We use AI Assist FAQs as a production knowledge source for customer service. Tawk.to supports importing FAQs in bulk, but I cannot find a way to export the existing questions and answers.
As the FAQ dataset grows, this makes maintaining the information unnecessarily difficult and time-consuming.
The problem
To keep AI-generated answers reliable, we regularly need to review the complete FAQ dataset. For example, we need to:
- find duplicate or near-duplicate questions;
- identify conflicting answers;
- find outdated information;
- review dates, prices, availability and conditions;
- update multiple FAQs after a service change;
- compare the current dataset with a validated master version;
- perform quality checks outside Tawk.to.
Doing this FAQ by FAQ in the interface does not scale well.
There is also a specific AI Assist issue here: duplicate or outdated information is more than a content-management inconvenience. If conflicting information remains available to AI Assist, it can affect the answers given directly to customers.
The customer does not see the underlying knowledge-management problem. They simply receive an incorrect or inconsistent answer.
Real-world use case
After operating AI Assist for some time, we want to periodically extract all existing FAQs, review and deduplicate them, correct outdated information, validate the answers and then maintain a clean master dataset.
At the moment, getting the existing FAQ content out of Tawk.to appears to require manually reviewing or copying individual entries.
For a growing FAQ dataset, this creates unnecessary administration and increases the risk that obsolete or conflicting information remains unnoticed.
Why this is important
Being able to review the complete FAQ dataset would help improve:
- customer experience;
- accuracy and consistency of AI Assist answers;
- knowledge quality;
- first-contact resolution;
- change management;
- quality assurance and governance;
- maintenance time and operational efficiency.
It would also make it much easier to maintain a single validated source of truth for customer-facing information.
Suggested improvement
A simple Export all FAQs and Delete alle FAQ’s option would solve much of this problem.
CSV would already be sufficient, preferably using the same basic structure as the existing FAQ import:
question,answer
If possible, fields such as FAQ ID, status, creation date and last-modified date would make the export even more useful.
An API for retrieving FAQ data could provide another solution for organisations managing larger knowledge sets.
Question to the community
How are other AI Assist users currently reviewing, deduplicating and maintaining larger FAQ datasets?
If there is already a way to export all AI Assist FAQs that I have missed, I would be very interested to know how you are doing this.
If not, I would like to suggest FAQ export as a feature request. It would close an important gap between bulk importing knowledge and being able to efficiently manage that knowledge throughout its lifecycle.