Lessons on AI Acceleration — David J Crawford

Over the past few months, I’ve had the privilege of leading our dedicated AI Accelerator Squad. We work in a highly regulated industry where mistakes carry real consequences, so this ain’t no sloppy vibe-code squad. It’s been one of the most transformative leadership experiences of my career — not because we’ve simply adopted new tools, but because we’re fundamentally reimagining what work looks like.

I’ve been working with AI for several years now, primarily from a product design and strategy perspective. But leading a team dedicated entirely to AI acceleration has taught me something unexpected: the leadership skills required to guide agentic AI teams aren’t that different from those needed to lead highly capable human teams. And that realisation has profound implications for how all of us will work in the years ahead.

The Productivity Shift: Beyond What We Imagined

When we started this journey, I expected productivity gains. What I didn’t anticipate was the magnitude of those gains or how they would reshape our understanding of what’s possible.

Our team has experienced productivity shifts that would have seemed impossible just a year ago. Tasks that once took days can now take hours. Processes that required multiple specialists can be orchestrated by a single person with the right AI support.

The Discipline Paradox: You Have to Do Things “Well”

Here’s something that surprised me: working effectively with AI requires more discipline, not less.

In traditional development, you could sometimes get away with shortcuts. Unclear requirements? You’d figure it out as you go. Inconsistent coding standards? The human developer would adapt. Vague documentation? People fill in the gaps with institutional knowledge.

But AI agents don’t work that way. They need clarity. They need structure. They need well-defined interfaces and consistent patterns.

This leads to a massive uplift in our general engineering discipline. We’ve had to:

Sharpen our requirement definitions beyond what we thought necessary
Standardize our processes to levels we’d previously considered aspirational
Document our decisions with unprecedented clarity
Define success criteria before we start, not after we finish

I love the irony here: the technology that promises to make things easier has forced us to become more rigorous. And that rigour will make us better at everything we do, not just AI-accelerated work.

Skill Democratisation: The Great Levelling

Perhaps the most exciting transformation I’ve witnessed is how AI is democratizing skills across our team.

Team members who once felt limited by technical gaps are now operating at levels that would have required years of specialised training. This isn’t about replacing expertise, it’s about elevating everyone’s baseline capabilities. The specialists are still essential, but now they’re focused on solving novel problems rather than routine implementation. As a college of mine said this week: “we’re weening Developers off ‘writing code’ and moving them more towards traditional ‘engineering’ ”.

We’re seeing a clearer separation between conceptualising work and executing work, a distinction that was always there but is now impossible to ignore. This shift is profound, it means knowledge work is evolving into something more valuable, something I like to think of as wisdom work — the ability to frame problems, make judgment calls, and orchestrate complex solutions.

Leadership Lessons: From Micromanagement to Orchestration

Leading individual contributors for many years has taught me one fundamental truth: micromanagement is the enemy of capability. There’s no point having talented people if you’re going to do all the important work for them or hover over their shoulders directing every move.

The leadership approach I’ve developed over the years is simple:

Set clear direction on what success looks like
Establish guardrails that define the boundaries of acceptable solutions
Inspect and adapt at suitable intervals
Trust the process in between

This is “guardrails, not gatekeepers” leadership and it works because it respects people’s autonomy while maintaining accountability.

But here’s what I’ve realised over these past few months: this exact leadership approach is what every individual contributor will need to develop as we move into an agentic future.

The Future: We’re All Becoming Leaders

Think about what’s coming. Individual contributors will still be accountable for outcomes, but increasingly they won’t be directly responsible for all the production work. Teams of AI agents will handle that more efficiently and productively than any single human could hope to be.

This creates a fascinating challenge: how do you lead when your team isn’t human?

The answer, I believe, is the same skills that make great leaders of people:

Clear Direction: AI agents need to understand the goal, not just the task. They need context about what success looks like and why it matters.

Well-Defined Guardrails: Just like human teams need to understand constraints like budget, timeline, quality standards, regulatory requirements, AI agents need boundaries that define acceptable solutions.

Structured Accountability: Companies need well-designed structures with clear lines of accountability. Our future AI-enabled teams will need the same. Realising meaningful value from agentic AI will require teams to formalise structured ways of working, both for collaborating with agents and for coordinating effectively with one another.

