Have you guys tested this in real-world scenarios? We attempted a similar integration a few months ago, adding payment and refund capabilities to our AI agents. For the chat agent, it works super well but for the call agent, there are just too many edge cases to take care of, and at times it feels like a tacky solution that can break anytime.
Some issues we identified in our cohort analysis:
- Users frequently mis-pressed numbers, leading to data entry problems.
- Many people were uncomfortable providing credit card details over a phone call, especially to an automated system.
- The system lacked intuitiveness, users struggled with putting the phone on speaker to press keys as instructed.
Overall a very bad user experience and basically a degradation in user experience compared to current user interfaces. It's just adding to the interaction cost of completing the task.
No matter how solid the NER models we use to protect user details from third-party LLMs and ensure compliance with regulations, we found that accountability remains a significant issue in cases of errors or fraudulent transactions.
Protegee AI (YC F24) is the payments API for AI voice agents.
Voice agents are great at conversations, but payments often get messy. Companies try all sorts of hacks, such as texting a payment link later, passing the customer to a human rep, or even giving up on taking payment altogether.
With a simple API, Protegee will collect the details, process transactions securely, and return the caller back to you with payment confirmation.
Today marks the start of B2AI - the agent economy. And AI2AI is coming soon.
Congrats on the launch, Xiaoyu Li and Kirthi Banothu!
🚀 https://lnkd.in/gDfS8TTN
Technical Co-Founder @ Happy Plum Language Learning | MongoDB, GitHub
2moDexter Horthy great work! I am wondering more ideas on this one