Every CX leader is being asked the same question right now: how much of this can AI do? The honest answer is that it's the wrong question. The right one is which tickets, and the answer is different for every ticket type you have.
My AI strategy didn't start with AI. It started at Uber, on the first in-house social media response team, while carrying Incident Response casework at the same time. The job every day was triage: what can be resolved right there on social, what needs to move to a DM, and what has to go to a phone call or an email with a specialist. Three lanes, sorted by how much judgment and how much risk a conversation carries. When AI agents arrived, I already had the framework. The lanes just got new names.
I sort every ticket type into one of three buckets.
01AI, end to end
These are the tickets where the customer already knows what they want and the answer lives in a system. Cancel my subscription. Where is my order. What's your return window. There's no judgment call, there's a lookup and an action, and a well-built agent does both faster and more consistently than a person at 11pm on a Sunday.
This bucket is where the automation numbers come from, and it's larger than most teams think. The catch is that "end to end" has to mean end to end: the agent needs the tools to actually cancel the subscription, not just explain how. An AI that can only answer questions about cancellation is a help center with a personality, and customers can tell.
The other thing this bucket needs is good source material. The agent learns from what you give it, and the fastest way to a reliable one is to point it at a wiki that's already clean: SOPs written as steps, with the approved wording embedded, and public policies that say what the internal ones say. The platforms make this easy, you can upload the SOPs or connect the knowledge base directly. What they can't do is fix documentation that was never clear to begin with. That work comes before any of this.
It's also where the business case lives. These are the most common tickets in any DTC queue, and they arrive at every hour. Handling them with AI gives customers 24/7 coverage on the things they ask about most, and it lets the human team stay the size it should be as the brand grows, instead of hiring in step with ticket volume.
02AI starts, a human finishes
This is the bucket where the design matters most. The split is simple: AI does the collecting, a human does the deciding. A customer writes in about an order that arrived damaged. The agent confirms the order, asks for a photo, checks the timing, and writes up a clean summary. Then a person looks at it, decides, and sends the replacement. The human touch is spent on the part that needs judgment, not on the intake form.
Done well, this bucket cuts handling time more than the first one does, because the human part of these tickets goes from ten minutes to one.
03Humans only
Someone had a reaction to a product. Someone is describing a safety issue. Someone is upset in a way that could become a brand moment. These tickets are a small fraction of volume and a large fraction of risk, and AI should never be the first thing that touches them.
Not because the model would say something terrible. Usually it wouldn't. But "usually" is the wrong standard for an adverse event, and the thirty seconds of friendly automated acknowledgment before a human gets involved is thirty seconds the brand can't get back. This bucket needs hard routing that pulls the ticket out of the AI flow before any reply is generated, and a human-owned SLA that is faster than the AI's.
Start where the wins are, and let the harder buckets earn their way in.
How I roll it out
Bucket one goes live first. Buckets two and three skip AI entirely on day one, routed straight to people. That gets the easy, high-volume wins immediately, with zero risk to the conversations that matter most, and it gives the team a few weeks to QA the agent on the simplest tickets before it's trusted with anything else.
Once the QA holds, bucket two ramps in: the agent starts gathering information on replacement-type tickets and hands them off. Bucket three never ramps. It skips AI from day one and keeps skipping it.
I've done this twice with Siena AI, for a functional beverage brand and a wellness supplement brand during my time consulting at Create CX. Both reached 40 to 65% ticket automation within three months of implementation, and both got there by sorting first and automating second. Before that, at Juice Beauty, the same thinking applied to Gorgias automation took self-service from 0 to 16% against a 12% benchmark. The number falls out of the mix of tickets a brand actually has. What you control is the sorting, and the sorting is the strategy.
