ASAPP

When the recent MIT report revealed that 95% of AI pilots had zero ROI, one thing was clear. The model wasn’t the red flag, but how the model was implemented. In his recent The AI Journal article, Peter Coutoulas breaks down the architectural difference between an AI system that stalls and an AI system that enables productivity. ↓ https://hubs.li/Q04jjJPm0

Next up on our #ASAPPien interviews is Aurora Morrelli, who talks about what surprised her most when she first joined ASAPP almost 5 years ago! ↓

A demo is not a deployment. Salesforce showed Finnair rebooking flights with AI. Six months later, it's still not live. The capability existed in the demo. The operating model didn't support it in reality. The bottle now isn't the AI but whether or not organization are actually built to scale it. Our latest newsletter covers what has to change: redesigning work before layering on AI, what governance looks like as an operating model, why trust matters more than accuracy, and what a real agentic platform looks like in production.↓ From Gina Clarkin, Chris Arnold, Panagiotis Coutoulas, Jeffrey Cochran, Theresa Liao, and Farshad S.

Most enterprise AI deployments look the same six months after launch as they did on day one. Containment flat. CSAT unmoved. Edge cases still surfacing. Farshad S. discusses how ASAPP CXP is built to address the stagnation by orchestrating five AI agents in a closed loop. Click below to see how the flywheel works in full.↓ https://hubs.li/Q04j6Q5k0

🍎 NYC! We are heading to Skift Data & AI Summit 2026 next week.🗽 Be sure to catch Nirmal Mukhi at his session, "How Agentic AI Is Solving the Billion-Dollar Hold Queue", where he'll break down why IVR, chatbots, and ticketing systems never truly resolved the interaction, and what's fundamentally different now. See us on June 3rd. ↓ https://hubs.li/Q04hXhLG0

This past weekend we wrapped up another wonderful meetup with Nokia, Docker, Inc Bengalore, and Collabnix - Docker, Kubernetes and Agentic AI in Bengaluru. 🪷 It was great to see over 250 AI/ML engineers, senior student researchers, and AI-first tech leader/founders gather to talk about AI agents, AI sandboxing, threat modeling, and model serving. Special shoutout to the #ASAPPien team, Alagu Prakalya P, and Navaneetha Krishnan K S, MLE at ASAPP, for his speaker session "Inference optimisation in the era of AI." Hope to see you soon again!

When millions of employees can't get a straight answer on their benefits, who's on the hook? We are thrilled to be partnering with a leading benefits platform serving 800k+ enterprises across corporate America, including Fortune 500 companies. The complexity of benefits is real, and it has to be handled right. Depending on who you are, which part of the org you sit in, or which union you belong to, your benefits response looks completely different. ASAPP's voice agents are now front-lining that experience by navigating enrollment, health insurance, and 401k conversations in real time and tailoring every response to the individual—and keeping human agents in the loop where it matters most. Because when the call is about someone's health coverage or retirement, "close enough" isn't good enough.

Next #ASAPPien up is Xiomara Ruth Sawin, who shares how ASAPP empowers her to take on new challenges and have ownership over her projects. ↓

Most enterprise AI teams treat security like a one-time audit: ship the model, check the box, move on. But adversarial attacks don't stop evolving, and neither should your defenses. Jeff Cochran deep dives into how we built Continuous Red Teaming into Customer Experience Platform (CXP), which conducts testing across 50+ vulnerability classes in real time. https://hubs.li/Q04hqDv00

Your SOC 2 certification doesn’t answer the question your compliance team actually cares about. It answers whether the system is secure. It does not answer what happens when the AI is wrong. Over the past year, I’ve had conversations with CTOs, CIOs, and operations leaders across major carriers, banks, and insurers. And I’ve noticed something interesting: When executives ask whether an AI system is “trustworthy,” they’re usually conflating two completely different things. The first is security. Is the data protected? Is the vendor certified? SOC 2. HIPAA. HITRUST. That problem has a clear answer. It’s documented, auditable, and validated by a third party. The second is accountability. When the AI Agent denies a claim incorrectly, misinterprets a policy, or gives a patient the wrong coverage answer, can you reconstruct exactly why it made that decision? Not just that it happened.The actual reasoning path. Step by step. In a format legal and compliance teams can hand to a regulator. Most enterprise AI deployments only solve for the first category.The vendor has the certifications. Procurement checks the boxes. The deal closes. Then six months later, someone from legal asks: “What exactly did the Agent do here?”. And the answer is usually a dashboard screenshot and a vague explanation no one can defend under scrutiny. Security tells you the vault is locked. Accountability tells you what’s inside it, who touched it, and how the decision was made. Those are not the same thing. I’ve watched enterprises blur that distinction through entire procurement cycles, and it becomes a very expensive problem once AI is operating at F500 contact center scale. The question I’d add to every enterprise AI RFP right now isn’t: “Do you have SOC 2?” It’s: “When your Agent makes a decision, can you show me the reasoning chain that got there?” Very different question. Very different answers. Very different vendors. ASAPP #ASAPPCXP #TrustedCX