Mustafa Suleyman speaks with urgency. The CEO of Microsoft AI sees machines matching and then surpassing human performance on office tasks far sooner than many expect. In recent months he has laid out a timeline that lands like a shock. Most professional work done at computers could be fully automated within 12 to 18 months.
Accountants. Lawyers. Project managers. Marketers. Suleyman names them directly. He does not soften the message. “Most of those tasks will be fully automated by an AI within the next 12 to 18 months.” That line, reported by Fortune in May 2026, set off waves of anxiety and debate across boardrooms and newsrooms alike.
Yet Suleyman is no doomsayer. He frames the coming change as part of something larger. Something he calls humanist superintelligence. The idea rejects the notion of machines that simply outthink humans in every domain without guardrails. Instead he wants systems built to serve people. To remain accountable. To stay in humanity’s corner.
This vision took clearer shape in a November 2025 essay on Microsoft’s own site. Suleyman wrote that if AGI means matching human performance across all tasks, superintelligence goes far beyond it. But direction matters. Microsoft AI pursues advanced capabilities that always work for organizations and individuals rather than replace them. The term humanist superintelligence signals intent. It is not a neutral mountain to climb. It carries a moral stake.
Suleyman’s path to this role traces back through DeepMind and Inflection AI. He joined Microsoft in 2024 to lead its consumer AI efforts. The company later elevated him to run Microsoft AI as a distinct unit. That move freed resources. It also signaled ambition. Microsoft had poured more than $100 billion into its partnership with OpenAI. The original agreement limited how large Microsoft could train its own models and barred certain AGI pursuits.
Then came a contract change about six months before Suleyman’s public comments in mid-2026. “We were only sort of set free from our contract with OpenAI about six months ago to formally pursue superintelligence,” he told audiences. The remark, picked up on X and in tech coverage, marked a turning point. Microsoft no longer sat in a supporting role. It would build frontier systems on its own terms.
At Microsoft Build 2026 the company showed what that freedom produced. Suleyman unveiled seven new models under the MAI brand. MAI Image-2.5, MAI Thinking-1, MAI Voice-2 and others. These span vision, reasoning, transcription and speech. The event, covered by Microsoft’s official channels and outlets like Bloomberg, underlined a push for self-sufficiency. The goal: state-of-the-art performance across text, images and audio by 2027.
Why the rush? Compute. Suleyman argues development will not hit a wall. In an April 2026 interview with MIT Technology Review he broke down three converging forces. Raw chip performance jumped more than sevenfold in six years thanks to Nvidia. Microsoft’s own Maia 200 accelerator delivered 30 percent better performance per dollar. High bandwidth memory stacked chips vertically and tripled data flow. Interconnects like NVLink turned clusters of hundreds of thousands of GPUs into single cognitive units.
The result looks like cognitive abundance. Warehouse-scale supercomputers. Hundred-billion-dollar clusters. Ten-gigawatt power draws. These are no longer hypothetical. Ground is breaking now. Suleyman sees this infrastructure enabling systems that deliver genuine product value at massive scale. He has redefined superintelligence in practical terms. Not necessarily god-like intelligence but models capable of transforming enterprises by the millions.
Critics question the framing. Some see it as moving goal posts when true AGI remains distant. Others worry the speed of deployment outruns safety work. Suleyman acknowledges the tension. He has warned about AI psychosis, where users form delusional attachments to convincing chatbots. Zero evidence exists today of machine consciousness, he stresses. Yet the perception alone can cause real harm. Industry needs stricter standards against overstated claims.
Automation predictions add fuel to those concerns. If white-collar jobs shift this fast, entire professions face restructuring. Suleyman insists 80 percent of existing roles will change rather than vanish. Microsoft research across 50 countries supports that view. The bigger risk sits in the skills gap. Sixty percent of workers in Asia have not received training to collaborate with AI tools. Retraining becomes the urgent priority.
But. The productivity gains already appear. GitHub Copilot users show 55 percent higher output in some coding tasks. The question is who captures that surplus. Companies? Workers? Society at large? Suleyman pushes for systems that augment rather than displace. Human oversight must remain central. Accountability cannot fade.
His earlier book, “The Coming Wave,” written with Michael Bhaskar, explored these dual edges. AI could drive radical abundance. It could solve climate problems through cheap optimization. Yet the same technology might accelerate its own improvement, compress progress timelines and enable misuse from bioweapons to autonomous systems. Containment, he argued, demands global coordination.
That perspective still informs his Microsoft work. Humanist superintelligence demands design choices from the start. Systems should refuse harmful requests. They must explain decisions. They require clear lines of responsibility. Suleyman rejects the idea that superintelligence is both inevitable and automatically desirable. Build only what stays aligned. Prove safety before scaling power.
Microsoft’s strategy blends partnership and independence. The OpenAI relationship continues. Billions flow and models improve. At the same time the company invests in its own stack. Custom silicon. Massive clusters. New model families. The seven MAI releases at Build represent early signals. More will follow. Analysts see this as turning Microsoft from integrator to full competitor in the frontier race against Google, Anthropic and OpenAI itself.
Timelines keep tightening. Suleyman once spoke of AGI in distant terms. Now he points to measurable jumps. Large language models recently cleared 85 percent on the ARC-AGI benchmark, double the score from 2023. Vision capabilities advanced enough in 2025 to reduce hallucinations in context-aware settings. Agents that act on behalf of users moved from prototype to practical tools.
So the conversation shifts. From whether AI will transform work to how fast and who prepares. Executives at Fortune 500 companies now ask different questions in strategy sessions. How do we redesign roles around AI colleagues? What new skills command premiums? Where do humans add judgment that machines cannot yet match?
Suleyman offers no easy comfort. Change this rapid creates dislocation. Yet he sees upside in abundance. Cheaper intelligence means more invention, faster science, broader access to expertise. The humanist qualifier aims to steer that abundance toward shared benefit rather than concentrated power or unintended catastrophe.
Success hinges on execution. Delivering those 2027 frontier models on time. Maintaining safety as capabilities grow. Convincing regulators, employees and the public that alignment efforts are serious. Microsoft’s scale gives it leverage in infrastructure and distribution. Its track record on responsible AI will face fresh scrutiny.
One recent interview captured the balance Suleyman strikes. He described superintelligent systems that must remain tools shaped by human intent. They stay accountable to oversight. They never run away or escape control. That standard sets a high bar. It also defines the race Microsoft now runs in public view.
The coming months will test the claims. New model releases. Pilot deployments in enterprises. Early data on automation uptake. Labor market reactions. If Suleyman’s 18-month horizon holds, 2027 will look markedly different from 2025. Offices filled with AI agents handling routine cognitive labor. Humans focused on strategy, creativity and relationships. Or perhaps something messier. Resistance, adaptation, unexpected breakthroughs.
Either way the direction is set. Microsoft under Suleyman no longer waits for others to define the frontier. It builds toward a specific future. One where intelligence multiplies human potential instead of rendering it obsolete. The bet is enormous. The stakes are higher still.
WebProNews is an iEntry Publication