AI is a multiplier
In strong systems, AI accelerates value, learning, and delivery. In weak systems, it accelerates defects, risk, and fragility.
Enterprise AI adoption starts with more than access to tools. It requires AI literacy, disciplined delivery, mature DevOps practices, clear governance, verification, value stream thinking, AI economics, human accountability, and the resilience to recover when things go wrong. Autonomy is earned across all of it.
This is a system conversation, not a tooling conversation.
But discipline is not the same as caution. The organizations that win will push AI leverage hard. They will simply insist on proof before they trust it with more. Increased autonomy should be earned through evidence, not granted on ambition.
AI autonomy must be earned, not declared.
AI will not make slow systems competitive. It will make the gap between fast-learning organizations and slow-moving ones impossible to ignore.
AI is changing the pace of competition
The question is no longer whether AI can improve individual productivity. The real question is whether the organization can turn new ideas into safe, valuable outcomes fast enough to compete.
As more digital value moves toward agentic, API-driven, and headless experiences, companies with long build cycles, slow feedback loops, unclear priorities, and fragile delivery systems will struggle to keep up. Competing in AI requires more than adding AI features to existing products. It requires an operating model that can sense disruption, make decisions quickly, experiment safely, deliver reliably, and learn from the market.
The competitive advantage is not just AI adoption. It is the speed and discipline of the system that turns AI-enabled ideas into measurable customer and business value.
Where I stand
In late 2024 and throughout 2025, much of the industry conversation around AI focused on adoption rates, usage metrics, and individual productivity gains. My focus was elsewhere: on the delivery system itself, how work flows from idea to production, how teams maintain operational resilience, and whether organizations were building the discipline to safely absorb acceleration.
That perspective is gaining traction now. More leaders are recognizing that AI adoption is a system conversation. I’ve been making that case since the beginning.
What I am seeing now
Two shifts are becoming clear.
First, software is moving toward more agentic experiences: agents, APIs, automation, and headless workflows that help customers get work done with less friction.
Second, the role of senior engineers is expanding. AI can reduce task effort, but it does not remove technical judgment, product context, quality thinking, operational ownership, or accountability. Experienced engineers become more important because they know how to guide, review, orchestrate, and validate what AI helps create.
That is why AI literacy has to move beyond tool usage. Teams need tool literacy, system literacy, and responsibility literacy.
Responsibility literacy also means being explicit about how humans and AI actually work together. In practice, AI shows up in one of a few roles: as a tool that executes a bounded task, an assistant that recommends while a human decides, a collaborator that iterates with a person, a monitored agent that works within limits and escalates, or an autonomous agent that acts end to end in a narrow domain. Naming the role for each use case is how you keep ownership clear as autonomy grows.
AI may shrink parts of the team, but it cannot shrink the work of judgment, learning, validation, and ownership.
The goal is not the smallest team. The goal is the smallest sustainable system.
The one-minute version
AI is reshaping how code is written and how technology organizations think about the full system that delivers software.
DevOps is the foundation
To truly leverage what AI can do, organizations have to get exceptionally good at DevOps so code can move from idea to production quickly, reliably, consistently, and securely.
AI reveals truth
AI reveals the truth about your delivery system. It exposes the gaps between what organizations say they do and what actually happens when code needs to reach production.
Think beyond code generation
If organizations apply AI only to code generation while ignoring the surrounding system, they risk achieving local gains while causing broader dysfunction.
AI adoption starts with literacy
The question for most companies is no longer whether AI matters. It does. The question is how to begin adopting it practically and responsibly.
That starts with exposure, hands-on use, and shared learning. Teams need time to understand the tools, compare options, and build judgment through real experience.
Teams build AI judgment through real experience and shared learning.
But fluency alone is not enough. Without engineering depth, judgment, and systems thinking, AI adoption gets shallow fast. Organizations still need strong fundamentals to make AI durable at enterprise scale.
Leadership still matters
The old model, which overvalued the technical leader who could personally grind out the most code, was already limited. In an AI-enabled environment, it becomes even less useful.
What matters more now is the ability to build strong teams, create repeatable systems, develop experienced people, and lead through change.
This work needs builders, not only managers: people who can design the technical, commercial, governance, and operating architecture that lets humans and agents create measurable value together.
