Navigating the New Era of Artificial Intelligence in the Workplace

In just a few years, AI has gone from a specialist topic to something everyone has opinions on. It has been praised as miraculous, condemned as dangerous, and debated everywhere from boardrooms to dinner tables. But between these poles of euphoria and dread lies the reality most of us now face: a technology powerful enough to reshape how work gets done, yet still deeply dependent on human judgment.

Too often, headlines focus on extremes, but the practical questions are far more grounded. How do we manage this technology? How do we and our teams work alongside it effectively? How do we capture its benefits without compromising our values or our goals? To find useful answers, we have to start by understanding where we are, what has truly changed, and what has not.

Artificial intelligence, after all, is not new. It has been with us since at least 1956, when a small group of computer scientists gathered at Dartmouth College to explore a deceptively simple question: Can machines think? Since then, the technology has seen decades of progress and setbacks, so-called AI summers and winters.

Today’s systems are powerful and raise new challenges, and we will take those challenges seriously throughout this book. But they are not as alien or unmanageable as they are sometimes made out to be. We already have decades of lessons and frameworks from earlier generations of AI and from other complex technologies, from aviation to automobiles to power plants, all of which can be adapted to this moment.

Understanding what’s the same and what’s different about this moment for AI and work, and how we can build on what we already know, is crucial. If you’re new to AI, I will offer you a map of essential concepts so you can navigate confidently. If you’ve been here for a while, I will reframe the challenge, moving the conversation toward leadership and collaboration rather than technical mastery alone.

After all, how we choose to work with AI, and who we’ll become in the process, is something we still get to decide.

For decades, AI was used behind the scenes, embedded in models that (for example) predicted customer churn or flagged fraud. Those systems were specialists, usually trained for one narrow task and confined to it. Today’s AI models are generalists. These “foundation models” are vast neural networks trained on oceans of data and capable of being adapted across contexts. The same model that helps a developer write code can also be harnessed to help a marketer write copy or an HR leader write a job description.

A second shift is that AI now speaks our language, literally. We no longer need to code or click through rigid menus to interact with AI. We can simply use everyday language, and the responses come back in polished, humanlike prose, making the technology broadly accessible. These systems can also generate new content, including text, images, and beyond—which is why this wave is often called “generative AI.” But that ease of use creates a new kind of business responsibility. We must now learn when to trust AI’s output and when to challenge or shape it.

The third change, and perhaps the most profound, is agency. AI no longer just analyzes or predicts. Now it acts. “Agentic AI” can plan steps toward a goal, call for the right tools, check its own work, and keep going from there. A single AI agent can draft an email, open a ticket, schedule a delivery, and log the transaction, all without a human clicking “send.” This can bring immense productivity potential but also raises new questions. How much autonomy should we give AI? What does accountability look like when decisions are distributed across humans and machines?

Put these shifts together, and it makes for potentially confusing implications for how work gets done. If AI can “reason” through tasks and carry out steps without a human at every turn (and if its outputs look and sound like a colleague’s message rather than robotic communication) the line can start to blur between using AI as a tool and AI being part of the team.

But just like previous waves of AI, generative AI (which creates new content like text, images, or code) and agentic AI (which can take actions independently to complete multistep tasks) are good at some things and bad at others.

An AI agent can draft a sophisticated legal memo. But it can also confidently misread a contract’s indemnity clause. It can write working code, yet stumble on a small, unspoken requirement a junior engineer would catch. Harvard’s Fabrizio Dell’Acqua (with his colleagues) describes this as the “jagged technological frontier” of AI.