Cresta's CEO advocates for a single AI platform that both assists human agents and powers full automation. This creates a powerful feedback loop: when an AI agent fails, the system observes the human's successful resolution, capturing data to improve the next AI agent iteration.

When faced with 1,000 support emails daily and a 12-person team, StackBlitz integrated Parahelp, an AI support tool. The AI agent handled 90% of tickets automatically, allowing the company to manage hyper-growth without hiring a 50-100 person support team, thus avoiding associated complexity and cost.

Vlad Tenev outlines a maturity model for customer support AI. Phase 1: Answering questions from a help center (inform). Phase 2: Reading customer data for context (read). Phase 3: Taking actions on an account, like issuing refunds (act). Most companies are stuck in Phase 1, but the real value lies in Phase 3.

Hostinger uses its AI agent, Cody, to resolve over 83% of customer support conversations. This isn't just about cost savings; it's a strategy to provide faster, more accurate solutions for repetitive issues like DNS problems. This frees up human agents to focus exclusively on high-value, complex edge cases that require deeper expertise.

Companies adopt AI not to reduce headcount but to address the chronic shortage of skilled customer service advisors. AI handles mundane tasks like password resets, allowing humans to focus on high-value interactions and act as brand ambassadors, ultimately elevating their roles.

The goal of AI in customer service isn't human replacement. Instead, use AI agents to handle predictable, repetitive queries instantly. This strategy frees up human staff to focus their time on complex, empathetic problem-solving where a personal connection is most valuable.

To navigate regulatory hurdles and build user trust, Robinhood deliberately sequenced its AI rollout. It started by providing curated, factual information (e.g., 'why did a stock move?') before attempting to offer personalized advice or recommendations, which have a much higher legal and ethical bar.

Prioritize using AI to support human agents internally. A co-pilot model equips agents with instant, accurate information, enabling them to resolve complex issues faster and provide a more natural, less-scripted customer experience.

A significant internal application for Meta's newly acquired Manus AI could be automating its notoriously difficult-to-reach customer support. These AI agents could handle common inquiries at scale, providing faster responses and reserving human agents for complex issues.