How to Transform Customer Experience at Scale ft. Jeffrey Bussgang & Teresa Anania

27 August 2025 | EPISODE 145 | 42 MIN | A Ticket Is Already a Failure. Here's What to Do About It.

[Un]Churned

May 07, 2026

In this episode, a Harvard Business School lecturer and VC meets a CCO who's actually running the AI playbook at 600,000 customers. The theory lands a lot harder when someone in the room is living it.

Listen on YouTube, Spotify and Apple Podcasts

Three Things We’re Taking Away

Start internal before external, always. Teresa’s team built agent assist for renewals reps, support engineers, and CSMs before touching digital agents for customers. The crawl/walk/run sequence matters: internal use cases build trust, surface data gaps, and prove the model before you put it in front of a customer.
The intelligent front door reframes the whole problem. Stop thinking about AI for support tickets and start thinking about all inbound customer demand as one routing problem. Teresa’s team redirected renewals quoting, support deflection, and community search under one framework. A ticket created is already a failure.
Imperfect data is not a reason to wait. Every company Teresa has worked at (Autodesk, Zendesk, Sophos) had messy data. The AI use cases they deployed improved the data while increasing productivity. The motion to clean your data and the motion to get value from AI are the same motion.

🎯 The [Un]Churned Take: The Crawl/Walk/Run Playbook for AI-Native CX

Jeff Bussgang, co-founder of Flybridge Capital and lecturer at Harvard Business School, and Teresa Anania, Chief Customer Officer at Sophos, bring a rare combination to this conversation: one person who has seen AI transformation across hundreds of companies, and one person currently doing it inside a 600,000-customer cybersecurity organization. Together, they make the clearest case we’ve heard for what the first year of serious AI adoption actually looks like.

The Pilot Ghetto Is a Real Place

Jeff has a name for what happens to most enterprise AI experiments: they end up in the pilot ghetto. Companies run dozens of experiments in parallel, generate learnings, and then stall out because the path to production requires budget, organizational buy-in, and a willingness to trust early-stage vendors that most large enterprises aren’t culturally built for. “All the innovation is coming from the startups,” he said. “If they wait forever for Salesforce and Oracle and Workday to bring their AI A-game to the table, they’re going to be waiting forever.” The companies pulling ahead are the ones willing to take a controlled risk on less mature vendors rather than waiting for the safe choice to catch up.

Fear Is Data, Not an Obstacle

Teresa ran an AMA with her team specifically on AI. The questions were predictable: what happens to hiring, what happens to my job, are we doing this at the expense of the customer? Her answer wasn’t to dismiss the fear. It was to bring actual agents into the session to describe what their job felt like now compared to before. “The before and the after. When you can show them that they can handle 10 customers but go deeper than they were ever able to go per day, that resonates. Because they took a job because they want to have more customer engagement, less administrative research in the background.”

The CSAT, tNPS, average handling time, and number of handoffs all improved after Sophos deployed their first AI use cases. Her team’s fear that AI would erode the customer experience turned out to be wrong in precisely the opposite direction. “Customers don’t want to have to always talk to a human.”

Rethink the Front Door, Not the Ticket Queue

Teresa’s most useful reframe was about where AI gets applied. Most CX leaders start with support tickets because it’s the obvious use case. Her team started there too, but the bigger unlock came from zooming out: what if you treated all inbound customer demand as a single routing problem? Community searches, portal visits, renewals quoting, support requests. “A ticket creation is already a failure because they didn’t find what they need.” An intelligent front door catches demand before it becomes a ticket, routes it to the right resource (human or agent), and improves the data with every interaction.

The renewals quoting agent is a concrete example. Reps used to search across six systems to build a quote. Now they type what they need and the agent builds it. The result: at least 25% improvement in productivity and headcount savings from attrition that aren’t being backfilled in reactive roles, reinvested into proactive ones.

QA at Scale Is the Underrated Win

Jeff and Teresa both landed on a point that doesn’t get enough airtime: AI’s ability to do quality assurance across every customer interaction, not just the sampled one in a hundred that traditional QA could reach. Teresa described being able to surface customer sentiment in real time, before a call and after, so that frustrated customers get identified before the conversation goes sideways rather than after. “What I love about AI is it allows us to do QA at scale. You don’t have to wait for the lagging indicator.”

Wrapping Up

The CCOs winning on AI right now aren’t the ones with the cleanest data or the biggest budgets. They’re the ones who stopped treating AI as a support ticket problem and started treating the entire customer experience as one routing challenge worth solving.

See you next week 🧠

Where to find Jeffrey Bussgang:

LinkedIn: https://www.linkedin.com/in/bussgang/

Where to find Teresa Anania:

LinkedIn: https://www.linkedin.com/in/teresa-anania/

Referenced in This Episode

The Experimentation Machine — Jeff Bussgang’s book on leveraging AI tools alongside timeless product-market fit principles. Written for founders but relevant for any executive navigating AI adoption. Available on Amazon
Flybridge Capital Partners — Jeff’s seed-stage VC firm, focused heavily on AI software companies. flybridge.com
Sophos — Cybersecurity company where Teresa served as CCO, currently serving 600,000 customers. Teresa’s team is one year into their AI transformation. sophos.com
Brighthire — AI-native interview intelligence tool Jeff referenced; tracks whether companies are actually hiring for the AI-native skills they claim to want. brighthire.ai
Replit — AI coding platform Jeff referenced as a tool for executives to vibe code prototypes and demos without deep engineering backgrounds. replit.com
Lovable — AI app builder Jeff recommended for quickly creating MVPs and art-of-the-possible demos. lovable.dev