23 posts tagged with "AI"

Imagine a finance team waiting for an invoice-dispute workspace: a queue, supporting documents, assignment rules, and a connection to its accounting system. In this illustrative scenario, the feature once consumed weeks of sprint capacity. Now one engineer orchestrates agents that scaffold the implementation, tests, and deployment configuration in hours, using established components and contracts.

That is not an unattended production release. By that afternoon, the engineer and domain owners have reviewed the behavior, validated integration assumptions, and approved and shipped a bounded production pilot. But the change in economics is clear: creating a credible version of a familiar feature requires far less implementation effort. A competitor with similar tools can do the same.

The difficult question is no longer whether someone can produce the queue and its buttons. It is whether the workspace helps people resolve the disputes that keep money, decisions, and customer relationships stuck. When building is no longer the moat, what is?

Picture an engineer reviewing an agent-generated pull request. It changes the payment client, retry configuration, tests, and deployment documentation. The engineer learned to work differently: follow the failing request through logs, reproduce the problem, edit each file, and build confidence one deliberate step at a time. Now the proposed solution is already on the screen.

The first reaction is not necessarily excitement or distrust. It might simply be: "That is not how I used to do this task." In this illustrative scene, the discomfort is not about whether the engineer can read the code. It is about the missing experience of having constructed it. A familiar route to understanding has been shortened, but understanding is still required.

That tension reaches well beyond software. When a tool produces work that used to demonstrate our expertise, we have to separate the method from the responsibility. The useful shift is neither "do everything yourself" nor "let the agent handle everything." It is a harder, more durable principle: embrace the leverage, own the verification.

Agents Do Not Remove Delivery Friction​

AI agents can read a repository, write code, run tests, and open a pull request. That is an impressive capability, but it does not make the surrounding engineering system easier to navigate. It makes every inconsistency in that system easier to reproduce at speed.

An agent given access to a large, undocumented environment faces the same questions a new engineer does: Which template is approved? How is a service deployed? Where do secrets come from? Which checks are mandatory? Who owns this dependency? What is safe to change? A capable model can infer some answers, but inference is not a reliable operating model for production software.

On Friday, July 17, 2026, I delivered a webinar on the Microsoft Reactor YouTube channel called Agentic DevOps Meets Accessibility: Inclusive Delivery with AI & Agents. The recording is now available, so you can watch it anytime.

The Real Shift Is Not Less Learning, It Is Different Learning​

Every time a new technology makes information easier to reach, the same worry resurfaces: will people stop learning altogether? Calculators supposedly meant nobody needed arithmetic. Search engines supposedly meant nobody needed to remember facts. AI now raises the same question, at a much larger scale, because it can explain a concept, draft an essay, and carry out multi-step tasks on its own.

The worry misreads what is actually changing. AI is transforming how quickly people can access information and produce a first draft of an answer. It is not transforming the underlying process by which a human being builds real understanding, develops judgment, or becomes capable of solving problems they have never seen before. That process is still slow, effortful, and deeply human.

Why Understanding Still Matters When AI Can Write the Code​

It has never been easier to turn an idea into working software. Describe what you want in plain language, and an AI tool can generate a web page, a script, an API integration, or even a multi-step agent that takes actions on your behalf. For anyone who felt shut out of software because they did not know how to code, that shift is real and welcome.

But there is a difference between software that runs and software you understand. AI can produce the first one in seconds. Only you can produce the second, and only if you know enough about how software actually works to read what the AI gave you, question it, and fix it when it is wrong.

Why the next chapter of FinOps is not about finding a cheaper model. It is about engineering systems that do not waste the tokens they already have.

A finance leader opens the monthly invoice for the company's AI platform and finds a number that does not match any story anyone can tell. Usage grew modestly. The bill grew sharply. Nobody switched to a pricier model. Nobody approved a new integration that anyone remembers. The line item simply grew on its own, the way cloud bills used to grow before anyone built a discipline around watching them.

Ask the engineering team what happened and the answer is rarely a single cause. It is a hundred small decisions: a system prompt that grew every time someone patched in a new rule, a retrieval step that fetches ten documents when two would do, an agent that retries a failing tool call five times before giving up, a workflow that hands a conversation between three specialized agents and resends the full history at every handoff. None of these decisions looked expensive in isolation. Together, they are the bill.

The Bottleneck Moved, and Most Teams Have Not Noticed​

For decades, writing code was the expensive part. Reading it was almost free by comparison. A developer spent hours producing a change, and a reviewer spent minutes confirming it. That ratio shaped everything: our tools, our processes, our sense of who was busy and who was waiting.

Agents inverted that ratio. A capable agent now produces a complete branch with code, tests, and documentation in the time it takes to write a thoughtful task description. Authoring became cheap. Reading did not.

Eight building blocks that turn a capable model into a system that ships real work: agents, subagents, skills, context, memory, prompts, notebooks, and projects.

Most people believe they are using Generative AI when they open a chat window and type a question. They are using the least interesting part of it. The chat box is the front door, not the house.

On Friday, June 12, 2026, I delivered a webinar on the Microsoft Reactor YouTube channel called DevOps Foundations for the AI Agentic Era. The recording is now available, so you can watch it anytime.