The Promise and the Reality

Every automation vendor will tell you their tool handles 70-80% of support volume. That's become the standard claim. Teams that get there typically need several months of tuning. And "70% automated" doesn't mean "70% of your problems are solved." Here's how a typical 12-month rollout plays out, using an illustrative team of about 3,000 tickets per month.

Month 1-2: The Optimistic Phase

Teams usually start by automating the obvious stuff. Password resets, order status lookups, "how do I cancel" questions, billing inquiries. These categories can make up about 45% of volume, and the intent classification was straightforward.

Results tend to be good. Response time for these categories can drop from an average of 47 minutes to under 10 seconds. Customers don't complain. Agents are relieved to stop handling the same five questions all day.

Cost savings are immediate. At 3,000 tickets per month and 45% automation, about 1,350 tickets are automated. With usage-based AI pricing in the $0.20 to $0.50 range, that portion costs a few hundred dollars a month, versus roughly $2,700 in agent time for the same tickets handled by humans.

Month 3-4: Expanding and Breaking Things

Encouraged by early results, teams push automation to cover more categories. Refund requests, feature questions, integration troubleshooting. That moves the rate from 45% to 70%.

And this is where things get messy.

The edge case problem

Take a customer who writes in saying "I need to cancel my subscription." The automation classifies it correctly as a cancellation intent and triggers the cancellation flow. Problem: the customer actually wanted to cancel a specific add-on, not their entire subscription. The message was ambiguous, and the system picked the most common interpretation.

At scale, this kind of error repeats. Each one is a customer whose full subscription was cancelled when they only wanted to remove an add-on, and some of them churn because the experience felt careless.

The tone-deaf bot

Or a customer whose payment was double-charged writes in frustrated. The automation correctly classifies it as a billing issue and sends the standard response: "I can see your payment details. Here's how to update your billing information." The customer isn't asking to update anything. They want their money back. The response is technically related to billing but completely misses the emotional context.

The false confidence trap

When automation handles most tickets successfully, the failures become invisible. They get buried in the 30% that goes to humans anyway. A systematic problem with one intent category can go unnoticed for weeks if nobody is reviewing the automated responses that preceded human escalation.

The Fixes That Actually Worked

Confidence thresholds

Stop treating every classification as equally reliable. Messages with high confidence scores (above 95%) get fully automated. Messages between 85-95% get automated with a softer touch: the response includes a "Did this help?" button, and a "no" click routes to a human immediately. Messages below 85% go straight to a human.

This single change can cut false-positive automation substantially.

Sentiment detection before action

For any action with real consequences (cancellation, refund, account changes), add a confirmation step. "I understand you'd like to cancel your subscription. Just to confirm, you want to cancel your full account, not a specific add-on?" One extra message. Dramatic reduction in errors.

Weekly automation reviews

Every Friday, one person spends an hour reviewing a random sample of 50 automated conversations. They flag anything that felt wrong: correct classification but wrong tone, technically accurate but unhelpful, missed nuance. These reviews feed back into the system configuration.

Where a Careful Rollout Lands

After tightening thresholds, the automation rate often settles slightly below the peak, because some categories get pulled back to human handling. CSAT on automated conversations usually trails CSAT on human conversations, but overall CSAT can rise because the speed gain on simple questions outweighs the gap.

In the illustrative example, roughly 2,100 automated tickets a month would cost about $4,200 if handled by humans. Usage-based AI pricing brings that to a fraction of the cost.

What to Do Differently From the Start

Start with confidence thresholds from day one instead of adding them after failures. The "automate everything" approach is tempting because the numbers look great, but the customer experience damage from edge cases takes months to repair.

Exclude high-emotion categories from automation entirely. Billing disputes, complaints about service quality, anything where the customer is already upset. Speed doesn't help when someone's angry. Empathy does, and automation doesn't deliver it convincingly.

Invest in the review process early. A weekly audit should start in month one, not after the first serious incident.

The Real Takeaway

70% automation is achievable. The question is what that 70% looks like. If you automate carefully, with confidence thresholds, confirmation steps for consequential actions, and regular human review, it's a genuine win. Faster for customers, cheaper for you, more interesting work for your agents.

If you automate aggressively and treat the number itself as the goal, you'll hit 70% faster and spend the next six months apologizing for it.