Most of what enterprises deployed between 2022 and 2024 was the former — tools that answer questions, generate content, summarise documents. Useful, but fundamentally reactive. Agentic AI is something structurally different: systems that plan, reason, and execute multi-step tasks autonomously, without continuous human instruction at each step.

That distinction matters enormously for anyone building or investing in B2B technology right now.

The adoption numbers are striking in their speed. According to Gartner, fewer than 5% of enterprise applications embedded agent capabilities in 2025. By the end of 2026, that figure is projected to reach 40% — one of the steepest adoption curves in enterprise software history. KPMG’s AI Quarterly Pulse Survey tracked agent deployment surging from 11% of organisations in Q1 2025 to 42% by Q3 — a nearly fourfold jump in under six months. These are not pilot numbers. These are production deployments.

On the investment side, the trajectory is equally clear. Agentic AI startups raised approximately $1.3 billion in 2023, $3.8 billion in 2024, and an estimated $6.5–7 billion across 2025 — roughly doubling year on year. The capital is not flowing toward LLM infrastructure. It is flowing toward companies that have built actual products on top of it — vertical solutions with real enterprise customers and measurable outcomes.

That last point is where the investment thesis gets specific. The market is not rewarding builders who can demonstrate a capable model demo. The hard part was never the model — system complexity has become the primary deployment bottleneck, with multi-agent orchestration, reliability, and traceability now surpassing all other challenges as organisations move from prototypes to production. The companies that will capture durable value are those that have solved the integration and governance layer, not just the intelligence layer.

For B2B founders, this creates a clearer brief than the generative AI wave did. The question is not “how do we add AI to our product?” It is: which workflows in our domain involve multi-step decision-making that currently requires constant human oversight — and what would it take to make those autonomous? Risk assessment, compliance monitoring, communications orchestration, customer workflow management: these are the categories where agentic architectures are already delivering returns above cost of capital, and where defensible positions can be built.

Gartner’s longer-term projection suggests agentic AI could generate close to 30% of enterprise application software revenue by 2035, surpassing $450 billion. That is a large number. More relevant for builders today is the near-term window: the companies that move from prototype to production in the next 18 months will be setting the category defaults that later entrants will have to displace.

At Forthtech, this is not an abstract observation. Our portfolio’s focus on AI and deep learning applications means we have been watching the applied layer of this space closely — and the shift from generative to agentic is the most consequential change we have seen since we began investing. The builders who understand what agentic really means, beyond the terminology, are the ones worth backing. And we are proud to say that we are working with visionary founders in that premise.