ChatBotKit gives teams a place to configure AI agents and the experiences around them. Strawberry can help inspect agent instructions, knowledge sources, test chats, and embed preparation in the signed-in tab, then put the open decisions into a usable review.
The companion can support agent instructions, knowledge sources, test chats, and embed preparation in the ChatBotKit tab, then use the surrounding browser context to make the next step easier to check.
01
Find the instruction gaps before users find them.
A useful ChatBotKit pass begins by separating what the workspace shows from the decision a responsible owner still needs to make.
Review this ChatBotKit bot with its attached knowledge sources. Identify ambiguous instructions, unsafe answer patterns, and missing escalation rules.
02
Check whether the agent knows what the website says.
For ChatBotKit, a sensible response starts with the workspace history and the source tabs that give it meaning.
Compare the ChatBotKit knowledge source against the product and policy tabs I have open. Produce a source-by-source discrepancy list for review.
03
Make test conversations useful evidence.
A ChatBotKit review earns trust by showing its evidence, its gaps, and the question that cannot be answered automatically.
Read these ChatBotKit conversations and group failures by intent, hallucinated claim, retrieval gap, and handoff issue.
04
Set up a careful recurring bot check.
When the ChatBotKit checklist is proven, save it as a skill and run a routine that prepares the same evidence-led review on the team’s own schedule.
Every Friday, prepare a ChatBotKit agent-health review from the selected bots and transcripts. Do not deploy a change, rotate an API key, or alter an embed.
Yes. Keep ChatBotKit open in a signed-in Strawberry tab and request a review of the agent setup or the conversations visible there.
It can prepare a proposed revision, while changes to instructions or live deployment should be explicitly approved.
Yes. It can organise selected conversations into repeatable failure patterns and concrete follow-up work.
Yes. Open the relevant ChatBotKit source and the current product or policy pages, then ask Strawberry to produce a source-by-source discrepancy list for review.
Yes. It can work through selected test chats and return ambiguous instructions, unsafe answer patterns, retrieval gaps, and missing escalation rules.
ChatBotKit Strawberry