I’ve known Lyle Pratt since he founded BetterVoice in 2010, and was then acquired by Inteliquent (wholesale voice services) in 2016. Lyle was part of Inteliquent / Voyant, which was acquired in 2021 by Sinch. Lyle then left to build Vida as LLMs (Large Language Model) were getting interesting in automating some business functions.
Through Lyle’s time in Inteliquent he’s been build machine learning applications, such as spam and fraud protections, basic chatbots, And with the rise of LLMs Lyle saw a wholesale change was going to happen.
A key point is not just the customer interactions, LLMs can be applied to the whole stack, all the backend functions. We saw that in the interview with Igor Boshoer from Clerk Chat.
Lyle sees all programmable platforms will require AI agents, UCaaS, CCaaS, CPaaS. And given how easy they are to use, the best course of action is building the full business application.
At the end of the day, a plumber looks at their phone and sees 20 missed calls, some will be spam, some will be good leads. Vida filters that, making the plumber’s life easier. They have also packaged the Vida application to make it easy for resellers, managed service providers, and value app reseller. Lyle is applying the lessons from BetterVoice, make it easy for the customer by delivering Vida in the easiest way for them.
Lyle then runs through the reseller experience with the Vida dashboard at the 5 minute mark in the video. Creating accounts, viewing usage, setting up white label domains, bundles, billing plans, creating and tracking agents, agent functions like taking a message or transferring a call or sending a SMS / email or sending a notification, or much more complex functions with webhooks like managing schedules or inventories.
Vida are also building some power integrations with the 7000 applications available from Zapier. If the business is larger and has its own IT shop Vida supports them as well through its API, that Vida is built on. They’ve made it very easy for channels of all types to use Vida.
For smaller businesses that continue to use their cell phone they have an app, so agents can be used for some of the callers, and the smaller business owner can also answer if they have the time, or its an important customer.
We then discuss how the scheduling function can be configured, given all the small business scenarios. It’s the most used function in Vida, and they’ve built significant functionality that is easy to use, and because it’s an LLM it takes just a prompt. However, most small businesses are not prompt engineers, hence enabling the common asks with a good prompt helps the customer.
They support RAG (Retrieval-Augmented Generation), the API and service is fully documented. On hallucinations adequate prompting is key. Customers can test their agents, ensure it does what they intend. When the agent does not have a prompt on a particular topic, transfer to a human avoids many hallucinations.
Some of Vida’s customers are lead-gen, and the agent can ensure the consent forms are complete in compliance with FCC rules.
Improvements coming soon are scraping your company’s website, vectorizing it, and making it available to the agent. So it the answer is not available from a prompt or RAG, or the website scrape, then transfer to a human. Currently with AI agents its a continuous evolution.
For 2025 developments include new outbound capabilities such as keeping CRM records up to date, sending outbound SMS reminders for scheduled events. With CRM integration, customers can be greeted by name, last interaction is available, also pipeline status if a deal is being discussed. Vida is expanding from inbound focused, to in and outbound.
Vida is a great example of using Voice AI in a focused way, and helping customers with the prompts to do what they need, as a small business is not a prompt engineer.