How Market Leaders Outpace the Competition with Agentic AI.

Lyft, Klarna, and Google's secretive X division have already done it.

They use agentic AI instead of reading about it. Now it's your turn.

Autonomous, capable of reasoning, planning, and acting. They aren't chatbots answering FAQs. They're emerging as a new class of employees. As productivity demands rise and human workers are overwhelmed by interruptions, market leaders are quietly building AI-native teams. Will you?

This piece unpacks five trends embraced by agentic leaders. Each chapter shows how leading firms are using agents today—and what that means for your organisation tomorrow. After exploring the trends, we'll return to the big picture: how to design for exponential capability.

Trend 1 – AI agents as a new class of employees

Market leaders don't just use agents. They onboard them. Lyft's February 2025 announcement barely made a ripple outside the tech press, yet it signalled a massive shift. By integrating Anthropic's Claude into its customer-care platform, the ride-sharing firm slashed average support resolution times by 87%. The AI "colleague" fields thousands of inquiries per day, hands off complex cases to humans, and is being trained across other departments. Whether they give agents names (Uber does) or keep them anonymous (Lyft hasn't decided), the intent is clear. Agents are not side projects but core staff.

Klarna shows how far this can go. In a single month, their OpenAI‑powered assistant handled 2.3 million conversations, equivalent to the workload of 700 full‑time employees. Customers now resolve issues in under two minutes, compared to 11 minutes before. Across 35 languages and 23 markets. The company's Kiki assistant is used daily by 87% of employees to draft contracts, check press sentiment, or research competitors. Klarna's CEO even boasted that the firm saved about $40 million a year and stopped hiring humans—before walking back the statement. The lesson? Treating agents as colleagues—with defined roles, escalation paths, and targets—builds trust and frees people to focus on judgment and empathy. How many "colleagues" like this have you hired?

Trend 2 – The future is today

The cost of intelligence is collapsing. Between November 2022 and October 2024, the price to run GPT‑3.5‑class models fell from roughly $20 to $0.07 per million tokens. Nearly a 300-fold drop. Nvidia says its latest chips deliver a 100,000× reduction in energy needed to generate tokens compared with a decade ago. That's why agentic leaders design for capability that doesn't yet exist.

PayPal, for instance, is building a Financial Operating System for AI agents that allows agents on both sides of a transaction to browse, buy, negotiate, and refund without human intervention. With more than 430 million users and $1.5 trillion in annual payment volume, the payment giant is positioning itself as the infrastructure layer for agent‑to‑agent commerce. Are you designing for the world that exists today—or the one that will exist in three years? When intelligence is this cheap and energy so abundant, 50×–100× capability gains aren't hype. They're baseline.

Trend 3 – Everyone is a target

Agentic automation isn't just for repetitive tasks. Darktrace's Cyber AI Analyst uses layered AI to question data, test hypotheses, and draw conclusions, autonomously triaging millions of security alerts. In 2024, it performed 90 million investigations, equivalent to 42 million hours of human analysis, escalating only the most critical incidents to people.

FutureHouse, a nonprofit lab, has launched a platform of superintelligent scientific agents that search and synthesize research across millions of papers. Agents like Crow and Falcon outperform human researchers in literature searches and can propose new experimental directions.

Waymo's fully autonomous ride-hailing service utilizes layered AI agents to perceive, predict, and plan in real-time. By early 2025, its vehicles had completed over 10 million paid rides and were on track to double that by year‑end. In other words, agents are already tackling complex cognitive tasks—cyber investigations, scientific discovery, urban driving—that we once assumed only humans could handle. If your knowledge workers are "safe", think again.

Trend 4 – Culture is king

Agents don't thrive in rigid hierarchies. Bold experimentation, intelligent failure, and psychological safety are prerequisites for success. Google's X exemplifies this. Early‑stage ideas are deliberately tested and killed to make room for genuinely transformative ones. Their Rapid Evaluation group prototypes scores of ideas. Only around 3 % move forward. Once a project reaches the Foundry, the expectation is that half will be killed within a year.

X teams run pre‑mortems to imagine failures in advance and follow Astro Teller's mantra: "tackle the monkey first"—focus on the hardest part before building anything. When a project ends, they say it's been "killed" to acknowledge the grief and move on. That brutal honesty frees capital and attention for bolder bets. Could your board stomach an explicit mandate to kill half of all AI initiatives?

Trend 5 – Become an agent boss

True market leaders aren't just throwing technology at old processes. They're rewiring their organisations around human–agent teams. Microsoft's May 2025 report calls such companies Frontier Firms. In its survey of 31,000 workers across 31 countries, 82% of leaders reported plans to utilize AI agents to expand workforce capacity within the next 12–18 months. Meanwhile, 80% of workers report lacking the time or energy to perform their jobs effectively, and employees are interrupted an average of 275 times per day.

Frontier Firms recognise that the answer isn't more tools, but new structures:

• Phase 1 pairs every employee with an assistant,

• Phase 2 forms human-agent teams, and

• Phase 3 hands over complete workflows to agents with humans supervising.

The vision isn't about replacing humans. It's about turning every employee into a manager of digital labour, building pipelines, governance, and integration layers so that people can delegate tasks, critique AI output, and make judgment calls. Are you ready to coach a machine, not just use one?

Designing for exponential capability

If today's agents feel like clever interns, imagine what they'll become. The AI arms race isn't about who has the biggest model. It's about who builds the most capable agents and the culture to harness them. Market leaders already treat agents as digital colleagues. They redesign work so machines can execute tasks autonomously while humans focus on direction and relationships. With employee burnout and productivity pressures mounting, the choice is stark: either augment your teams with reasoning, acting AI, or watch competitors do it first.

Are your systems ready to accommodate an exponential drop in the cost of intelligence and a surge in capability? Will your employees have the skills to supervise digital labour?

Agents won't replace all of your people. They will replace the tasks that distract. They will create space for strategy, creativity, and human connection.

If you let them.

Your move