Experienced engineers become force multipliers when paired with AI. Their value is in what they can shape, guide, review, and orchestrate across increasingly AI-enabled workflows.
The manager’s role is expanding
As delivery teams shrink and AI agents take on more of the task work across product, design, coding, quality, architecture, and security, the managers supporting those teams need to evolve alongside them.
That evolution runs in two directions.
The first is scope. Engineering managers now need enough fluency to understand how agents are built, how they are used, and how they affect team capability and delivery flow. Budgeting is changing too: the cost conversation is no longer only about direct report salaries but also about the tooling, access, and infrastructure that agent-supported teams require. Outcome visibility remains non-negotiable. Smaller teams with more AI assistance do not reduce the manager’s responsibility to understand what is being delivered and whether it is creating value.
The second is depth. The work being delivered today remains highly technical. Services, agentic connections, design patterns, and system interactions still require deep systems thinking and technical judgment. Managers who want to lead effectively in this environment may need to broaden toward V-shaped skills: deepening their understanding not just of people leadership but of product thinking, engineering architecture, and the technical nature of the systems their teams are building and operating.
The question for senior engineering leaders is whether they step up into that expanded role or remain anchored to a model of people management designed for a different kind of team.
Humans remain accountable for what AI helps create.
The leaders who create the most value with AI will be the ones who understand people, systems, flow, and change.
AI across the value stream
I say this not only as a technology leader, but as a technologist and active user of AI. I use it. I see how it can accelerate thinking, improve analysis, and increase the leverage of experienced professionals.
But I also see the temptation to confuse more activity with more value. More AI-generated work does not automatically improve outcomes. Without strong systems and disciplined ways of working, AI can create the appearance of progress while increasing waste and risk underneath.
Guardrails are the structure that enables speed.
There is also a cost to all of this that often goes unmeasured. AI is becoming part of product economics, not just a tooling line item. Token spend, model choice, agent infrastructure, and tool sprawl all show up in margin, cost to serve, and product profitability. The teams that win will not chase the lowest AI spend. They will manage for the best cost per outcome.
The economic goal is not lower AI spend. It is better cost per outcome.
AI is a value stream conversation. If enterprises want the full benefit, they have to think beyond build speed and look at the full path from idea to outcome.
AI, operations, and the human loop
Much of today’s AI discussion centers on whether engineers will be replaced or teams will become much smaller. Some of that may happen, especially if organizations optimize the system that moves software from idea to production.
But software delivery does not end at production. When the enterprise system fails, who gets the call? Who understands the architecture, the dependencies, and the safest path to recovery?
AI may increasingly help diagnose incidents and accelerate response. But at an enterprise level, operating and maintaining production systems still requires human judgment, accountability, and deep contextual knowledge.
Speed without operational resilience is just risk in a different form.
Software does not stop at deployment. Enterprise responsibility includes operating, maintaining, restoring, and improving systems in production.
How I approach AI adoption
My approach is practical, and it starts with a simple rule: match the level of AI autonomy to the level of consequence.
Match autonomy to risk
Not all work deserves the same level of AI independence. Low-risk work can move fast with light review. High-risk work, security, billing, compliance, permissions, production, keeps clear human accountability and stronger checks. Know what data can be used, what cannot leave the environment, and what requires a human in the loop before autonomy expands.
Choose tools with intent
Different use cases require different tools, trust levels, and constraints.
Build literacy through real use
Hands-on learning, shared practices, and room for experimentation.
Integrate into the delivery system
Better requirements, stronger tests, safer reviews, faster debugging.
Make verification the new bottleneck
When AI can generate faster than people can review, verification and validation become the constraint that matters. Build the checks for correctness, security, regressions, and operational impact directly into the workflow, sized to the risk of the work.
Measure the system
System-level outcomes matter more than individual output metrics.
The real question
I am optimistic about AI. I have been using it since February 2023 and have seen significant gains in the recent wave of frontier models and agents. I can already see how it is reshaping how modern organizations work.
But I am also practical. AI will not rescue a broken operating model. It will reveal it. The organizations that benefit most will be the ones with the strongest systems, clearest priorities, healthiest teams, and most disciplined path from idea to outcome.
Is your system strong enough for AI to make you better, or will it simply make your existing problems move faster?
Let’s talk
If you’re navigating AI adoption at the enterprise level, I’m happy to compare notes.