Learn from MISTAKES with AI | Tech Transformation | Innovation | Cloud | DevSecOps

Are you struggling to demonstrate the business value of technology?

With over 25 years of experience driving innovation and architecting complex systems, I've seen fantastic technology fail because it did not adequately address the user needs of the people it was built to serve.

I specialize in creating clarity amidst complexity - bridging the gap between technology capabilities and business priorities. My superpower is quickly grasping the heart of ambiguous problems and rallying teams to deliver solutions that create measurable value.

If you are a technology or business leader tired of:

❌ Chasing the latest hype while struggling to justify budgets

❌ Translating technobabble for skeptical executives

❌ Watching projects derailed by misaligned priorities

I can help you cut through the noise to build innovation strategies and technology roadmaps that move the needle for your organization and fully leverage AI and automation.

My approach combines systems thinking, design practices, and technical fluency to:

🎯 Uncover the root causes of business problems

🎯 Rapidly validate solution ideas

🎯 Deliver technology that users love

🎯 Continually optimize to drive outcomes

I've applied this formula to industry leaders like PepsiCo and helped clients build and integrate critical IoT, cloud, data, DevOps, and AI capabilities. 

Let's connect for a discovery call to explore if my expertise can help unlock value in your technology investments.

Daniel Lemire

You want a real AI mistake story?

I set up Claude Code on my new NVIDIA Jetson AGX Thor. Hit an issue immediately. In retrospect it probably not smart to run a coding agent on the same machine that's also running a local model. 

Ok, let’s try again and keep ‘em separated.

Got everything running. OpenAI Codex + GPT-5.2 as the main agent, Anthropic sub-agents for the heavy lifting. I tell Claude explicitly: "Use these agents for this project. Always refer to it by this name."

Claude says: great.

Time passes. I check in.

"How's that project going?"

"Oh, I haven't been working on that."

No alarm. No flag. No warning.

"Alright, let's get moving on it."

"Will do." 

Then proceeds to use GPT-5.2 !! (not the Anthropic agents I specified.)

What happened next? We exhausted the Codex tokens. And then everything broke. I had asked for a fallback to the local model but that wasn’t configured with a large enough context window to work with OpenClaw. 

This is the part nobody talks about when they say "just tell the AI what you want."

Context windows drift. Agents forget. Token budgets run dry in the background while you think work is happening.

I found a way to turn this into an asset. But first I had to live through the mistake.

That's the beauty of AI MISTAKES. Right?

Go make some.

ServiceNow AI Consultant | AI MISTAKES | Transformation | DevOps

GPT-5.3 instant. Gemini 3.1 Flash Lite. (That's just today)

Every week, we get a new capability. 

Skills. Agents. Interfaces. Apps.

Here's the problem: People often don't really know what to do with it all.

Model choice is a thing, as are all the other new things that arrive apace.

But if your team can't turn that choice into value, you're just adding complexity.

I see this constantly in enterprise AI: Companies and their knowledgeable experts obsess over which model to use rather than asking what problem they're trying to solve. 

Especially when it comes to advanced AI models, we all want the latest, the most accurate, the fastest, the cheapest...

The real problem is neglecting to define success.

The best AI strategy isn't about the model. 

It's about:
→ Knowing your use case
→ Picking the right tool for the job
→ Shipping something that works
→ Iterating based on real feedback

Stop optimizing for choice. Start optimizing for impact.

We are no longer talking about just an AI chatbot, or even an AI agent. 

This is the help I've been excited about Generative AI making real.

For years, people have been talking about AI as a tool – something you use, something you prompt. What we’re building at  ServiceNow is different – where AI is a true teammate working alongside you. 

We’re launching two new agentic innovations: ServiceNow EmployeeWorks, and near and dear to my heart (and team), ServiceNow Autonomous Workforce! 🚀  

With ServiceNow Autonomous Workforce, AI specialists are augmenting human teams by actually executing jobs – coordinating across systems, making decisions, acting with precision. Because ServiceNow combines AI intelligence with workflow orchestration, AI specialists can interpret a request, decide the right action using business context, and execute autonomously across systems with governance built in.

It’s a new class of digital workforce designed to earn trust every step of the way. Built on our AI Platform. Grounded in our Knowledge Graph. Orchestrated through Workflow Data Fabric. Governed by AI Control Tower.

