Decagon

Speech-to-speech models sound amazing in theory…until they’re deployed at enterprise scale. S2S models bypass explicit text generation, which means less latency and more expressive interactions. But in real-world deployments, following workflows precisely, executing actions correctly, and scaling without breaking matter more than just how it sounds. That’s where speech-to-speech models fall short today. Here’s why: 🔴 Reliability — S2S models often skip instruction steps or behave unpredictably, especially when workflows are complex. 🔴 Accuracy — Without an intermediate text layer, it’s hard to validate outputs or catch hallucinations without adding latency. 🔴 Cost — S2S models process every slice of audio, not just words, which makes them more expensive at scale. Speech-to-speech is exciting technology. But for CX at the enterprise level, reliability and auditability are non-negotiable. Read more on S2S models in our blog below. ⬇️

Huge congratulations to ŌURA on raising $900M in Series E funding at an $11B valuation! 👏 From sleep tracking to holistic health insights, ŌURA continues to lead the way in proactive health and personal data innovation. For them, delivering member interactions that feel as personal as their product experience is essential. We’re proud to partner with them on delivering intelligence and care in every customer interaction!

We’re building our founding Sales team in New York! 🗽 We’re looking for Enterprise Account Executives to join our NYC team. If you’re a top-performing seller with a builder’s mindset and want to help define the future of CX with AI, we’d love to talk. 📍 In-person role based in New York 🔗 Submit your application at the link below. Tag someone you think would be a great fit!

We’re heading to Money20/20 Las Vegas! 🎰 This year, we’re bringing something new: the Decagon Theater, an in-booth stage for conversations that move the industry forward. We’re inviting our community, partners, and customers to join us in exploring how AI is redefining customer experience in financial services. We’re curating sessions that explore: 🔹 Emerging trends shaping CX in fintech 🔹 Real-world case studies and customer stories 🔹 How AI can transform service into a true differentiator Attending Money20/20? Request a demo below. ⬇️

We just hosted our first new hire mixer in San Francisco, and it was a reminder of just how fast the Decagon team is growing! If you’re interested in joining one of the most driven teams in AI, let’s talk. We’re hiring for Strategy & Operations, Agent PM. This role is all about building from 0 → 1: you’ll join our Agent Product Management team to stand up new functions, improve how we serve customers, and create systems that scale. We’re especially excited to meet folks who have: ⚡ Startup or high-growth experience, with a track record of spinning up new initiatives from scratch 🤝 A collaborative mindset and ability to partner across functions 📈 Curiosity to dive deep into product and customer needs to find leverage points Application link is in the comments below. ⬇️

How do you build AI agents that homeowners actually trust? That was the question facing Valon, which is building the AI-native operating system for regulated services. With call volumes surging at the start of each month and the stakes of every conversation incredibly high, they needed AI that could be fast, accurate, and compliant. With Decagon, Valon designed voice agents built for trust. From precise number handling to regulatory-grade accuracy, every detail was tuned to eliminate errors and avoid the “robotic IVR” experience. Valon has been able to: ☀️ Achieve ~90% CSAT, proving AI can drive efficiency without sacrificing customer experience ☀️ Deflect 50%+ of complex inbound calls, even while onboarding 80,000 loans in a single month ☀️ Keep average answer times under one minute, even during call spikes 2–3x higher than normal Together, we are showing what’s possible when innovation and regulation work hand in hand. Read more about the partnership below. ⬇️

Most enterprise AI pilots fail to scale — but it doesn't have to be that way. At DevDay, Jesse Zhang from Decagon and Varun Anand from Clay told a packed audience how to design AI systems that actually work for the enterprise. Watch the full reply on YouTube (🔗 link in comments) and check out the five lessons they shared with our moderator Kimberly Tan from Andreessen Horowitz: 1️⃣ Build proof, not promise. “If we get to the end of a pilot and people are unclear what actually happened or how much money they’re going to save or make, then it’s going to be a really tough sale. If they can’t quantify it, it’s a tough business case.” – Jesse 2️⃣ Start with data, not features. “We start with data as the wedge when we sell into enterprises. It’s measurable — you can run a data test and determine who’s better. Once we win on that basis, we use Clay to automate the workflows that follow.” – Varun 3️⃣ Guardrails are non-negotiable — but they should be user-owned. “We created what we call AOPs — agent operating procedures. They’re kind of like SOPs, but for AI. That way, non-technical people can set and customize their own guardrails safely.” – Jesse 4️⃣ Don’t scale GTM faster than your product deserves. “You can really paper over real problems in the business that product should be solving with people. As we scale go-to-market, we’re trying to scale engineering one-to-one.” – Varun 5️⃣ Build from curiosity, not consensus. “The most important thing is to not over-index on anyone’s advice. Figure out where your competitive advantages are, what you’re strong at and what you’re weak at, and go from there.” - Jesse Full DevDay AMA replay in comments ↓

What a way to kick off SF tech week! At OpenAI’s DevDay, we hit a major milestone: one trillion tokens processed with OpenAI models. Those tokens represent millions of real customer interactions powered by our AI agents every day. It was a great moment seeing our Co-Founder and CTO, Ashwin Sreenivas, recognized during the announcement. Our Co-Founder and CEO, Jesse Zhang, also joined Varun Anand, Co-Founder of Clay, and Kimberly Tan from Andreessen Horowitz for the opening keynote. Together, they explored how enterprises are adopting AI at scale and creating lasting impact. Thanks OpenAI for an incredible start to the week and an exciting moment for our team! Watch Jesse's full keynote in the comments below. ⬇️

Decagon is heading to London for Customer Engagement Summit 2025! 🇬🇧 Our Co-Founder and CEO, Jesse Zhang, will be taking the stage at 12:10 with Arrive’s Product and CX leaders, Alex Keynes and Hanna Fredriksson, to share how Arrive went from evaluation to global rollout of AI support agents. Jesse will also be hosting a Decagon Roundtable at 10:45 in the Mezzanine: "Redefining Customer Experience: A Buyer’s Guide to Agentic AI." If you’ll be there, we’d love to meet you! Book an onsite demo at the link below. ⬇️