The death of SaaS, the dawn of agents

The AI Imperative: Des Traynor Unveils the Dawn of Agentic Software at Web Summit Lisbon 2025

Des Traynor, Co-founder and Chief Strategy Officer of Intercom, and creator of Finn.ai, addressed Web Summit Lisbon 2025, proclaiming the “death of SaaS” and the dawn of AI agents. He asserted that the AI era is a present reality, demanding an “adapt or die” response from software companies. Successfully navigating this transition promises category leadership and substantial financial rewards, while inaction leads to rapid irrelevance.

Mr. Traynor advised rebuilding existing SaaS products as AI agents or for new ventures to target established, “boring” areas with AI-first solutions. He emphasized the importance of ambitious product definition, noting that while integrating a simple co-pilot is straightforward, transforming entire workflows with a comprehensive agent is more challenging but offers greater potential. Defining clear boundaries for the agent’s scope is crucial to avoid feature creep, considering factors like distinct buyer personas or complex data integrations.

A key challenge lies in understanding the true complexity of the problem domain. Many new entrants in customer service, for example, fail by expecting perfectly structured queries, whereas real-world interactions are often messy and incomplete. Mr. Traynor stressed that agents must be engineered to handle this chaos, respecting the inherent difficulties of the domain to avoid building a product unfit for purpose.

Learning from past AI failures is essential. Previous attempts often faltered due to high setup costs, complex configuration, and poor pattern matching. While generative AI offers new capabilities, understanding these historical shortcomings helps in building robust solutions. The goal is to move beyond simple “button bots” to create genuinely effective and reliable AI agents that address real user needs.

Building agents involves critical trade-offs between agency (the ability to handle diverse tasks) and predictability (adhering to defined actions). Customers desire both maximum capability and absolute reliability, which are often conflicting. Mr. Traynor highlighted the need for careful architectural and UI design to balance these aspects, ensuring the agent is trustworthy, especially for mission-critical business processes. Unreliability, even from highly reliable components, can compound, making robust engineering paramount.

Launch is merely the starting line for an AI agent. Post-launch, the focus shifts to continuous optimization. Customers will require tools to control the agent’s tone, measure its performance, and fine-tune its behavior for their specific business context. This necessitates developing an “optimization engine” that supports ongoing analysis, training, testing, and deployment, enabling customers to perfect the product over time.

The organizational impact of AI agents is profound. Companies must anticipate new roles, such as “DevOps agentic coordinator,” and redefine success metrics, as traditional KPIs become obsolete. Educating customers about these organizational transformations, including changes to budgeting and forecasting, is crucial. Mr. Traynor concluded by affirming the immense value of this transition, citing Finn.ai’s rapid growth as evidence, despite the inherent challenges and chaotic nature of the AI revolution.