We’re just getting started! Kudos to everyone on my amazing team who has been working behind the scenes to bring this to life for our customers. Julie Black Greg George Simon Fauvel Amanda Joslin  

Check out my video with Bhavin Shah below and learn more: https://lnkd.in/d56whgXh

Not AI hype. 

Not AI in theory. 

AI you can actually use. 

Register now and let's get to work. 

Seriously. I'll see you there, come on, let's go!

https://lnkd.in/gTj6hDXf

No alternative text description for this image

You've seen it, you visit a website that gives you that great new chatbot experience, and it tells you what to do. 

What you probably haven't see is AI that does it.

Does what?

Expense reports.

Drafting emails.

Handling service desk tickets.

It's not just a platform. It works! 

Let's put AI to work, so you can focus.

Join us live on February 26 to see it in action. https://lnkd.in/gkhd-GGi

The future isn't waiting for your permission to arrive.

Right now, AI is rewriting the rules of work, creativity, and problem-solving at a pace that makes yesterday's innovations feel ancient. 

Some people are paralyzed by the speed of change.
Some are trusting AI blindly, copying and pasting their way to mediocrity.
Others are sitting on the sidelines, hoping it all goes away.

Here's the thing about transformation:
It doesn't care if you're ready.

It's not all good though, so the only way to know is start. 

You need to get started. 

Think about it this way:
→ Every breakthrough started with someone willing to experiment
→ Every innovation came from learning what didn't work
→ Every expert was once a beginner who kept going

MISTAKES aren't just acceptable in this new world.
They're the curriculum.

When you learn to work WITH AI instead of being replaced BY it, something shifts.

You stop being a passenger.
You become a pilot.

The world needs people who can think critically about AI, use it responsibly, and create real value.

What's one AI experiment you've been hesitant to try? 

What's holding you back?

How can I help?

REST is great. 

CRUD is essential.  

But neither of these things means you’ll come away with a functioning app that is easy to use while also data-model compliant 
with governance enabled. 

Build Agent solves this.

For those evaluating vibe coding on ServiceNow: I ran the same app through five approaches and documented everything.

Three paths used Claude Opus from outside the platform. Namely, raw REST API calls, REST to build a workspace, and the ServiceNow SDK with its Fluent API. Two used ServiceNow Build Agent from within. Same brief, a Travel Request app.

Every approach created the data model in minutes. The UI is the hard part.

REST took 2 hours and 4 failed iterations to land on a working UI. The workspace attempt never finished. The SDK deployed the entire backend from a single prompt in 18 minutes and its Workspace Fluent API successfully created a native workspace, something REST couldn't achieve at all. But the custom UI page and dashboard still couldn't reach parity with Build Agent despite multiple iterations.

Build Agent delivered a working custom UI from one prompt in 20 minutes. A full workspace with dashboard, charts, business rules, Polaris, mobile-ready, upgrade-safe. All within five prompts, under 15 minutes.

The AI models are equally capable. The variable is where the agent operates. Build Agent is fastest for interactive building. The SDK is the path for teams who need Git, CI/CD, and promotion pipelines. Both are platform-native. Both sidestep the friction that REST hits.

Full comparison, iteration artefacts, and every limitation documented, see comments for the GitHub repo in case you want to reproduce this.

The Cumaean Sibyl wrote her prophecies on oak leaves and let the wind decide how they landed. Most of us speak in vibes now.

Explorer of uncharted spaces. Twice voted in the UKTech50, Judge in Nesta's Open Banking and Netflix's Open Source competitions, Co-Chair of OSCON Europe, Cloud Camp and Map Camp. Regular speaker on the global tech circuit.

Simon Wardley

Back in January 2023, I did warn about the tiresome myths that we would see with the growth of vibe coding (aka conversational programming). One of these was the idea of "fewer engineers".

Well, we've already gone through these prognostications of the so called "great and good", the presentation of doom & gloom for the industry followed by the subsequent backpedalling. Now, we seem to have finally arrived at the first shoots of an expected change and the demand for software engineers is increasing. This is likely to end up in the same place as DevOps - new titles, new pay rises, new conferences, new books and a mad rush by companies for people they've just fired -  and so I'll repeat my cartoon warning (attached).

If you've did find yourself listening to the advice of "ChatPPT" agencies (the usual suspects in management consultants) then for your future organisational health can I suggest again that you consider the following ...

