Table of Contents
If last week Upland quit messaging, this week call-tracking analytics powerhouse Invoca declaratively entered the text-messaging market with the acquisition of Symbl.ai.
For almost twenty years, Invoca has been building intelligence on top of the call, including things like sentiment analysis. It would seem that the current focus on the multi-modal conversation and the combination of AI was just the wave Invoca was waiting for.
Symbl.ai has been building custom LLM-based solutions around text transcription, and yes, text messaging. The latter is a harder problem to solve than one would imagine.
This Is Personal
From 2016 to 2019, I led a skunkworks project that tried to detect fraud in text messages. The key was to catch it in the text message composition phase rather than after it had entered the network. But something about the 160-character limit caught even Google’s NLP flat-footed.
I was working with a pair of 10x programmers. The kind you give an extremely hard goal; throw out all agile best practices about velocity, work-in-progress, and managing scope and then leave them be.
This problem had not been tackled before. At every check-in, I’d only hear about how stuff did not work. Google couldn’t get the sentiment right because there wasn’t enough data in the payload. Text messaging is highly contextual, and it was nearly impossible to build context with just 3-5 words. More on this later.
Symbl.ai seems to have solved that problem. Its API allows you to run sentiment analysis. It also allows you to be compliant with personally identifiable information (PII) and keep track of conversations across text and calls. It was a 2024 honoree of the Twilio Startup Searchlight awards.
With this acquisition, Invoca CEO Gregg Johnson and team have used M&A to hack time. They’ve decided to buy, and overnight added text-messaging capability. Moreover instead of adding the plumbing to the platform, you are adding the product and letting the product (conversational AI) determine how the pipes will be used. It’s the classic “Make one decision that makes a hundred decisions for you.”
Finally
As for the skunkworks project, we did finally get it right. The ultimate test came when our NLP-tuned engine took a seemingly innocent real text (short and full of abbreviations) conversation about rates and a hotel room—with none of the usual keywords that give away intent—and correctly identified the use case as “adult.” These days, I can only imagine what they’d do with this information. Both brothers have since moved on to be successful blockchain innovators. But I digress.
Congrats to Gregg and team on the acquisition. One can expect more news from Invoca in the days ahead.