There's a running joke on the campuses where AlphaSights recruits. Everyone has been interviewed by them, but nobody has any idea what they do. The title might be something like Research Analyst, the job description has all the investment buzzwords, but as with many other entry-level jobs, the people these firms hire rarely know what they're actually signing up for.

There's a question floating around the industry right now: can AI replace the expert network analyst? To understand where the market is headed, we have to start with where firms are today. We have to explain what the expert network analyst actually does.

Your Day as an Analyst

Congratulations! You've just graduated college, and it's your first day at AlphaSights. So…what now?

Whatever your interviewers told you, your day begins with sourcing. An associate at a hedge fund just sent you a project. He wants to speak with five former regional managers at a large QSR chain. In many ways, an analyst is a lot like a BDR, and your job is to book meetings with these people. Since AlphaSights has been doing this for decades, you start with the company's existing network, an established bench of experts who usually participate in projects when asked. You pull a list of profiles, draft a few emails, fire them off, and wait.

The catch is that every project is a bit new. Maybe AlphaSights has two of those managers in their existing database; as an analyst, you still need to deliver the five that the client asked for. This is where the real job begins. You open your laptop, dust off your trusty LinkedIn Sales Navigator license, and start scrolling. You scroll and scroll and scroll until your fingers hurt.

Your day is spent finding and recruiting people who match a client's description. It's surprisingly difficult to judge a person's experience from their job title alone, so you cast a wide net. You stalk hundreds of profiles, sending as many messages as you can without triggering LinkedIn's anti-bot limits. Only some people respond, so you keep at it. Entire days are spent finding people who match a description and sending quick messages.

The responses you do receive need to be screened, so from time to time, you'll hop on a quick five-minute call to verify that a potential expert worked in the right region, at the right time, with the right operational visibility. These calls can feel a bit silly. There you are, it's your first job out of college, and you're reading off a script, pre-qualifying an expert who runs a $100M P&L.

Then comes scheduling. The fund operates on Eastern time, the expert is in Phoenix, and it's your job to mediate between them and find three slots that work in the next forty-eight hours. Basically, you're the white-glove version of Calendly. And when the expert no-shows (which is often), you run the whole song-and-dance all over again.

As for the interview itself, that's generally handled by the hedge fund, not you. In the rare cases when you actually do run an interview, the process is quite rote: asking pre-written questions, live-transcribing the conversation, and trying your best not to interrupt. Afterward, it takes another 30 minutes to turn your call into a client-ready transcript.

That's the job in a nutshell. Web crawling and logistics.

Which Parts Require Human Judgment

It's a grueling, manual job. And that raises a question: what parts can be automated?

This is broader than just an AI question. As GTM data improves, the task of finding people and filtering for the right experience keeps getting easier. The same trend has already taken root in B2B. The task of finding the right people is now better suited to a skilled data ops team than an army of analysts.

LLMs are just an accelerant. Think about what's actually happening when a research analyst looks at a LinkedIn profile and decides that person is a good fit for a project. A twenty-three-year-old with limited work experience is comparing a resume in a strange industry to an equally strange project description and using some quick rules of thumb to say "yeah, she might be a fit." That description isn't meant to be derisive either — that guess is genuinely good enough, and the rest of the process is set up in such a way as to make sure the right people still make it to the end of the funnel. The point is that this cheap, snap judgment we're relying on people for is the kind of thing Claude is actually pretty good at. Given the breadth of knowledge today's models have, I'd even argue that LLMs are better at this than most analysts.

The same points about automation are true of the outreach as well. While automation violates LinkedIn's terms of service (which is one of the clearest advantages to manual processes), other platforms are more forgiving. Those same GTM tools make finding email addresses easy, and the actual notes are templated anyway. Reaching out automatically makes a lot of sense.

Everything else — the sourcing emails, the screening checklist, the scheduling, the note-taking, the payment chasing — is administrative work that the industry decided, at some point, to staff with humans. It doesn't need to be.

The Real Question

The real question isn't whether AI can replace the expert network analyst. It's whether we've been asking the analyst to do work that ever required a person in the first place.

The expert is irreplaceable. They have spent decades operating a business, sitting in board meetings, watching markets move. No LLM can reproduce that. The investor is irreplaceable too. They're the ones paying for it all, making judgments based on the output of research.

The analyst, as the role has been defined, is something else. It is a layer of human friction stitching the two essential parts of the process together. The analyst is not the expert. The analyst is overhead.

AI doesn't replace expertise. It replaces overhead.

What This Means in Practice

If you're a source, this is good news. Expert interviews can feel rushed and transactional because the model is built around managing a human bottleneck. Remove the bottleneck and the experience improves: faster scheduling, cleaner conversations, payment the same day.

If you're an investor, removing the human bottleneck means you can finally run the research you actually want to run. Back when I was a consultant, our teams would have happily done hundreds of interviews if the costs and operational burdens weren't prohibitive. Less overhead means more research. (Incidentally, this is why the news isn't all bad for the analysts either. For the foreseeable future, AI alone appears to be less efficient than AI and humans together, and cheaper research means more demand.)

At Clairvoyant, we're building this infrastructure. The expert drives the conversation. The investor makes the call. The administrative middle just gets faster, fairer, and out of the way.