Regular Inspection: You can’t just set an AI agent loose and hope for the best. You need checkpoints, reviews, and opportunities to course-correct, just like with human teams.

The Leadership Skill Gap

The uncomfortable truth is, most individual contributors haven’t developed these leadership skills yet. They’ve been focused on personal execution, not team orchestration. They’re experts at doing the work, not at directing others to do it.

But in a world where AI agents handle much of the execution, everyone becomes a leader. Everyone needs to master:

Strategic thinking over effective execution
Outcome definition over task completion
Quality oversight over direct production
Process design over manual operation

The research backs this up. Only 1 in 6 leaders say they’ve fully established strategies to upskill employees as “agent bosses”. We have a massive skill gap to close. techcommunity.microsoft.com.

What We’re Learning in Practice

Start with Outcomes, Not Tasks

When team members first start working with AI agents, they tend to think in terms of tasks: “Write this function,” “Create this report,” “Generate this design.” But effective AI leadership requires thinking in outcomes. Start with a business case and bring the AI in on day zero so it has full contextual understanding of the desired outcome.

Embrace Experimental Mindsets

In regulated industries, we’re trained to minimize risk. But working with AI requires a different balance with constrained experimentation within clear boundaries.

We’ve created a “safe space” where team members can try new approaches, learn from AI interactions, and develop their judgment about when to trust AI output and when to intervene. This psychological safety is essential for building confidence with new ways of working.

Design Workflows, Don’t Just Augment Tasks

The biggest gains aren’t coming from making existing tasks faster, sure we’re doing plenty of that but the biggest wins are from reimagining entire workflows. This requires systems thinking, understanding how work flows through the organisation and where AI can fundamentally restructure the process.

The Competitive Advantage

Organizations that figure this out early will have an enormous advantage. Not because they have better AI tools, everyone will have access to similar technology. The advantage will come from having teams that know how to lead AI effectively.

This is about culture, skills, and leadership models. It’s about developing a workforce that’s comfortable operating at a higher level of abstraction, focused on strategy and orchestration rather than mechanical execution.

In our regulated environment, this advantage is even more pronounced. The teams that can move quickly while maintaining compliance, that can innovate within constraints, and that can scale impact without scaling risk, those teams will dominate their markets.

Preparing for What’s Next

As I reflect on these last few months leading our AI Accelerator Squad, I’m convinced we’re witnessing the early stages of a fundamental shift in how work happens.

The future belongs to organisations that can:

Develop leadership skills across their entire workforce, not just management
Create structures that support hybrid human-AI teams, with clear accountability and coordination
Maintain high standards of discipline and rigor that enable AI to work effectively
Foster cultures of learning and experimentation within appropriate guardrails

This isn’t about technology replacing humans. It’s about humans evolving to work with technology in fundamentally new ways, and that evolution requires leadership skills we’ve previously reserved for managers.

The Call to Action

If you’re an individual contributor today, start developing these leadership muscles now. Practice thinking in outcomes. Learn to set clear direction. Get comfortable with delegation and oversight. These skills will be essential in the agentic future.

If you’re a leader, recognize that your role is shifting. You’re not just leading people anymore, you’re designing systems where humans and AI collaborate effectively. Your job is to create the conditions for success: clarity, structure, safety, and accountability.

And if you’re an organization, invest in building these capabilities across your workforce. The competitive advantage won’t come from having AI, it will come from knowing how to lead it.

Final Thoughts

Leading our AI Accelerator Squad has been humbling and exhilarating in equal measure. We’ve experienced productivity gains that seemed impossible, developed discipline we didn’t know we needed, and unlocked capabilities we didn’t realize we had.

But most importantly, we’ve glimpsed the future of work, and it requires leadership skills from everyone.

The age of agentic AI isn’t about humans becoming less important. It’s about humans becoming more strategic, more creative, and more valuable. It’s about all of us learning to lead.

And that’s a future worth building.

What’s your experience leading or working with AI systems? How are you developing the skills for an agentic future? I’d love to hear your thoughts.

— David J Crawford