Generative AI chatbot FAQ

Get answers to your common chatbot questions.

What is the Hootsuite generative AI chatbot?

The Hootsuite chatbot leverages generative AI capabilities that allow you to quickly respond to customer queries in real time, giving customers the answers they need right away. Our chatbot draws responses from your organization’s pre-approved knowledge bank file to formulate contextual, intelligent responses, on the spot.

What type of AI engine is used for the chatbot?

We harness the power of our AI Service Provider’s, OpenAI's GPT or Microsoft’s Azure OpenAI, as the AI engines for the chatbot. Leveraging our expertise, we've designed the backend prompts and formulas to seamlessly interact with our service provider’s AI model to provide customized chatbot content for your customers on the social networks that you use.

As AI technology is in a dynamic space, we may also change or leverage additional providers or AI engines to provide the best customer experience.

How is our data used in the chatbot and do you share our data with your AI engine provider?

Your customers will be able to ask the chatbot questions and receive relevant responses, even if the question includes spelling errors or uses emoji. The inputs are sent to our AI service provider, who references your pre-approved knowledge bank file to generate a reply.

While our AI service provider may store your data for a temporary period (unless legally required to retain it for longer), storage is limited to the purpose of providing the services and maintaining the security of their systems, such as identifying abuse. Our AI service provider is not permitted to use your data for any other purposes or for any other customer. Your data, including any inputs you may provide, are not used by OpenAI to train their AI models.

We only use and store your input data and the AI-generated output for providing and improving our AI powered services. This allows you to access your data and past activity on our platform and helps us to make improvements and develop new features that leverage AI and machine learning technologies. Whenever we process your data, you can be assured that we will continue to safeguard your data. We will not share your data with any other customer or other third party (except to provide and improve the services).

How do you protect our data when using AI service providers?

We have strict privacy and security policies in place, and our AI service provider is contractually obligated to comply with a data processing agreement. We also take great care to ensure the security and privacy of all user data, and have implemented measures to protect against unauthorized access or use. For further information on our data protection measures, please review our Trust Center.

How do I know that the chatbot responses will be accurate?

We understand that accuracy of responses is important so we’ve rigorously tested the chatbot to ensure that it conforms to customer service standards and can handle various customer inquiry scenarios, from questions to complaints.

We are also implementing tests and evaluations based on the latest research into hallucination mitigation, to reduce hallucination risks. The chatbot only references your pre-approved FAQ library to generate responses, which further reduces the risk of hallucinations or responses containing inaccurate information.

The chatbot will be able to detect context and intent and provide responses that draw from the information you upload. In addition, there are predefined settings (in the case that the answer is not found in the knowledge bank file) that determine when the chatbot will hand off a conversation to an agent for more complex cases.

How do you ensure the chatbot is safe to use?

Hootsuite has conducted quality assurance testing and reviews product usage and efficiency to determine the effectiveness of the chatbot in meeting customer needs.

Our evaluation and testing framework seeks to ensure that outputs are safe, and we’ve implemented guardrails in the chatbot and our processes to ensure compliance. We do this by generating thousands of synthetic conversations targeting each feature of the chatbot, in order to assess its performance in the areas of compliance, brand safety, security, and topic limitation.

We also rely on our AI service provider’s own filters for negative results, including, bias, toxicity, political polarity, insult, threat, and profanity; and we have implemented our own test cases as a second layer of filtering.

Will I have oversight of the chatbot’s responses?

Yes, the chatbot’s responses are based on your pre-approved knowledge bank file, so if you ever need to add, remove, or update the content of your knowledge bank file, you can always edit and upload a new one.

If you have any issues with the response given by your chatbot, you can update the knowledge bank file to add or remove content as necessary, or you may need to update the tone or style that the bot will reply with.

For additional oversight, all chatbot conversations with your customers will be recorded and available to view at any time by your team. You can find them in the Resolved queue in Inbox 2.0.