AI: plenty of promise for affiliates, but also some pitfalls

Having recently undergone a large-scale overhaul of the tech used by our business, it made sense to explore the ways we could use AI to make our business more efficient at the same time.

As a business, we have always been very process-driven and keen to explore tools that allow us to automate more processes. We see this as a way to free up people's time so they can focus on solving complex problems that fully utilise their talents and capabilities, rather than as a means of potentially replacing them.

At QiH, we’re still at an early stage of our AI journey. As yet it’s too soon to say whether it will have a transformative effect on either our business or the affiliate marketing industry more widely. What is already clear though is that it shows real promise in certain areas.

Drilling down into the data

Historically, some of our data analysis required us to gather huge amounts of data and look at it manually. In some cases, AI has helped us get tasks that used to take six or seven hours down to an hour. However, it’s worth pointing out that there’s still a human element needed to sense check for quality.

One of the key areas we see AI being of value to our affiliate business is in positioning partner offers. We’ve built our own algorithm to assess exactly what position different partners should have on different pages to drive the most value for both parties. However, adding an AI component has meant it can evaluate many more data points, allowing our algorithm to develop further and be more effective.

At this stage, we wouldn’t use an AI model alone to make our page ordering decisions as we’re still learning and developing our knowledge of what works. But by comparing the outcome of the model with our knowledge, and by tweaking and modifying it each time, it’s gradually requiring less human intervention.

We also see the potential for AI to be useful in personalising offers to improve our player targeting. As a business, we are now getting a vast amount of data from various different sources, including traditional marketing sources, Facebook, Google Ads and interactions on our websites. The purpose of capturing all that data is so that we can personalise our targeting. AI is helping us fast-track our analysis of that data without having 100 analysts on hand, meaning we can very quickly present the most appropriate offers to players who’ve only recently interacted with our sites.

At QiH, we've already benefited from an uplift in the business based on the way AI is helping our data analysis. We’ve seen a 30% growth in revenue due to our improved decision-making on positioning the right offers so our traffic is matched up with the right journeys. We’ve also seen a 15-20% uplift in retention due to improved player targeting.

Getting creative with content

Content is the other obvious area where AI has potential. We’ve found that an initial draft of something like an article or email that would ordinarily take someone several hours to write can be produced much more quickly using AI tools. However, we still require human intervention to refine the final product.

Internally we’ve developed processes that we see as having the balance right between utilising AI and making sure content always appears to be human.

However, one learning for us has been that you need to be cautious about external suppliers. We’ve had freelance writers deliver copy that has clearly been generated by AI. This highlights one of the negatives of AI, that it can sometimes lead to lazy work, with people thinking they can just tap what they want into a tool and that’s the job done.Considering the significance of content in affiliate businesses, we would never publish copy that was evidently AI-generated.

We’ve also seen significant benefits when it comes to images. Our creative team would previously sometimes spend a couple of hours looking for quality images, but AI tools can often cut this down to 20 minutes.

We’ve also seen benefits when using AI to alter images, though this is one area where we’ve also seen some comical results. For example, someone asked an AI tool to change one of the team’s headshots so that he was wearing a suit and tie, but he ended up looking like someone who had walked straight out of the 60s!

Broader uses

While data and content are the most obvious areas we’ve seen early results, we are also looking at how we could use AI in areas such as HR, finance and compliance.

Compliance requirements are a key focus for us, for example, we need to stay on top of T&Cs for offers we’re promoting from our partners. We have been using AI to detect changes across their documents, as well as changes on our partner pages that we direct players to. Gambling is a highly regulated industry, so it's crucial that we are notified of any revisions to ensure our ongoing compliance. To have people looking at the volume of partners we work with every day would be not only costly but also arduous, whereas AI tools can do a lot of the heavy lifting here.

We also see it as having interesting applications in detecting and preventing fraud, which has been quite prevalent in gambling in the past.

Overall, we’re approaching it with an open mind but also adopting a measured approach. AI is not a magic tool or silver bullet where you just plug something in and expect the output to be perfect. As with all new systems and tools, you’ve got to put the time in to understand what works and what doesn’t.