newsfilter.io
Conference Presentation, Keynote

Building the Next Generation of Software | Des Traynor, Fin | RAISE Summit 2026

RAISE SummitDes Traynor
Watch on YouTube →

Intercom/Finn Transformation Overview

  • Context of Crisis (August 2022): Intercom faced collapsing revenue growth, imminent debt, and the existential threat of AI rendering its core SaaS customer support seat model obsolete.
  • Strategic Pivot: The company executed a total overhaul, changing the CEO, strategy, product, sales model, pricing, culture, values, and engineering processes.
  • Current Financial Status:
    • Total revenue now exceeds $400 million.
    • Revenue specifically from the new AI product, "Finn," is near $100 million.
  • Acquisition News: Intercom (rebranded as Finn) has entered into an agreement to be acquired by Salesforce.
  • Core Philosophy: Survival required treating the new venture as a "restart" of the entire company; partial changes were deemed insufficient.
  • Market Warning: Such a radical pivot inevitably results in dissatisfaction among some customers, employees, shareholders, and the media.

Pillar 1: AI Changes What You Build

  • Product Irrelevance: AI renders many traditional software categories irrelevant; products must evolve into new, highly relevant versions or disappear.
  • Market Consolidation: The era of "point solutions" is ending; products are converging into all-encompassing agentic systems, forcing companies to redraw product boundaries.
    • Strategic Scoping: Product scope must be carefully defined to avoid competing with dominant platforms (e.g., a meeting agent should not attempt to build its own Slack).
    • Risk of Over-scoping: Expanding beyond a specific buyer type or system access level creates dangerous competitive friction.
  • UI Paradigm Shift:
    • From UI to Intent: Traditional command-line-to-UI evolution is being superseded by "Chat as UI," where users express desired outcomes via text prompts rather than navigating menus.
    • Dynamic Interfaces: Software now renders dynamic UI based on user queries (e.g., a chat input asking for sales performance data) rather than presenting static dashboards.
    • Design Goal: The objective is to minimize user input while maximizing automatic execution and reliability.
  • Future Product Architecture:
    • Products will evolve into "systems of strategy" rather than "systems of UI."
    • Key success metrics will shift from feature count to the reliability and speed of the AI's core logic.

Pillar 2: AI Changes How You Build

  • Reliability Challenge: AI is probabilistic; chaining multiple components (each 95% reliable) results in a total system reliability of ~85%, which is insufficient for mission-critical tasks like expense tracking.
    • Solution: Companies must aggressively decompose problems to ensure component-level reliability before productizing.
  • Development Methodology Overhaul:
    • Old Workflow: Pick a problem -> Design solution -> Build -> Ship.
    • New Workflow: Identify what AI can make reliably -> Productize -> Design -> Ship.
    • Empirical Rigor: Building now requires scientific rigor, causal analysis, and empirical evaluation (A/B testing) to verify reliability.
  • Design Iteration Speed:
    • Engineers and designers can rapidly explore 10+ design directions to find viable paths before refinement, a speed impossible in the pre-AI era.
  • Invisible AI: Major product improvements can be achieved through backend AI logic without any changes to the visible user interface.

Pillar 3: AI Changes Company Momentum

  • Productivity Surge:
    • Engineering velocity tripled across the organization.
    • Defect backlog was halved.
    • The volume of product changes released to customers doubled.
  • Workforce Evolution:
    • Designers: All designers now write code, own the frontend, and "vibe code" roadmaps for customer validation before engineering begins.
    • Engineers: Mandatory AI usage increased individual performance by 3x; top engineers now use "swarms" of AI agents rather than manual coding.
  • Quality and Cost Metrics:
    • 94% of source code is now written by the AI model "Claude."
    • 20% of Pull Requests are reviewed and approved by AI.
    • Despite a 130k/week spend on cloud tokens, the cost per feature shipped is decreasing.
    • Code quality improved with fewer breaking changes.
  • New Bottlenecks: Rapid AI-driven development has created a new constraint in marketing and sales capacity to promote the accelerated output.
  • Competitive Differentiation:
    • Feature parity is achieved in minutes via AI scraping and prompting; long-term competitive advantage now relies solely on execution speed and the ability to pivot quickly.
    • "Yesterday's unique features are today's copy-paste prompts."

Forward-Looking Statements & Resources

  • Prediction: Speed of adaptation (changing one's mind) and speed of execution are the only remaining defensible differentiators in the AI era.
  • Resource Availability: Detailed engineering case studies and AI research are published at ideas.finn.ai and finn.ai/research.
  • Strategic Advice: If a company is not seeing changes in what they build, how they build (reliability/agency), and their momentum (speed), they are not implementing AI "hard enough."