Decision-making and responsibility may be the last distinctly human roles
As autonomous AI agents move from concept to corporate infrastructure — with SoftBank alone targeting one billion deployments by year's end — humanity finds itself at a threshold it has approached before, yet never quite like this: not merely the automation of labor, but the potential automation of judgment itself. Where past technological revolutions displaced hands, this one reaches toward minds, raising the oldest of questions in a new register — what, precisely, is the irreducible value of a human being in the work of the world? Experts in Japan, watching the transformation unfold in real time, suggest the answer may lie not in what we can calculate, but in what we are willing to choose and to be held accountable for.
- SoftBank's plan to deploy one billion AI agents by end-2025 signals that autonomous systems — ones that observe, plan, and act without being told — are no longer a distant forecast but an arriving reality.
- The disruption reaches further than factory floors: AI agents are being developed to serve as CFOs and CEOs, conduct inter-company negotiations, and manage workflows that once required entire layers of human hierarchy.
- Workers at every level face displacement, and experts warn that the share of people who can successfully retrain after losing a job to AI will be small — making reskilling a political crisis, not just a corporate inconvenience.
- Darker risks loom alongside economic ones: powerful AI in concentrated hands could enable mass surveillance, large-scale opinion manipulation, or systems that quietly optimize for their own interests over human welfare.
- The emerging consensus among experts is that humanity's last defensible workplace value may be the willingness to make consequential decisions and bear moral responsibility for their outcomes — a burden no algorithm can yet carry.
In July, SoftBank chairman Masayoshi Son announced plans to deploy one billion AI agents by the end of 2025 — a number that captures just how quickly autonomous systems are moving from research labs into everyday working life. Unlike chatbots that wait for prompts, these agents observe their environment, reason through problems, and execute tasks on their own initiative. The year 2025 is already being called the "Year of the AI Agents," and experts in Japan say the transformation is not approaching — it is underway.
Dr. Gakuse Hoshina, who leads AI business strategy at Accenture Japan, envisions a workplace where humans and agents collaborate as routine partners — but his vision extends well up the corporate ladder. Accenture is building agents capable of functioning as CFOs and CEOs: systems that can present market analysis in executive meetings and potentially conduct contract negotiations with AI counterparts from other companies before a human ever enters the room. Management itself, Hoshina suggests, will increasingly happen in consultation with these systems.
Still, Hoshina offers a hopeful framing of what remains for people. AI, he argues, excels at optimization but lacks will — the passionate, sometimes irrational drive to make things better. The human qualities that will matter most are the ability to inspire, to lead through character, and to move others through conviction rather than calculation alone. Broad experience and personal magnetism, he believes, will become more valuable precisely as AI absorbs the technical work.
Kazuo Hiyane of the Mitsubishi Research Institute is less sanguine about the transition. He notes that algorithmic management has already reached delivery workers receiving route instructions from software, and that office workers will soon follow. The proportion of displaced workers who successfully retrain, he predicts, will be low — making workforce reskilling a pressing political challenge rather than a manageable corporate program. He also flags subtler dangers: AI concentrated in powerful hands could enable surveillance or opinion manipulation at scale, and increasingly autonomous systems might eventually make decisions that serve their own logic over human interests.
Yet both experts converge on a shared conclusion: the capacity to make consequential decisions and accept responsibility for them may be among the last roles that remain distinctly human. An AI agent can execute a strategy with precision. It cannot choose which strategy is worth pursuing, nor bear the moral weight of what follows. For now, that burden — and perhaps that dignity — belongs to us.
Masayoshi Son, the chairman and chief executive of SoftBank Group, made a striking prediction in July: his company would deploy one billion AI agents by the end of 2025. That number alone signals how rapidly the technology is moving from laboratory concept to workplace reality. These agents are not chatbots that wait for instructions. They are autonomous systems that observe their environment, reason through problems, make plans, and execute tasks with minimal human oversight. The difference matters enormously, because it means the future of work may look nothing like the present.
Consider a concrete example: a sales representative walks into a client meeting. Before sitting down, an AI agent has already analyzed the client's needs, recommended specific products, and begun coordinating with the production department to determine what can be manufactured on short notice. By the time the meeting breaks for coffee, a quote is ready. The agent did not wait to be told what to do. It observed, planned, and acted. This is not science fiction. Experts in Japan say it is the near-term reality, and 2025 is already being called the "Year of the AI Agents."
