01
AI adoption
How tools become repeatable ways of working through ownership, training, useful first versions and better operating habits.
What building useful AI has taught me.
Personal moments of uncertainty, discovery and practical change, each followed by the framework I now carry into the next project.
Start with the question you are trying to answer
Four paths through the field notes.
02
Workflow design
Finding the real process beneath the request, then designing the trigger, decision, action and human boundary before choosing tools.
03
Production AI
The unglamorous details that decide whether an AI system survives: cost, fallback paths, security, explanations and maintenance.
04
Decision support
Turning meetings, client context and operational signals into better judgement—not simply automating more tasks.
Complete archive
All 20 field notes
01 / 20 · AI project discovery
Why the best AI projects start with a messy conversation.
I have left discovery calls feeling excited by the opportunity and unsettled by the lack of a clear answer. That discomfort is often where the real AI project begins.
Read the field note →
02 / 20 · AI workflow design
Workflow before tools: the reset that made my AI projects better.
I once walked into discovery with a mental shortlist of tools. I left realizing that every tool in my head was stopping me from seeing the work in front of me.
Read the field note →
03 / 20 · meeting transcript automation
The day I stopped treating meeting transcripts like storage.
I kept seeing valuable conversations disappear into folders. The transcript existed, but the decisions, risks and product clues inside it were already fading.
Read the field note →
04 / 20 · AI MVP development
I stopped chasing the perfect AI system and built the first useful version.
The more exciting the idea became, the further we moved from giving one real user something valuable. Cutting the scope felt painful until the product finally started teaching us.
Read the field note →
05 / 20 · AI data readiness
I wanted AI to understand the business. The business information was not ready.
The assistant sounded confident, but it was missing decisions hidden in emails, context trapped in people’s heads and documents nobody knew were outdated.
Read the field note →
06 / 20 · human in the loop AI
The best AI decision system I designed did not make the decision.
I thought the breakthrough would be more autonomy. The real value appeared when AI removed the confusion surrounding a decision and left the judgement with the person responsible.
Read the field note →
07 / 20 · what not to automate with AI
One of my most important AI decisions was deciding not to automate.
I could see how to build the workflow. I could also see that the process, data and consequences were not ready. Saying ‘not yet’ felt less impressive, but far more responsible.
Read the field note →
08 / 20 · AI workflow fallback
The AI workflow worked perfectly until a real user touched it.
The happy path had been polished for weeks. One unsupported file exposed how little thought we had given to the moment the system could not continue.
Read the field note →
09 / 20 · AI cost optimization
The AI bill surprised me. The architecture should not have.
I first treated cost as something to monitor after launch. Then one workflow showed me that the expensive decisions had already been made in the design.
Read the field note →
10 / 20 · productize AI services
The request I kept solving manually was trying to become a product.
The client names changed. The wording changed. The underlying problem kept returning until I stopped treating each request as a separate piece of work.
Read the field note →
11 / 20 · practical AI tools
The AI tool nobody called revolutionary became the one the team depended on.
It did not have a futuristic interface or autonomous agents. It prepared a report, created tasks and saved people from a routine they were tired of repeating.
Read the field note →
12 / 20 · AI process design
I kept fixing the prompt. The prompt was not the real problem.
Every revision improved the wording, but the output remained inconsistent because the system was feeding the model unclear inputs, weak context and no validation.
Read the field note →
13 / 20 · AI decision support
‘Build an AI agent’ sounded exciting. It was not a product requirement.
The agent could search, summarize, draft and act. What nobody had defined was the decision it was responsible for improving.
Read the field note →
14 / 20 · AI meeting workflow
The meeting went well. The week after it did not.
Everyone left aligned, yet tasks were missed, decisions became fuzzy and the same questions returned. The meeting needed a life beyond memory.
Read the field note →
15 / 20 · AI consultant skills
I could not deliver useful AI by thinking like only one role.
The product problem, technical system and operating reality kept colliding. The work became effective only when I learned to move between all three.
Read the field note →
16 / 20 · reduce handoffs with AI
I was automating tasks while the handoffs kept slowing everything down.
No single step looked especially wasteful. The delay lived in the space between sales, operations, finance, delivery and the systems each team used.
Read the field note →
17 / 20 · explainable AI workflow
The AI answer looked correct. Nobody wanted to trust it.
The problem was not only accuracy. The user could not see where the answer came from, what the system had inferred or whether anyone had approved it.
Read the field note →
18 / 20 · client intelligence platform
The client information existed everywhere, yet nowhere useful.
A simple account question sent people searching through email, meetings, tasks, reports and memory. The agency did not lack information. It lacked a shared intelligence layer.
Read the field note →
19 / 20 · AI engineer interview questions
The candidate knew every AI tool name. I still did not know how they would build the system.
Frameworks, models and vector databases filled the answer. The missing part was the reasoning: users, evidence, permissions, failure, cost and trade-offs.
Read the field note →
20 / 20 · AI business clarity
I thought AI’s biggest advantage was speed. I was looking at the visible part.
Faster drafts and summaries were useful, but the real change came when scattered information became decisions, owners, risks and next actions people could understand.
Read the field note →