Conference Presentation, Keynote
Des Traynor, Co-Founder & CSO @ Intercom & Creator of Fin.ai: The Death of SaaS, The Dawn of Agents
- Agents are predicted to be the most significant technological shift in history, creating an immediate "adapt or die" imperative for B2B SaaS companies where failure to build competitive agent products results in business extinction within approximately two years.
- The market is characterized by rapid growth, with Intercom's specific agent business projected to grow 5x to 4x year-over-year against a backdrop of hundreds of millions in existing revenue, while the broader customer service transformation is deemed inevitable and occurring soon.
- The strategic scope of agents aims to expand from narrow tasks like meeting notes to executing entire organizational functions such as all support, marketing, or payments, though expansion is advised to cease when encountering different buyers, permissioning levels, brand identities, or established competitors like Slack.
- Success in this category depends on overcoming historical failures of previous AI systems, such as the inability to handle unstructured data, phatic replies, user intent changes, and the lack of empathy in messy real-world interactions, requiring bespoke architectures rather than simple RAG bots.
- A fundamental trade-off exists between the high agency and adaptability of generative AI and the necessary predictability and control required for production environments, where daisy-chained workflow steps can significantly degrade overall reliability and where risks include agents acting erratically or causing financial loss.
- Building robust multi-agent architectures is technically more demanding than simple "thin wrapper" solutions, yet they offer superior competitive moats; however, progress often stalls after initial marketing or fundraising efforts due to the difficulty of moving from occasional functionality to live deployment.
- Companies must fundamentally restructure their operations to accommodate agents, including redesigning software architecture, pricing models, sales strategies, and core metrics, while simultaneously educating users on how an AI-driven workforce operates and developing analytics to monitor agent performance.
- The current landscape is described as the most volatile and difficult period in two decades of software sales, driven by ambitious visions that may outpace deliverable products, as seen in historical examples where major tech figures warned against the pitfalls of product convergence.