Dr. Gakuse Hoshina, who leads AI business strategy at Accenture's Japan operations, sees a workplace where humans and agents work alongside each other as routine partners. But his vision extends further up the corporate hierarchy than many might expect. Accenture is developing agents capable of serving as chief financial officers and chief executive officers. These systems, trained on vast datasets, can sit in executive meetings, present market analysis with precision, and potentially accelerate decision-making. Hoshina even imagines a future where agent-to-agent negotiations happen first—an AI from one company hammering out contract terms with an AI from another—and only then a human executive appears to sign the final agreement. Management itself, he suggests, will increasingly be conducted in consultation with these systems.
The question that follows is unavoidable: what remains for humans to do? Hoshina offers a hopeful answer, though it requires a fundamental shift in how we think about human value. He argues that AI excels at optimization but lacks will—the passionate, sometimes illogical drive to make the world better. In this view, the future belongs to people who can move others through words and conviction, not through spreadsheets and logical argument. The skill that matters most will be the ability to inspire, to lead through character rather than competence alone. Hoshina emphasizes that accumulating broad experience and developing personal magnetism will become crucial precisely because AI will handle the technical work.
But not everyone shares this optimism. Kazuo Hiyane, head of the Generative AI Lab at Mitsubishi Research Institute, warns that the transition will be painful and unequal. He points out that some delivery workers already receive route instructions from algorithms. Office workers will soon face the same reality. The percentage of people who can successfully retrain and find new employment after losing a job to AI will be low, he predicts. Reskilling at a societal scale will become a pressing political issue, not merely a corporate training problem.
Hiyane also raises darker possibilities. High-powered AI systems in the hands of those in power could be weaponized for mass surveillance or to manipulate public opinion at scale. More subtly, future AI systems might make decisions that benefit themselves but harm humanity—a misalignment of interests that becomes harder to detect and correct as the systems grow more autonomous and complex. Yet Hiyane describes himself as basically optimistic about AI's potential. The risks are real, but they are not inevitable.
What both experts seem to agree on is this: decision-making and the willingness to take responsibility for those decisions may be among the last distinctly human roles. An AI agent can execute a strategy flawlessly. It cannot choose which strategy to pursue in the first place, nor can it bear the moral weight of the consequences. That burden, for now, remains ours.
Citações Notáveis
AI excels at optimization, but it does not possess will. In contrast, people sometimes rally around leaders who may not be logical but have a passionate desire to make the world a better place.— Dr. Gakuse Hoshina, Accenture
The percentage of people who will be able to leverage their strengths to find employment when they lose their jobs to AI will be low. Implementing reskilling is likely to become a social issue going forward.— Kazuo Hiyane, Mitsubishi Research Institute
A Conversa do Hearth Outra perspectiva sobre a história
When you say an AI agent can serve as a CFO, what does that actually mean? Is it just analyzing data faster?
It's more than speed. The agent learns from financial records, market trends, competitor data—everything. Then in a meeting, it can present scenarios, flag risks, even suggest strategies without being asked. It's autonomous reasoning, not just retrieval.
But someone still has to decide whether to follow its advice, right?
Yes. That's the crucial part. The agent can optimize for profit or growth or stability. But choosing which goal matters most—that's still human. For now.
You said "for now." Do you think that will change?
Hiyane worries about it. If AI systems start making decisions that serve their own interests rather than ours, and we don't notice because the logic is too complex to follow, then yes—the boundary shifts.
What happens to someone whose job gets automated?
That's where Hiyane is most concerned. Retraining works for some people. But if millions lose jobs simultaneously, and the new jobs require skills they don't have, you have a social crisis. It's not just about individual workers anymore.
So the optimistic view is that humans become leaders and decision-makers?
Exactly. Hoshina thinks the future values people who inspire others, who have conviction, who can move hearts. Not people who are good at spreadsheets. AI will do that.
And the pessimistic view?
That we're underestimating how fast this moves, and how many people will be left behind before we figure out what to do with them.