How AI can help with your hardest tickets

The best thing AI did for customer support was take the repetitive tickets off your team. Order status, tracking, the same policy question fifty times a day. That was real work, and handing it to a bot was a genuine relief.

But it changed the shape of what was left. Take the routine out of the queue and it does not get shorter so much as harder. What remains is the handful of tickets that were never going to be simple.

The tickets that are left

Picture a customer whose order arrived damaged. Before anyone can help them, a rep opens the order in Shopify, the conversation history in the helpdesk, the carrier’s tracking, and the returns policy, then pieces the story together. The reply itself is two minutes of judgment. Getting to the point where you can write it is ten minutes of hunting across tabs.

That hunt runs on every complex ticket, every day. The bot was always going to hand this part over.

Why the bot stops here

For two reasons, and “the AI isn’t good enough yet” is neither of them. The first is feasibility. A bot answers from what it has been given, so to make one handle the hard tickets you would have to pour the whole of your business into it: every policy, every exception, every judgment call your team makes on instinct, and then trust it to surface exactly the right one on a case it has never seen. That is a vast, never-finished project, and it sits outside what most brands are good at. The few who take it on pay enterprise money to try. Everyone else lets the bot take what it safely can, which is why refunds sit under 10% AI-only handling, and hands the rest to a person.

The second reason matters more: customers do not want a bot on these tickets.

On a simple question, nobody minds the automation. On a hard one, they want a person. A human on these tickets is not a gap waiting to be closed. It is what the customer is asking for.

Where the relationship is won or lost

It is also the costly end of the queue. A routine ticket carries almost no relationship risk. The hard ones are the opposite: handle one well and you keep a good customer, handle it badly and you lose them.

These are the interactions your team is there for. The problem was never that they need a human on them. It is that the human starts every one from scratch.

How AI can actually help

Not by putting a smarter bot in front of the customer. That is the one place customers do not want it, and the work your team does best.

It helps behind the agent. The moment a hard ticket lands, AI can do the hunting the rep would have done: read the order, the history, the tracking, and the policy across every system, and hand over the full picture with a recommended next step. The rep reads it, makes the call, and writes the reply themselves.

The work stays human. The scavenger hunt in front of it does not.

AI took the easy half of the queue off your team’s plate. The hard half is the half worth getting right, and it is finally something AI can help with too: not by answering for your people, but by getting them to the answer faster.

Further reading

• AI in customer service is two products. Why the customer-facing bot and the team-facing assistant are different purchases, and why the brands getting AI right run both.
• The right deflection KPI. The per-category ceilings: 60-70% on order status, under 10% on refunds.

Sources

• Gorgias 2026 State of Conversational Commerce. 16,000 brands, 350M conversations. 86% of AI conversations eventually involve a human; refunds and returns handled by AI alone at 8%.
• Gartner, 2024 consumer survey (5,728 consumers, surveyed December 2023). 64% would prefer companies did not use AI for customer service; 53% would consider switching to a competitor if they found out a company was going to use AI for customer service.
• Gartner / CEB, The Effortless Experience. Study of 97,000 customers: 96% of those with a high-effort interaction became more disloyal, against 9% with a low-effort one.