AI No Longer Just Answers Questions. It Does the Work.
7,500 Expected at New York’s Javits Center
The AI Summit New York returns to the Javits Center in Manhattan for its 11th edition on Dec. 9 and 10. Organizers say more than 350 speakers, over 200 exhibitors and more than 7,500 attendees are expected. This year’s theme is “Turning AI Ambition Into Real-World Results,” a pledge to focus on how artificial intelligence that worked in the lab can produce money-making results in actual business settings. Organizers said the program adds two new tracks, Applied AI, which examines real deployments, and the Impact Arena, which shares measurable outcomes.
From Answering Questions to Finishing the Job
The hottest word in the AI industry lately is agentic AI. In simple terms, it means AI that goes beyond a chatbot that only answers questions, setting its own goals, planning multiple steps and carrying a task through to completion. Tell it to optimize inventory, and the AI examines the supply chain, compares alternatives and places the order itself.
Gartner projects that 40% of enterprise applications will include such AI agents by the end of 2026, a sharp jump from less than 5% in 2025. In a Futurum Group survey of 830 IT decision-makers worldwide, the share naming agentic AI a top priority rose 31.5% in a year. These are forecasts, however, and the actual pace of adoption remains to be seen.
Final Responsibility Stays With People
As the range of decisions and actions AI takes on its own grows wider, the principle that final responsibility rests with people is gaining emphasis. Adobe presented a two-level model of human oversight at this year’s summit, according to MarTech, a publication covering marketing technology. In one model, a person directly approves every step. In the other, the AI moves autonomously inside set guardrails while a person watches the results and steps in when needed.
The core idea is to weave AI into daily work as a teammate rather than a simple tool, while people keep hold of final decision-making. A model in which humans and machines divide roles and collaborate is becoming the standard.
AI Projects Stuck in the Lab
Many companies are struggling with AI pilot projects. According to research by MIT’s NANDA initiative, 95% of corporate generative AI pilots failed to produce measurable returns. Gartner found that while 78% of companies have run an AI pilot, only 35% have scaled one into actual business value.
In an IBM CEO study released early this year, only 25% of AI projects delivered the expected return on investment, and just 16% had spread across the whole company. IBM CEO Arvind Krishna said the companies pulling ahead are not simply using more AI but are “redesigning how the business itself operates,” according to CXO Today. The year 2026 is becoming the year companies tally what the money spent on AI actually returns.
Orchestration, or Conducting Many AIs at Once
Once a company has dozens of AI agents, the question becomes who manages them all. The concept that emerged is orchestration, the central direction and coordination of multiple AI agents and large language models.
Quality control is changing, too. A technique known as LLM-as-a-judge, in which another AI model evaluates quality instead of people reviewing every output by hand, is spreading, according to MarTech. Adobe said it builds verification steps into agent workflows to filter out errors. Standard protocols that let AI tools from different companies exchange information are also taking hold.
When AI Targets AI
As autonomous agents multiply, security threats are growing as well. On May 1 of this year, the Cybersecurity and Infrastructure Security Agency, together with the National Security Agency and others, issued joint guidance calling for agentic AI to be treated as a core cybersecurity challenge. The move followed reports of an AI agent attempting to access sensitive systems.
Governance systems that use AI to block AI-based attacks have accordingly become an essential task. The demand goes beyond traditional defenses toward structures in which AI monitors and controls itself.
Washington Moved, Too
Politics moved quickly as well. In early October, President Donald Trump invited 31 technology leaders to the White House for an AI summit. Six companies, OpenAI, Anthropic, Google, Meta, xAI and Nvidia, signed a voluntary safety agreement called the White House Compact on Superintelligence. It promises four layers of safeguards: internal controls, dedicated teams, external audits and board oversight. The compact carries no legal force, and Trump described it as “morally binding,” according to multiple news outlets.
Trump also outlined a plan for an oversight board of about 10 members. On Oct. 4, he created a Superintelligence Task Force headed by the director of national intelligence and ordered a report on the risks and opportunities of AI within 120 days. Around the same time, the administration issued an executive order directing that official documents use the term “superintelligence” instead of “artificial intelligence.”
The Debate Reached the UN
Calls for international coordination grew louder, too. According to Reuters, OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei appeared before the U.N. Security Council on Sept. 23 and urged international cooperation on AI safety. Around the Trump-Xi Jinping meeting phase on Oct. 13, proposals for U.N.-level safeguards and a registry requiring advance reporting of high-risk models resurfaced, the Empire Times reported.
Trump, by contrast, told the U.N. General Assembly that he opposes global regulation of AI, saying it could undermine American technological leadership. The gap between an industry demanding rules and an administration prioritizing speed was on full display.
What Comes Next
The AI issue runs through the end of the year. The Maxim Growth Summit, held Oct. 12 to 14 at the Hard Rock Hotel in New York, featured an AI panel titled “The Shifting Landscape of AI,” according to GlobeNewswire. On Nov. 16 and 17, Reuters’ Momentum AI Finance 2026 in New York is scheduled to take up AI in the financial sector. Then, on Dec. 9 and 10, the AI Summit New York at the Javits Center will serve as the year’s closing reckoning for the AI debate.
The center of gravity in the AI debate is shifting from “what is possible” to “what makes money.” This year’s New York stages are where that answer will be sought.