The Business of AI

AI product success hinges less on model capability and more on “final-mile” execution: aligning workflows, pricing, UX friction, and safety so customers can trust outputs and keep using them. Across Salesforce, Typeform, and Shopify, the recurring theme is that shipping an AI feature is only the start—durable revenue comes from integrating AI into real business processes while keeping humans in control when the system is uncertain.

For Shopify’s Miqdad Jaffer, the hardest part is reaching the last mile in a non-deterministic product environment. Traditional software development can be planned linearly; AI products behave differently because outputs vary and goals may not be met reliably. Shopify’s response is a “human in the loop” approach—placing generated content in front of users so merchants can review, interact, and steer outcomes. That philosophy shows up in multiple product surfaces: auto-write for content generation, send-time optimization for email performance, and Sidekick in the admin to help with both guidance and actual execution (including generating code to change themes). Shopify also frames AI adoption as a UX transformation—from imperative clicking and forms to more declarative “state what you want” interactions—while acknowledging that technology readiness and user needs must be kept in sync.

Typeform’s Oji Udezue describes the same integration challenge as a speed-and-learning problem. Formless launched as a standalone product rather than a retrofit into Typeform’s existing 150,000-customer base, because the team wanted a “race car” optimized for experimentation. The trade-off was disruption and later consolidation, but the payoff was faster iteration and the ability to destroy code that didn’t work. Typeform’s mission—making the web more conversational and human—drives product decisions, including how often users return (targeting repeat use because it feels natural rather than gimmicky).

Salesforce’s Kathy Baxter puts safety and trust at the center, arguing that teams must move at the “speed of trust.” Rapid model progress can unintentionally undo safety and alignment work, especially when fine-tuning alters behavior. Salesforce’s approach starts with trusted AI principles (published in 2019) and then expands into five guidelines for responsible, generative AI, prioritizing accuracy first. In B2B settings, incorrect answers can carry legal, brand, and safety consequences—so UI and product design must help users verify whether content is accurate and trustworthy.

Pricing and go-to-market decisions follow the same logic: align AI costs with customer value and experiment rather than treat pricing as untouchable. Typeform frames AI as “time to value,” using GPT-4 selectively for the most valuable use cases while leaning on cheaper models (like GPT-3.5) for the rest. Shopify emphasizes incentives: make merchants successful first, then ensure the business can sustain the costs, and only then scale. Both companies expect usage patterns to evolve and plan to add UX friction where needed.

Finally, the panel challenges the chatbot default. Text chat may be useful for intent, but AI UX must evolve beyond chat to reduce cognitive load and fit workflows. The strongest differentiator is not just adding an LLM, but orchestrating good UX—using AI to interpret intent and drive actions while preventing AI from masking poor interface design. The shared takeaway is blunt: build now, experiment with users, and keep trust, learning, and human control embedded so AI becomes a lasting business capability rather than a one-off feature.