a non-confrontational prioritization technique where people quietly place their dots on what they think are the top ideas for solving the problem.  Ideas with the top number of dots are discussed until a top option is selected.    

The value of divergent thinking cannot be underscored.  Spending as little as a few hours of collaborative brainstorming coming up with ideas is a drop in the bucket of the thousands of hours spend in building a solution.  The tiny investment is worth ensuring that many options are made available. 

Also, the divergent and convergent pairing can occur in a continuous manner.  For every customer need expressed in a user story, in the beginning of an iteration or sprint, divergent thinking can be used in sprint planning to consider options prior to jumping into the sprint where you build the customer need. When the next iterative or sprint begins, we move back to divergent thinking to consider options for the next step of user stories prior to embarking on the sprint.    

It is also very important to be explicit about whether you are in a divergent or convergent thinking mode. Some people naturally gravitate toward divergence while others gravitate toward convergence.  It can reduce issues in interpersonal relationships to explicitly state whether you are in a divergent or convergent mode so that it is clear whether you are soliciting options or closing down options.  

Finally, when you have an organization that is constantly pushing you to move forward, convergence is stressed whether consciously or unconsciously and this is a detriment to innovation and being open to numerous possible solutions.  It may take an organization to explicitly initiated divergent thinking to achieve more innovation and ideas.  

Friday, January 31, 2025

Leading with Agile Principles!

What does it mean to be Agile? It starts with aligning with Agile values and principles. In this short guide, I provide insight into the 12 Agile principles look like in action. The goal is to identify what evidence would look likes to determine if there is alignment with these principles, if a culture change may be occurring, and to prove the assertion that you are indeed Agile. 

Let’s take a closer look at each of the Agile principles and what they might look like in action (each in their own one-page article).

It is up to you to determine what supporting evidence looks like when a company believes in each of the Agile Principles. It is worth experimenting with this as it will help you better understand and embrace the Agile principles and put them into action in your organization. Good luck on your Agile journey!