How AI can deal with your post-peak backlog of tickets
5 Jan 2026
5 min read
So it’s peak, and you have a backlog of customer service tickets. What do you do now?
Don’t worry, you’re not alone. It’s a story we’ve heard a lot over the years. Ecommerce brands who keep to their SLAs all year, do lots of capacity planning for peak, and then see SLAs go out of the window as soon as ticket volumes go up, and response times balloon to two weeks or more.
Backlogs are a vicious cycle. The longer it takes you to handle tickets, the more those customers come back again and again on the same issue. So the backlog gets bigger and bigger.
The good news is that AI can help you get those backlogs under control. Here’s how we approach the problem.
What kind of backlog is it?
This sounds like a silly question, but it helps shape the focus of the project. Broadly speaking there are two types of backlog – general or specific.
By general backlogs we mean when there is a general increase in tickets across all types. This would be because the volumes of tickets overall have got higher than you can comfortably manage.
By comparison, a specific backlog is when there is an increase in a particular type of ticket that you are not able to shift. This is most likely something unexpected such as a product defect or a major carrier outage. It’s something that you haven’t planned for and so haven’t got good processes in place.
Let’s take a look at both examples in turn.
General backlogs
The first step we run is an automation analysis. We use our Intent Detection through AI to categorize tickets into different use cases. So we can map out which tickets are related to order queries, which are related to returns, which to damaged products and so on.
This gives us an understanding of where the opportunities for automation lie and where we can start to tackle the backlog.
To help brands get started, we have Genius Flows. These are our best practice workflows for common e-commerce use cases, such as order queries and WISMO (Where is my order). These enable brands to be automating quickly – in fact one customer was able to connect everything in less than 30 minutes.
From that point, you can soft launch the automations, and see how they would deal with real enquiries. When you are ready to launch, you can then roll them out and start answering all the backlogged tickets in those categories.
There are also some other low hanging fruit to get through, namely spam and business enquiries. Although you are probably not responding to spam queries, you still have to check through and remove them, and it makes your backlog look bigger than it needs to. So we can automatically close those for you.
The net effect of this is that you can quite quickly be automatically taking away a good proportion of your backlogged tickets, and preventing more tickets being added to the pile, while your team gets through the rest. Using this approach, AllSaints managed to reduce their backlog during peak by 50%.
Other brands have achieved 30% automation in 30 days, which makes a significant dent in your backlog.
Specific backlogs
These sorts of backlogs usually occur when you already have some use cases handled by AI and automation, but a longer-tail use case suddenly becomes more prominent. Or it could be an edge case that you didn’t predict but has come out of nowhere to be very common.
In this case, the use case gets bumped up the priority list for automation. What we would do in this case is work out what the desired response and process that your human agents would follow and then replicate that in DigitalGenius.
Imagine that you have changed your packaging provider and it’s not gone well: customers are complaining that their boxes are arriving damaged. You’ve detected the intent behind this message and can see it’s creating a backlog.
So you need to build out a process for your AI Agent to follow to resolve these queries, and then test it. Once you’ve done that, you can then apply it retroactively to the backlogged tickets, and cut them down to size.
Once again, this can be a quick process providing you know what you want to say, and you have a clear workflow to follow.
The problem that a lot of AI agents have with new edge cases is that they are not trained to handle them. Prompt-only agents tend to want to answer every question, and if it’s one they’re not familiar with, the agent will end up giving an unvalidated answer. This could be fine, but most likely it’s going to be imperfect or even wrong.
Our approach means that you are validating your approach before you start automating these tickets. This ensures that your customers get the right answer, and they are not coming back again and increasing your backlog further.
Preventing future backlogs
The advantage of using AI effectively to resolve tickets is that the backlog should rarely get too high to be manageable.
Tackling use cases one by one, and going for the biggest use case means that you can quickly get to a good automation rate, which immediately limits the height of the backlog.
On top of that, having 24/7 AI responses means that your agents are not seeing a huge backlog number because many of their tickets have been handled overnight.
While specific use cases can suddenly blow up and mess your inbox up, you can leave those tickets to one side until you have an automation worked out to handle them. Then, you can clear it in one go.
To find out more about how DigitalGenius can clear your backlog, speak to us today.
