For the last several months, we have been building something we believe will fundamentally change how investment professionals conduct primary research.
Today we are announcing Checkmate: the AI-powered expert call platform.
The best investors know that their most differentiated ideas come from primary research, directly from the source. They spend hundreds of hours a year on expert calls to get that edge.
Hedge funds, PE firms, and consultants have already started adopting AI to conduct interviews. But today's solutions are essentially survey bots. They can't replicate the analyst-level depth needed for real diligence.
Checkmate is the first AI that actually digs deep for information to uncover critical insights. This is possible with our deep thinking architecture, which analyzes a conversation in real time to come up with the best follow up questions. It's so effective, that many analysts believe the quality of our expert calls to be better than their own.
We believe the future of primary research is large scale, rapid, qualitative research. Check out our launch video to see exactly how our AI pivots mid-interview to chase down high-value insights.
Checkmate is live today, and we are currently onboarding new clients in waves. If you're looking to automate your due diligence, channel checks, or any kind of primary research, I'd love to personally show you what we've built! Send me a DM, or request access via the link in the comments.
There's a lot of buzz around AI-powered interviews and market research voice agents right now.
Most of them work like smart surveys. They ask a set of predetermined questions, limited depth follow-ups, and wrap up in 10-15 minutes. That's fine for some use cases, but it's not how expert calls work.
When an analyst is doing primary research, they're hunting for something they don't know yet: an insight, a pattern, or some edge. The questions are open-ended and the conversation can go anywhere. Often times, the most valuable moment is the one you didn't plan for.
One PE associate I talked to described a call where they went in asking about a cybersecurity vendor's product positioning. Halfway through, the expert, who was a former IT buyer, mentioned that many folks he knew were switching to another up and coming vendor. That vendor wasn't even on their radar, but that offhand comment reshuffled the entire thesis.
In a well-run expert call, when a spark comes up, the analyst recognizes it and the whole conversation shifts. The successive follow-up questions matter more than the original script.
Building AI that can actually listen, identify what's interesting, and go deeper is a different problem from building a survey agent. Most tools aren't even trying to solve this!
A lot of firms assume AI-conducted interviews would be a compliance headache. Turns out, it might actually be an advantage.
For buy-side investors doing primary research, the big risk is encountering material non-public information (MNPI). You can talk to industry experts all day, but you can't trade on information that hasn't been publicly disclosed.
Most firms already have protections, such as compliance teams to vet questions. Often times a rep joins the call live. Analysts are trained on what to listen for.
But even with all of that, an expert might share something they shouldn't, unprompted. When that happens, the call usually gets killed on the spot. Sometimes it's worse - one hedge fund analyst told me they were building a position in a stock, and after an expert let something slip, they had to walk away from the trade entirely.
When an AI agent conducts the interview, the transcript can be reviewed (by AI or by a human) before it ever reaches an analyst. MNPI gets flagged and scrubbed, and a call that would've been tossed is now usable. The analyst gets the insight, without the exposure.
@BuysideAIreview Haven't seen that effect so far but it's possible. I've actually found that humans can be more honest when they are talking to AI, maybe because of less fear of judgment
AI can now do in 10 minutes what used to take a full day of investment research.
That sounds like a win. But it creates a new dynamic... if everyone has the same capability, where does your investment edge come from?
Public information is getting commoditized fast. Any investor can spin up a deep research report on a company in minutes, looking at 100+ sources on the web instantly, including public company filings and earnings calls. Your competitors are running the same queries you are.
So it must be proprietary data, right? The problem is, most "proprietary" data isn't that proprietary. Every top firm pays for the same Gartner subscriptions, transcript libraries, and alternative data vendors. You're all looking at the same dashboards. What's actually left? Primary research - expert calls, customer interviews, channel checks. This is where most differentiated investment ideas come from. The firms figuring this out are already shifting where their analysts spend time.
1,000 hours a year - that's how much time some buy-side investment professionals spend on expert calls. And that's just the calls themselves, not even including prep, notes, or scheduling.
Over the last few months, I've talked to 70+ hedge fund, PE and management consulting professionals about how they run their investment process. On the high end, people are spending 20+ hours a week on the phone with experts.
But no one I talked to thinks this time is wasted. Primary research is where most differentiated ideas come from. In a world where any information on the web is readily accessible, true alpha comes from what folks are saying on the ground.
I've been thinking a lot about what this means for the finance and research industries. More on that soon.
Dome (@GetDomeAPI) is a unified API for prediction markets like Polymarket and Kalshi. Trade and analyze data across multiple platforms with one simple API.
https://t.co/YsZk6Elw2Y
Congrats on the launch, @kurushdubash & @princeofbothell!
I love using OpenAI deep research. It’s usually the first thing I turn to when I need to quickly understand a new field. But it really struggles with reliability - sometimes the right answer just doesn't exist on the web, or it shows up in low quality sources (like SEO-driven aggregator sites). Benedict Evans explains it well: https://t.co/NpZX9HSXTr
As a hobbyist user it's fine, but I also wonder how many reports sound right, and are actually based on faulty data. Until these tools can somehow assess the reliability of a source, professional analysts won't be able to rely on them.
Try out a demo of our adversarial voice agents!
At Labyrinth, we're building a voice agent testing and simulation platform that lets you run your agent against other agents that simulate your customers. Agents can be configured to generate randomized scenarios for your use case, or grounded on real transcripts of calls. You can also set up an automated testing pipeline of voice simulations along with evals.
In this demo, you can speak directly with an agent to give you a taste of the product. Try one of our preset scenarios or choose your own! All you need is your phone.
If you're building voice agents and thinking about automated testing, or have any feedback or questions, let me know in the replies or shoot me a DM.
Link in the next post. 👇