If you watched our Q2 roadmap podcast, you’ll know we have major work planned for Conversational AI this quarter. First out the door are webhooks, which mean far more than you might initially think. In short, an agent can now do real work, and you can integrate it with your customers’ systems.
Personalisation with pre-call webhooks
Previously, any number mapped to one of your agents would connect the call immediately. The agent would speak and follow your pre-configured prompt. This is ‘ok’, but what if the caller has called 15 times already , is a VIP, or simply has a birthday? Now, for every call to one of your agents, we’ll send a webhook to your end-point and allow you to respond to dynamically reconfigure the agent. The caller will hear ringing for up to 5 seconds, giving you ample time to consult internal CRMs or other systems.
What you return to us can be a personalised greeting, potentially a whole new agent prompt, including dynamic data which we’ll get to in a second. So less “Hello, how can I help you?” and a bit more “Hi Mike, happy birthday! How’s the Pro Plan working out for you?”.
Configurable fields
In previous demos, agents have been given a goal such as getting someone’s contact details for a newsletter subscription and they pursue it well. However, that data was unstructured and didn’t lend itself to injecting into workflows or updating other systems like CRMs. Now, after every call a post-call webhook is triggered, sending details of the call to your configured endpoint. This will include the specific fields you requested, e.g. callback_requested=true, but for each one it’ll also include a human auditable rationale which saves trawling through transcripts if any verification is needed, e.g. “Customer explicitly stated they wanted a callback”.
I’ve long talked about TypeForm for voice and this is the start of that. The above example is crazy simple, but imagine an insurance claim form or some kind of KYC process where customers can complete complicated forms through a natural language chat. The chat can be loosely formed and jump around but the agent won’t lose sight of the fields that are needed.
Available today
Those in our beta program will already have access to Conversational AI; these features are being rolled out as an extension to agent configuration in both the API and portal. We look forward to seeing what you build with it!