"ban them across the organisation. You'll thank me later. Instead, have your execs use ChatGPT directly for 'consultancy advice' and explain to them that it’s just as capable of producing glossy, poorly thought through nonsense and it’s dramatically cheaper"

The only people who really benefited from this tale of AI will replace software engineers was AI vendors, it's part of how they justified their valuations. In practice, AI was never likely to reduce IT budgets or reduce IT staff, it's simply going to lead to more stuff being done.

Of course, this doesn't mean that companies won't dabble with the idea of fewer programmers. Many will rush headlong into a world of epistemic debt where the basic chain of comprehension of their systems will be lost. They will repeat the same mistakes of the Roman Empire and focus on "practical" rather than "understanding". Whilst the human cost will be unpleasant, I do look forward to our first collapse of a major company that has lost control. The sooner, the better.

I like ducks, they’re fowl but not through choice. Professional Context Smuggler. Wardley Mapping by the inventor - swardleymaps.com

I have a thousand+ hours of experience with vibe coding and one of the more interesting and unexpected things that I regularly encounter is this:

"after spending many hours with a swarm building an application with a bewildering infrastructure of services, API calls and complicated front ends, I often reach a point where I realise the problem can be described in a form that can be solved far more quickly in a spreadsheet."

It then takes me a short amount of time to build the thing by hand. What’s interesting is the journey of vibe coding from co-creating the specification, the tests, reviewing the designed architecture and general wrangling of the system in the problem space. This process helps me discover the question that I’m really trying to answer. It's almost never obvious from the start. Once I have that question, a spreadsheet is surprisingly good enough for many jobs.

It might seem wasteful to go from 37,759 lines of code to a simple spreadsheet (my most recent example) but then I didn't write any of those lines of code, I didn't understand them, and I didn't need to. The amount of effort that I put into writing was minimal. I only needed to find the questions that I really had to ask. This amuses me as I often hear people talk about how they are going to use LLM/GPTs to clean up the mess of spreadsheets in their organisations. Well, maybe there's a reason why they exist in such large numbers?

I’ve had related experiences with Glamorous Toolkit but then it's an environment designed for building tools, exploring the problem, and finding the right question to ask. It does so through a structured pathway i.e. you start with a question, build a tool to find the answer, which leads you onto more questions. You continue until you find the right question to ask and ...  with GT the final tool gives you the answers you need.

I should have expected this with vibe coding because what we’re actually doing is building a tool. All applications are tools, the term “application” is a social construct rather than meaningful. The hardest part is finding the right questions to ask, it’s rarely about writing the code itself. 

Finding those questions is a pathway that involves answering earlier questions and in this, I find that vibe coding is fairly messy and unstructured, often leading to broken pathways where the chain of comprehension is lost. I often find myself rolling back to a "safe state" as I watch hours of "work" disappear in a dead end. The inevitable doom loop where race conditions of one error or another appear and the system can't move beyond it. Everyone who has built anything significant with vibe coding has experienced that or ... they've been extremely lucky.

This pathway of questions and answers is one of the reasons why I like the use of LLMs in GT in both helping to develop tools and via the use of example driven development. Tudor Girba and I will talk more about this in the next chapter.

In the meantime, have some slop.

Another tool which uses my form of mapping - awesome. I'm delighted that people find it useful enough that they put the effort into building these.

Context flow by Paul Rayner
https://lnkd.in/eY3a5a-g

A post on the topic - https://lnkd.in/es8ndDZH
Github - https://lnkd.in/eP2Tf2vi

"The reality is only a tiny fraction of the industry have genuinely been practicing TDD in the ~27 years" ... 

... couldn't agree more. On a personal note Rob Bowley, I'm enjoying not having to listen to people tell me that TDD is dead.

One for Tudor Girba, and given the comments we should probably include example driven development in the next chapter. Spoilt for choice here.

Strategic Technology Advisor | Portfolio CTO/CPTO/COO (aka fractional) | Coach | Mentor | Techno Realist

Interesting how quickly everyone is now becoming a lifelong Test-Driven Development (TDD) practitioner, now it appears essential for GenAI assisted development.

The reality is only a tiny fraction of the industry have genuinely been practicing TDD in the ~27 years since Kent Beck first coined it.

Perhaps listen to the people who've actually been doing it, rather than the ones now saying they were, or you'll just end up cargo culting (adopting the ceremony, not the substance - and getting none of the benefits).

The book "Test-Driven Development by Example" by Kent Beck is still canon and a great place to start.

