AI@Work: 3 new patterns of work define AI-first companies

These days, it would be hard to find anyone who needs to be convinced that AI is changing the future of business-and society at large. For business leaders it's become a question of how quickly you can weave it into the core of your organization. However fast you're doing it, your competition is trying to do it faster.

The share of US employees using AI tools at work-whether it's occasionally, frequently, or daily-has already doubled in two years. That's quite an adoption surge. And the most forward-looking organizations, the Frontier Firms, are moving beyond experimentation to set the tempo and rules of the game for the AI era.

3 new patterns of work for the AI era

Frontier Firms are human-led and agent-operated-buying intelligence like electricity, putting it to work like an employee, and compounding it like interest. These organizations are upending every assumption, rebuilding work from the ground up for human-AI collaboration. And we're already seeing new patterns of work emerge that maximize the value of that collaboration.

Today you can see these patterns most clearly in software development, but we'll soon see them pop up across industries, functions, and businesses of all sizes. Established companies are already applying them to those functions with the highest number of clearly defined and measured processes like sales, service, and finance. And they're showing up in areas where AI has distinctive strengths, like marketing and content creation.

How the patterns come to life
Let's take a look at each pattern, and use software development to understand what they look like in practice.

1. Human + AI assistant
The pattern: An individual pairs with an AI assistant to remove drudgery and accelerate productivity.

In software development: Work starts with AI. The AI assistant suggests code and tests, so developers spend less time on busywork, freeing them to concentrate on design, customer value, and quality-the work that actually moves products forward.

2. Human-agent teams
The pattern: Agents join teams as digital workers for specific steps or workflows, collaborating alongside people to scale impact.
In software development: AI is inserted into existing workflows. Chores-testing new code, reviewing every update, checking compliance-no longer eat up days during a sprint. Agents take the first pass, drafting summaries, flagging issues, and suggesting fixes. Developers make the final calls, but the team moves faster and with less friction.

3. Human-led, agent-operated
The pattern: The optimal Frontier Firm template-redesigning workflows so agents run them end to end. Humans set goals, guardrails, and intervene only when necessary.
In software development: Imagine a release pipeline that runs largely on autopilot. Development starts with goals expressed in plain language that agents turn into drafts and prototypes. This empowers small teams to go from idea to demo in days, accelerating the feedback loop. Agents build, test, deploy, and monitor in well-scoped pipelines with human-defined guardrails; developers focus on what to build next and step in when edge cases arise.

It's not a linear progression-it's a jagged frontier

These patterns don't show up in a linear or universal progression across the organization. Many firms see all three at once, in different corners of the business. A marketing team might lean on assistants to draft campaigns, engineering runs agent-driven testing, finance experiments with fully automated reporting.

But not every pattern isright for every workflow, and how fast each function can move depends on time, budget, and capacity.

The real takeaway is that software development doesn't just add AI on top of old routines-it redesigns its work around it. At a personal productivity level, that means starting every task with AI. At a process level, it means mapping workflows, deciding where AI can plug in, and then refining or even rebuilding those processes end to end. A single product team might use AI to draft code in the morning, collaborate with agents to test it that afternoon, and push a release through an agent-automated pipeline by evening.

They not only coexist, but compound, each pattern reinforcing the others to accelerate scale and impact.

The new patterns we see at Frontier Firms-and more widely in software development-will soon ripple across all work. The choice for leaders now is to either ride the wave-or get swept away.

For more insights on AI and the future of work, subscribe to this newsletter.