Since its explosion in popularity, Artificial Intelligence (AI) has become the single biggest lever businesses are looking to pull to improve their services. In the iGaming industry, AI has been transformative in player interactions, but knowing when to use it, and how much to use it, still remain key areas of thought. 

In a webinar hosted by SBC, iGaming Account Executive at Comm100, Karen Woods sits down with iGaming Consultant Iris Den Boer, Head of Operations at Reactivaction Kylie Tanti, Head of Safer Gambling at BV Group Lisa Corti, Head of ESG with JVH Gaming Liesbeth Oost and Maria Boelius, Chief Revenue & Operations at Wildz Group.

Why ‘Ai-first’ doesn’t mean ‘Ai-only’

On the operations side, Wildz Group’s Maria Boelius highlights how AI support is currently involved in about 90% of the conversations, escalating them to a human if its confidence in answering questions drops. An AI co-pilot is also utilised in chat translations, allowing Wildz Group to deal with customer queries, without having human agents that speak these languages. 

Ultimately, AI agents act as a first line between operators and customers, with humans sitting just behind them, ready to take the reins at a moment’s notice. 

Woods then prompts Head of Safer Gambling at BV Group’s Lisa Corti to speak on AI usage on the player safety side. Again, AI was used similarly as with ops, as a front-line filter that can then escalate issues to player safety agents. 

“Instead of having to go through a customer service agent on chat, they can use the AI bot to find us that way”, Corti said, “it’s been great at allowing us to filter out noise, so that we can better concentrate on the customers that we do need to see”.

For Head of ESG at JVH Gaming Liesbeth Oost, the answer is straightforward, “AI-first is more of an operational choice. AI-only is not a choice in our industry. We have a greater duty of care to our customers than other industries and increased AI usage doesn’t mean decreased operator responsibility.”

What gets automated? 

The panel was in agreement here that balance is key. AI usage dominates lower value, repetitive tasks, like password resets and general FAQ’s, but as Head of Operations at Reactivication’s Kylie Tanti said, “when there is more complexity to a question […] things related to vulnerable players, player complaints for example, where trust is involved, players need to speak to human agents”.

When asked about her views on what should be automated, Dan Boer highlighted something she felt the panel had so far missed – AI’s effectiveness in compliance, noting that “any high volume task with predictable outcomes and clear compliance flows can be really useful for this”. 

She continues; talking about AI’s ability to pick up on micro changes in tone of voice or cues and how that can be used to accurately escalate compliance issues that even experienced customer service representatives can miss. “AI is a system and it doesn’t typically miss cues and indications” she said, concluding that AI use should be focused on creating a safety net that human teams can fall back on. 

Where is it working?

Tanti was quick to point out how conversations around the impact of AI have largely underplayed its role in internal support. She highlighted how it helps teams improve efficiency, reduces the time it takes for people to find information like player data and history, and how it can summarise this information allowing support teams to go into conversations with players better prepared. 

Another area that gets overlooked is in quality assurance, with Oost saying, “it can check what human agents do, follow up on how well they did it and what they can do better next time.” She goes on to emphasize how teams training their humans to work alongside AI will have a competitive advantage to teams that aren’t.

Where is there friction?

Woods posed a question to the panel. Why do they think the handoff between IA and a human agent is the weakest link in player experience? Boelius points to the importance of making sure your AI agents are properly calibrated to actually initiate the handoff to a human operator when its confidence in dealing with a conversation dips. 

“It’s important that when the chatbot recognises the conversation is reaching a dead end, or is going around in loops, that it actually transfers the conversation to a real agent” she said, “it’s even more important that there is a summery, because when a customer is being transferred to a human, they are likely already frustrated, so the agent needs to be looped in on what has happened’.

Another interesting point of friction came from users who themselves refuse to engage with AI chatbots altogether, even if the AI bot is able to fix their problems easily.  Boelius laughed this off, but it points to a real point of friction as AI adoption grows, it’s likely that this cohort of users will grow along with it, meaning operators need to ensure their embedding of AI continues to be seamless and effective. 

Watch the full webinar here: https://www.bigmarker.com/sbc-webinars/ai-human-agents-how-to-find-the-right-balance-in-high-stakes-player-interactions