Everything Andrew Massay said on any show that made the record, most notable first. Each card names its show and opens the statement there.
Massay: Disclosing AI upfront increases voter conversation length
“I think what we've found is really interesting is more and more people will actually be comfortable and have longer conversations when you disclose that they're talking to an AI agent in the beginning, which you, yeah, you wouldn't think so.”
Massay: Google ranks third behind OpenAI and Anthropic in AI models
“Right now, I think Google is kind of on the lower, is probably number three next to Anthropic and OpenAI.”
Massay: Peer-to-peer political texting does not require explicit opt-in
“The law basically allows you to tap messages, essentially, and not have an explicit opt-in from that recipient.”
Massay: Both Political Parties Are Increasingly Adopting AI for Campaigns
“We've seen over the last six months to a year that more and more people that may have been hesitant previously are starting to come around to being okay with using AI in their campaigns. We do see it on both sides, really.”
Massay: Surviving Strict Political Regulation Enables Expansion Across Other Industries
“It's also the hardest, right? It's so regulated and so secure that if we can survive in this space, then we can definitely bring it to other industries.”
Massay: More Intelligent AI Models Will Make Security Holes Easier to Find
“And you're seeing now too, with some of the stuff Claude is putting out, or Anthropic rather, security holes are going to be easier to find with the more intelligent models.”
Massay: Claude Models Are Really Great for Engineering and General Knowledge
“I actually find that Claude models are really great for engineering and also just general general knowledge stuff as well.”
Massay: Local AI models will gain traction in high-security applications
“The local models are going to be, I think, more, more popular when it comes to very secure applications, and they don't want to send their data over To somebody like an Anthropic or an OpenAI or, you know, they'd rather keep it all in-house.”
Massay: Convos saved over 300,000 volunteer hours in Wisconsin election
“We calculated that it would have saved, that we saved over 300,000 volunteer hours responding to all of the messages that we did, which, I mean, you need how many volunteers to actually put in that many volunteer hours.”
Convos scans campaign text conversations to generate voter alignment scores
“You know, every time somebody has a text conversation, we're able to scan that whole chat and get so much intelligence out of it and, you know, find out what topics they talked about and save those and get an alignment score.”
Massay: Every political texting campaign must register with The Campaign Registry
“Every campaign has to be registered with an independent body called the campaign registry.”
Convos Sent 2.5M Texts to 1M Voters in Wisconsin Supreme Court Race
“The recent case study that we're about to publish was a Wisconsin race for a Supreme Court justice who got elected, and we sent out, I believe, 2.5 million texts to about a million voters out in Wisconsin.”
Massay: Convos drove 20,000 deep conversations from 1M voter texts
“Out of the million voters we reached out to, I believe we had 20,000 meaningful conversations. Meaningful meaning, like, four or more messages.”
Massay: Convos AI agents operate strictly within campaign-provided source points
“We do use a variety of different foundational models. But the way we've built them into the system just from day one has been with that safety in mind, and so ultimately, from a high level view, the way it works is when it's constructing a message that it want…”
Massay: Convos is operated solely by its two co-founders
“It's me and my co-founder Tom.”
Massay: Convos' vision is building direct voter brand recognition
“I think the grand vision for this thing is where the voters themselves actually are familiar with the brand, Convos, and that when they get one of our texts for from a political campaign, they almost know that they can text it back, and they can have a great c…”
Massay: California law requires AI disclosure in the first text message
“For example, California is one of those where you have to disclose. It's not even, you're not even given a choice. You have to disclose that it's an AI agent in the first text.”
Massay: Convos does not permit manual human text handoffs
“So there's no way to manually text back because we kind of believe in our agent to, you know, run that conversation. And also the agent is able to respond within 14 to 15 seconds.”
Massay: Convos AI automatically terminates conversations with hostile voters
“And if it's, if they're hostile, or it knows that it can't achieve its motive in any way, shape, or form it'll actually automatically shut down the conversation and stop responding.”
Convos uses AI to detect misspelled and natural language opt-outs
“Previously, when it comes to opt-outs, I'm sure you've seen it when you get a text, it says, you know, reply stop to stop to end or stop to opt-out. If you don't text back stop exactly, it doesn't count. It doesn't register that that's an opt-out. Whereas our …”
Massay: Convos Powers Two-Way AI Texting for Political Campaigns
“Yeah, so at Convos, we are an AI-powered platform that powers text messages with political campaigns, and so the idea there is instead of sending out blast texts that don't, you know, encourage any replies, we send out messages for political campaigns or the p…”
Massay: PubSent rebranded partly over search results and spelling similarities
“Looking up PubSent, you know, didn't always garner the best results because it's very close to the spelling of other words that aren't great, and so that was one thing. Also, I think Convos just overall reflects what we do better.”
Convos identifies 'knowledge gaps' when voters ask about unindexed policies
“Because in addition to topics that we surface to them, we surface them, we surface to them things that they didn't add to their source points, and we call those knowledge gaps, and then they can go back and say, oh, wow, we didn't have any policy information o…”