Everybody’s talking about Zendesk Answer Bots. AI was more than half of the 10 CX trends in 2024, and so far definitely living up to expectations in terms of dominating the conversation and administrators’ time and efforts. CX teams that are on the path to implement Answer Bot in their help center come across one big challenge - can you let these AI agents loose in production without testing them in a sandbox first? 

Most Zendesk teams update their most recent Guide articles in production rather than the sandbox, meaning articles in the sandbox Guide are often out of date. This is no good for us, given that we want to test our new bot against real, up to date articles, to check its accuracy and performance.

Without a sandbox environment that mirrors the help center production instance, conducting large-scale, production-comparable validation tests becomes impossible. This is particularly problematic for new generative AI functionalities, which rely on Guide to suggest solutions to customers accurately.

Testing AI and bot functionalities in the production environment is far from ideal and can lead to subpar customer interactions. This not only impacts customer satisfaction but also undermines the reliability of the bot's responses.

The best approach here is to use a sandbox that replicates production, and see that our AI bots actually deliver on their promise. This would enable safe and thorough testing without affecting live customers, allowing for seamless bot and AI functionality validation.

With Salto - you can easily and automatically replicate your production help center to your sandbox, including articles that exist in production but not in sandbox. This means that with very simple steps you can have a test bed for AI agents that is actually a replica of your production.

‍

With up to date articles living happily in the sandbox, you can begin to build your bot flows. An essential part of most bot flows is the ability for the bot to look up information from Zendesk Guide articles.

Zendesk’s Answer Bot can then feed this back to the customer either by linking to the relevant article or, more recently with Zendesk AI, by pulling one or two of the most helpful lines from a Guide article, and using them naturally within conversation.

AI agents may very well be the way of the future. To make the safe transition and incorporate them in your CX framework be sure to account for appropriate testing capabilities in your setup.