How Zendesk Bot Selects Articles for Recommendations
How are articles selected for recommendations by Zendesk's bot?
Articles are selected based on their proximity to the question's concept on a conceptual map. When a question closely matches an article, they become 'neighbors' on this map, making it clear that the bot should recommend that article.
The selection process is fine-tuned by Zendesk's data science team using a 'threshold knob' that determines how closely concepts must match to be considered similar. This threshold is not adjustable by users and affects all accounts globally. A higher threshold means fewer, more relevant recommendations, while a lower threshold increases the number of recommendations but may include less relevant articles.
More related questions
How does Zendesk's bot use natural language processing?
Zendesk's bot uses natural language processing (NLP) to understand and evaluate articles. It mimics human behavior by reading each article in your help center to grasp the main concept. The bot creates a conceptual map where each article's concept…
Does Zendesk's bot learn from end user feedback?
No, Zendesk's bot does not learn from end user feedback in real-time. Although it uses a machine learning model, the bot's recommendations are not influenced by feedback from users or agents. Feedback is used in other ways, such as providing…
Is AI-powered search better than keyword search in Zendesk?
Yes, AI-powered search generally provides more accurate and relevant article recommendations than keyword search, especially when questions are asked in full sentences. AI-powered search uses natural language processing to understand the context…
Can I train Zendesk's bot by repeatedly marking articles as relevant or irrelevant?
No, you cannot train Zendesk's bot by marking articles as relevant or irrelevant. The bot is designed to function without requiring training from users. It is pre-trained to understand natural language and will consistently recommend the same…
Do labels influence article suggestions in Zendesk's bot?
Labels can help create a list of approved articles for Zendesk's bot to pull from, but they do not influence the weight or frequency of article suggestions. Labels are used to define a whitelist of articles for the bot to consider, but they don't…
How can I improve the performance of Zendesk's bot?
To improve the performance of Zendesk's bot, monitor your autoreply activity and analyze which articles perform best and worst. Ensure your help center articles are concise and well-organized, with titles phrased as short sentences or questions….
Does Zendesk's bot suggest articles from restricted categories?
No, Zendesk's bot does not suggest articles from categories restricted to agents and admins. These articles are not accessible to end-users. If you experience restricted articles being suggested, check the article permissions and consider creating…
Does Zendesk's bot search across multiple brands?
Zendesk's bot only searches within the help center of the brand it is associated with. It does not search across multiple brands. While cross-brand searches can be enabled in search settings, the bot is limited to its designated brand's help…
Interested indeflectingover 70% of your Zendesk support tickets?
Zendesk Support Tickets
Zendesk Messaging (live chat)
Join1,000+ companies reducing their support costs and freeing up support agents for more important work
“We needed an AI agent integrated within our current tools. My AskAI was the only solution that wasn't going to disrupt our operations.”
Zeffy
“At the end of last year I was given the challenge - how can we provide the same or better service, without hiring anyone?”
Zinc
“My AskAI blew everybody else out of the water. It made the selection process very easy for us.”
Customer.io($50M+ ARR)
“It now resolves 71% of queries (over 35,000 every month), meaning more time solving complex issues and improving UX.”
Freecash
“We needed an AI agent integrated within our current tools. My AskAI was the only solution that wasn't going to disrupt our operations.”
Zeffy
“At the end of last year I was given the challenge - how can we provide the same or better service, without hiring anyone?”
Zinc
“My AskAI blew everybody else out of the water. It made the selection process very easy for us.”
Customer.io($50M+ ARR)
“It now resolves 71% of queries (over 35,000 every month), meaning more time solving complex issues and improving UX.”
Freecash
Reduce support costs.Spend more time on customer success.
