July 17, 2026
CCW 2026: How AAA Started its Agentic Journey With Riptide
David Dungay recently spoke to Cody Siens, Product Manager, Intake, AI and Automation at AAA Auto Club Group, and Doug Marinaro, CEO of Riptide at CCW to find out how AAA started its agentic journey and why they started with the complex end of the customer interaction.
Most organisations pick the easy stuff first. Password resets, order status, the queries that sit close to the bottom of the complexity curve and carry very little consequence if they go wrong. It is a sensible way to build confidence, and it is the reason so many programmes stall the moment they run out of easy stuff.
AAA Auto Club Group went the other way. It started with roadside: 30,000 service events a day, 13 states, 14 million members, and a customer who is standing on a hard shoulder having what is quite possibly one of the worst afternoons of their year. If there is a harder place to begin, it is difficult to picture.
Sitting down with both sides of the partnership at CCW in Las Vegas, the logic became clearer — and rather more interesting than another containment story.
Advocacy, not automation
Riptide CEO Doug Marinaro frames the distinction sharply. In a conventional deployment, the system talks to the customer, the customer talks back, and when the system reaches the edge of what it knows, it either gives up or invents something. Either way, you get an escalation.
The advocacy model inverts that. When Ripley — Riptide’s advocate — does not know the answer, it works out who does, and asks them. It generates questions into the organisation and opens a multi-party conversation with everyone needed to close the issue, staying with the member the whole way through.
“What if you could get every member their own customer advocate?” as Marinaro put it — one representing them inside the business, carrying their questions and concerns to the people who can actually resolve them.
Siens’ first reaction, by his own admission, was defensive. Can they actually do it? What changed his mind was seeing that this was not a handoff mechanism. The advocate meets the member at the front door and never leaves. Not a switchboard operator — something closer to a colleague who always knows who to phone, and who is still there when the member thinks of something else an hour later.
The data problem, sidestepped
The dominant story in the industry right now is the payback gap. Deployments are not returning what was promised, and fragmented data is the usual culprit. Roadside is about as fragmented as landscapes get.
Marinaro’s argument is that the “clean your data first” doctrine describes a problem that can never actually be finished. The data keeps changing. A great deal of it lives in people’s heads, and much of it is situational — where is the technician right now, how long until they reach me, what is the state of my vehicle.
Advocacy routes around it. The advocate does not need every answer, because it knows who to ask. It does not need to perform every task, because the person still responsible for that task can perform it. At AAA, Ripley sits deeply inside Salesforce and refreshes the same information any agent would see every 15 seconds — but for everything that never lived in Salesforce, and every task the club chose not to automate on day one, it simply asks a human. As more is connected, the advocate contains more and advocates less, and the curve improves from there.
The result was a working proof of concept inside 30 days. Siens describes containment and deflection arriving almost immediately, and a chat with a director on the day of launch that produced a handful of prompt adjustments — worth another five or six percentage points of containment that same night.
The numbers
AAA also ran the sequence in the right order: internal team first, in a safe environment, then out to members. Marinaro’s favourite proof point is what happened when it was briefly switched off during the internal phase. Everybody started screaming. Where is my Ripley?
Asked to reduce the partnership to a single figure, the two picked different ends of the same story.
Siens’ leading indicator is 3—3x concurrency. One call to one person becomes one advocate holding three conversations at once, which he identifies as the sweet spot rather than the ceiling. His lagging indicator is a 400-basis-point improvement in customer satisfaction, landing within the first week or two and repeating club after club.
Marinaro offered a metric Riptide has coined for this: the Advocacy Index — the share of interactions that never require human-to-human contact. At Auto Club Group, it runs consistently at 95%. Only one roadside issue in twenty reaches the point where a person decides they need to call the member directly. And when the advocate does escalate, it escalates into that 3x concurrency rather than into a queue.
Do members actually like it? Siens says yes, and the reason is unglamorous: immediate acknowledgement. Someone is on the way. Once the panic subsides, the questions start — and occasionally a request for a joke, which Ripley obliges before steering the conversation back to the service.
What the pair would tell CX leaders
Marinaro’s advice starts with motive. If the reason for deploying is cost, the biggest opportunity has already been missed. Satisfaction is the primary driver of profitability and share, and the technology should be focused on the experience, not just the cost line.
Siens’ is about the overrated obstacle. Walk the keynote halls, and you will hear, repeatedly, how hard this is and how long the data work takes. His experience says otherwise: you still need your business partners and IT, and you still need the workforce to understand this is help rather than replacement — but the advocacy approach let them go to market faster than the received wisdom suggested was possible.
Which is perhaps the more uncomfortable finding for anyone still queuing up their low-hanging fruit. Starting at the hard end was not the risk. Waiting was.