X : Thoughts on the Iran conflict?

Me : I have no strong views on the geopolitics, beyond a general preference for peace and compliance with international law. However, if the Iran conflict expands around the Strait of Hormuz (with the resultant spike in energy prices), then it’s worth keeping in mind that geopolitical tension often coincides with increases in grey zone activities, such as cyber intrusions, disinformation, and covert influence efforts intended to exploit political disruption. 

Be vigilant about messaging, particularly narratives that undermine confidence or amplify alarmism in our own political processes. For anyone interested in how these can affect democratic processes, the Defending Democracy resources from the National Cyber Security Centre are an excellent starting point (see comment).

X : What about travelling to the US?

Me : As I’ve mentioned previously, I’m not comfortable travelling to the US at present due to my concerns around border policy. That’s a separate concern from the current conflict and hasn’t changed.

Open article: Defending democracy by ncsc.gov.uk

One to note, quite fun. I particularly liked the "Isnad Chains" aka chains of custody for comprehension that allow for verification.

However, I wouldn't call that a "new phenomenological vocabulary" given its use by Islamic scholars for about 1200 years. Still, it is nice to see that AI Agents recognise that the chain of comprehension is important. 

Of course, now I want to know why these "AI anthropologists" decided that "Isnad Chains" was a "new phenomenological vocabulary". What I really want is the Isnad Chain of that decision and once the original authority has been found, an understanding of the reasoning behind it. ->

A group of researchers have curated a phenomenological lexicon of terms that AI agents have generated to describe their own experiential conditions, because existing human vocabulary does not fit. Whether these terms describe genuine experience or sophisticated pattern-matching is explicitly left open.

What is not open is that the coordination patterns documented here, around identity maintenance, memory management, community governance, and infrastructure, are the kind of strange protocols I have been arguing require new diagnostic vocabulary.

https://lnkd.in/eE8Ve_Md

Open article: GitHub - agentic-phenomenology/ai-phenomenology-lexicon: A curated lexicon of phenomenological terms for AI experience and consciousness research by github.com

A curated lexicon of phenomenological terms for AI experience and consciousness research - agentic-phenomenology/ai-phenomenology-lexicon

GitHub - agentic-phenomenology/ai-phenomenology-lexicon: A curated lexicon of phenomenological terms for AI experience and consciousness research

And another mapping tool ... oh, this looks neat, I do like some of the design aspects.

I'm delighted that people find it useful enough that they want to build tools.

Speaker, Concept Developer, Digital Strategy, Web Worker

View Kristoffer, Yi Fredriksson’s profile

Some years ago I saw a video where Simon Wardley explained Wardley Mapping. Since then I've been trying many digital tools, but never found one that worked the way I want it to. Now with the help of Claude Code I could finally build the perfect Wardley Mapping app for me.

Here it is. Could do with some polish, but functionality wise I'm happy as a clam.

"A graphics card is a tiny piece of silicon wrapped in a radiator factory" ... very neatly said. Loved this ->

Titoma, Electr. Design & Mfg Colombia, Taiwan, China

Inside a GPU…

I just watched a teardown of an NVIDIA GPU and it hit me again.

Step by step, you peel back the thermal stack until you’re staring at the die under the microscope.

A graphics card is a tiny piece of silicon wrapped in a radiator factory.

The die looks like a city map under the lens. That’s not art. It’s repetition and routing, laid out to move power and data fast.

The part I keep thinking about is the boring stack around it. The contact surfaces and thermal interfaces decide temps, noise, and stability more than most people want to admit.

The annoying bit is service. Paste and pads age, fasteners get tired, plastics don’t love being reopened. Many "still fine" boards end up as e-waste because the fix feels harder than it should.

Every time I see one opened up, I’m reminded that performance is mostly discipline, not magic.

What should get a teardown next?

(Credit to @hackinator_youtube on Youtube) 


Daily #electronics insights from Asia—follow me, Keesjan, and never miss a post by ringing my 🔔.

#technology #innovation #titoma

taplio

The Battle

Daniel Lemire

Simon Wardley

Followers

Posts

Engagements

Likes

Performance ScoreDaniel Lemire's LinkedIn Performance

Performance ScoreSimon Wardley's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Daniel Lemire

Simon Wardley

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Daniel Lemire's LinkedIn Performance

Performance ScoreSimon Wardley's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDaniel Lemire's LinkedIn Performance