What Happened
The artificial intelligence industry is experiencing a fundamental shift from “assistive” AI systems that suggest actions to “agentic” AI systems that can autonomously make decisions and execute complex multi-step workflows. According to recent industry research, 23% of organizations are already scaling agentic AI systems in production, with an additional 39% actively experimenting with these autonomous agents.
In a significant endorsement of this trend, Microsoft and NVIDIA announced the launch of an “Agentic Launchpad” in the UK and Ireland in November 2025. The initiative aims to support startups building autonomous AI systems that can plan and execute tasks without constant human oversight, signaling that major tech companies view agentic AI as the next frontier of artificial intelligence development.
This evolution represents a paradigm shift in how AI integrates into business operations. Rather than serving as sophisticated autocomplete or recommendation engines, agentic AI systems can independently research topics, make decisions based on changing conditions, and execute multi-step plans—moving from tools that assist humans to systems that can operate autonomously within defined parameters.
Key Details
- Adoption Rate: 23% of organizations scaling agentic AI systems; 39% experimenting
- Microsoft/NVIDIA Initiative: Agentic Launchpad launched in UK and Ireland for AI startups
- Key Capability: AI agents can plan and execute multi-step workflows autonomously
- Market Timing: Shift occurring as AI infrastructure matures and costs decrease
Why It Matters
For Consumers
This technology shifts the software interaction paradigm. It turns tools into teammates. Instead of asking an AI assistant for suggestions and then manually executing tasks, consumers will increasingly delegate entire workflows to autonomous agents. Imagine telling an AI agent “plan my vacation to Japan” and having it autonomously research destinations, compare flight and hotel options, create an itinerary, make bookings, and handle changes, all without requiring step-by-step approval.
Early examples are already emerging. TripAdvisor is using ChatGPT to generate personalized travel itineraries, while Instacart has launched AI tools that help families save money through smart cart recommendations and real-time pricing optimization. These represent the first wave of consumer-facing agentic AI, with far more sophisticated applications expected in 2026.
For the Industry
The shift to agentic AI is creating new competitive dynamics across the tech sector. Companies that can build reliable autonomous agents will have a significant advantage over those stuck in the “assistive AI” paradigm. This is why Microsoft and NVIDIA are investing in startup accelerators. They recognize that breakthrough agentic AI applications may come from nimble startups rather than established players.
Substantial technical hurdles remain. Agentic AI requires low-latency inference, massive context windows, and robust “guardrails” to prevent autonomous hallucinations. It also demands the ability to reason with incomplete information. These are capabilities that go far beyond current large language models. This is driving innovation in areas like “context engineering” (optimizing how AI systems integrate into workflows) and real-time information retrieval.
For AI infrastructure providers, the shift means different computational requirements. Agentic systems need to run continuously, pull real-time data, and make rapid decisions. This creates demand for different types of computing resources than the batch processing that dominates current AI workloads.
For Businesses
The 23% of organizations already scaling agentic AI systems are gaining significant competitive advantages. These autonomous agents can handle tasks like customer service escalations, supply chain optimization, and financial analysis without human intervention, dramatically reducing operational costs while improving response times.
However, the 39% still in the experimentation phase face a critical decision: move quickly to production or risk falling behind competitors who successfully deploy autonomous agents. McKinsey’s research suggests that organizations scaling AI programs across operations see measurably better results than those stuck in pilot purgatory.
The risk calculus is complex. Agentic AI systems that make autonomous decisions can also make autonomous mistakes. Organizations need robust governance frameworks, clear boundaries for agent authority, and fail-safes to prevent costly errors. This is why many companies are proceeding cautiously despite the technology’s promise.
The Backstory
Autonomous agents have existed in theory for decades. However, only recently has the technology matured enough for practical deployment. The breakthrough came with large language models like GPT-4 and Gemini, which demonstrated reasoning capabilities sophisticated enough to handle multi-step tasks.
However, early LLMs were fundamentally reactive; they responded to prompts but couldn’t initiate actions or adapt to changing conditions. The shift to agentic AI required additional innovations: the ability to use tools, access real-time information, maintain context across long interactions, and make decisions based on goals rather than just responding to inputs.
OpenAI’s release of GPT-5.1 in November 2025, with both “Instant” and “Thinking” versions that can independently decide when deeper reasoning is needed, exemplifies this evolution. Google’s Gemini 3 similarly introduced enhanced reasoning capabilities that can “reliably complete 10 to 15 coherent logical steps” without human oversight.
The Microsoft/NVIDIA Agentic Launchpad builds on this technological foundation, providing startups with the computational resources and expertise needed to build production-ready autonomous agents.
Expert Reactions
Industry analysts at McKinsey emphasized the scale of the shift:
“Agentic AI represents the most significant evolution in artificial intelligence since the transformer architecture breakthrough. Organizations that successfully scale these systems will operate fundamentally differently than their competitors, with autonomous agents handling entire workflows that currently require teams of knowledge workers.”
Technology strategists at ThoughtWorks highlighted the infrastructure implications:
“The move to agentic AI is driving unprecedented changes in the tech stack. These systems need continuous access to computational resources, real-time data pipelines, and sophisticated error handling. These requirements are forcing a complete rethinking of AI infrastructure.”
What’s Next
The agentic AI market is poised for explosive growth as the technology matures and more organizations move from experimentation to production deployment.
Timeline:
- Q4 2025: Microsoft/NVIDIA Agentic Launchpad begins accepting startup applications
- Q1 2026: First wave of production agentic AI applications expected from accelerator participants
- Mid-2026: Industry analysts predict 40%+ of organizations will be scaling agentic systems
- 2027: Agentic AI expected to become standard feature in enterprise software platforms
Key developments to watch include how organizations handle governance and safety for autonomous agents, whether startups or incumbents lead in agentic AI innovation, and how quickly the technology moves from narrow task automation to broader autonomous decision-making.
The World Health Organization’s recent call for legal safeguards for AI in healthcare highlights the regulatory challenges ahead. As AI systems gain autonomy, questions about liability, transparency, and human oversight will become increasingly urgent.
The Outlook
The shift from assistive to agentic AI is not just an incremental improvement. It is a fundamental reimagining of what AI systems can do. The 23% of organizations already scaling these systems are gaining competitive advantages that will be difficult for laggards to overcome.
What’s particularly significant about the Microsoft/NVIDIA Agentic Launchpad is the signal it sends: the tech giants believe autonomous agents are the next major platform shift. They are investing in the startup ecosystem because they recognize that breakthrough applications often come from companies unburdened by legacy systems.
However, the technology is still early. The 39% of organizations experimenting with agentic AI are discovering that building reliable autonomous agents is significantly harder than deploying assistive AI. The systems need to handle edge cases, recover gracefully from errors, and make decisions that align with organizational goals—challenges that require sophisticated engineering and careful governance.
The organizations that will succeed are those that treat agentic AI as a strategic transformation, not just a technology deployment. This means rethinking workflows, establishing clear boundaries for agent autonomy, and building the infrastructure to support continuous AI operations.
The Bottom Line
This shift from assistive AI that suggests actions to agentic AI that autonomously executes complex workflows represents a fundamental transformation in how businesses operate. While challenges around governance, safety, and reliability remain, the competitive advantages for early movers are substantial.
🦋 Discussion on Bluesky
Discuss on Bluesky