SuperMe Raises $20M to Scale Its Network of AI Expert Profiles

SuperMe has raised $20 million in a round led by Kirsten Green at Forerunner, with Greylock Partners participating, to scale a network of AI expert profiles built from the actual knowledge of real operators, VCs, and founders. Instead of getting one generic chatbot answer, you ask a question and SuperMe assembles a virtual panel of specific experts, pulls them into a group chat, and answers from their accumulated playbooks. It matters right now because the startup is betting that the next big AI moat isn't a smarter model, it's verified human expertise with the human removed from the scheduling conflict.

---

What Is SuperMe? The Expert Network, Rebuilt for the Chatbot Era

The idea is almost embarrassingly simple: the best career and business advice lives inside a few hundred heads in Silicon Valley, and access to those heads is gated by warm intros, $500-an-hour consulting fees, or luck.

Image: SuperMe turns the informal "who should I ask?" network into a searchable product.

SuperMe's founders, Casey Winters and Ludo Antonov, started the company in 2024 to ungate it:

• Winters has held executive roles at Grubhub, Pinterest, and Eventbrite, and advises companies including Airbnb and Canva part time.
• Antonov was head of engineering at several startups, including Pinterest, Lyft, and Whatnot.
• The two met when Winters was advising Whatnot, right as ChatGPT was going mainstream, and decided the moment was right for a network business wrapped in AI.
• Winters' framing to TechCrunch: the goal is to make "the network of who the absolute best people are more transparent" and unlock the insights "locked inside their heads."

You type a question, SuperMe picks the experts best suited to it, and spins up a group chat of their AI profiles. You can also DM a single profile, or return to a group later for follow-ups.

The Core News: $20M, a Small Team, and a Free Product That Won't Stay Free

The funding is the headline, but the operating details are what tell you how serious this is.

Image: The product is closer to a routing layer for expertise than to a general-purpose chatbot.

• The round: $20 million, led by Forerunner's Kirsten Green, with Greylock Partners joining. Forerunner has backed consumer-defining brands; Greylock brings deep enterprise and infra bench strength.
• The team: roughly 13 employees, hiring on two fronts, model quality (conversation and expert matching) and expert scouting for uncovered subject areas.
• The business model: free today, but the company plans to cap free questions and charge beyond that. It is also designing incentives for experts to keep contributing knowledge.
• The distribution trick: SuperMe supports MCP (Model Context Protocol), so you can query it from inside Claude and ChatGPT rather than only on its own site.

That MCP support is the quietly important line in the story. It means SuperMe is positioning itself as a tool the big chatbots call, not a destination you have to remember to visit.

Why This Matters: AI's Real Problem Is Not Intelligence, It Is Trust

Generic models are confident about everything and accountable for nothing. Ask ChatGPT for pitch deck feedback and you get plausible advice from no one in particular. Ask SuperMe and the answer is attributed to the AI profile of a named investor.

A real example from the product: "Analyze my pitch deck. Which parts of the story are the weakest, and which questions should I be most prepared to answer?" The panel could include profiles of longtime VCs like Mercedes Bent, Sarah Tavel, and Josh Elman. That is a qualitative shift in how the output feels, even if the underlying model is similar.

The "so what" for founders, especially outside the US, is bigger than it looks. The reason a Delhi or Bengaluru founder can't get 20 minutes with a top Pinterest operator isn't talent, it's time zones, price, and proximity. Compressing that distance into a chat window is a genuinely useful product if the answers hold up.

Under the Hood: Data Ingestion, MCP, and the Privacy Bet

How SuperMe builds an expert profile is the part most readers will want to copy.

Image: Most of the training signal comes from content experts already published, which lowers the onboarding ask.

The ingestion pipeline, step by step

1. Public footprint first. SuperMe scans LinkedIn, YouTube, Substack, and personal blogs for what an expert has already said in public.
2. Private context, optionally. Experts can connect Google Drive, Notion, Gmail, and meeting notetakers like Granola for material they would never publish.
3. Profile construction. The system builds a conversational persona tuned to that person's frameworks, not just their biography.
4. Matching at query time. A router decides which profiles are relevant to your question, then assembles the group.
5. Answering with guardrails. SuperMe says some ingested data includes personal information and trade secrets, but it ensures none of that surfaces in responses.

The uncomfortable question hiding in step 5

"Trade secrets won't appear in answers" is a policy, not a proof. Every enterprise AI vendor in 2026 is making a version of this promise, and the entire prompt injection problem exists because policies are breakable. For experts, the incentive to connect Gmail is high and the verification is invisible.

The Competitive Landscape: Everyone Wants to Clone the Smartest Person in the Room

SuperMe is not alone, and it is not even the first. The space now has three overlapping lanes.

The strategic risk is obvious: OpenAI and Anthropic can ship a "verified expert" layer as a feature. The strategic hedge is also obvious: a curated, consented, attributed network of named people is a supply-side asset that takes years to build, and that is exactly what 13 people and $20 million are for.

India matters here too. Topmate has already normalised paid 1

expert calls for Indian creators, and the appetite for monetised expertise is proven. A SuperMe-style product localised for Indian operators, investors, and D2C founders is a very viable regional play.

What This Means for AI-Tool and AI-News Publishers

This story is a content goldmine because it sits at the intersection of funding news, product reviews, and a live ethics debate. Five angles worth publishing this week:

• "I tested SuperMe's VC panel against ChatGPT": Run the same pitch deck question through both and publish the side-by-side. Comparison posts like this rank for years and are cheap to produce.
• SEO play on the expert-clone cluster: Build a hub page targeting searches like "AI clone of [expert name]," "Delphi vs SuperMe," "can I talk to an AI version of a VC," and "MCP AI expert network." Low competition, high intent.
• The consent and trade secrets angle: A piece on what happens when your Notion docs become someone else's chatbot answer is a strong newsletter lead, especially with DPDP Act compliance now top of mind for Indian founders.
• The MCP distribution angle: SuperMe shipping inside Claude and ChatGPT is a concrete case study for your "MCP is the new app store" narrative. Tie it to other MCP launches for a listicle that keeps updating.
• Expert-network economics explainer: Compare the $500/hour human model with the near-zero marginal cost AI model, with a table. Finance-literate audiences share these.

Also note for your own operations: if you write tool reviews, your published reviews are exactly the kind of data these systems ingest. That is both a distribution opportunity and an attribution problem worth covering honestly.

Challenges Ahead: Incentives, Hallucination, and the Consent Problem

The round is real. The hard parts are realer.

• Expert incentives are unsolved. Winters' team is still "considering" how to pay contributors. Until that's concrete, profiles risk going stale, and stale expertise is worse than no expertise.
• Attribution without accountability. If a cloned VC profile gives you bad advice, who is liable? The expert did not say it, the model generated it, and the platform hosted it.
• Hallucination inside a persona. Users will over-trust an answer that sounds like Sarah Tavel. Confident mimicry is a bigger risk than generic confidence.
• Privacy is a promise, not an architecture. Ingesting Gmail and Granola for trade secrets raises the stakes of any leak from embarrassing to litigable.
• Free to paid is a cliff. Capping questions on a product people are still learning to trust may kill usage before habit forms.
• Thin coverage. With a small team scouting experts, whole categories will return a shallow, mismatched panel and quietly disappoint.
• Platform dependency. Building on MCP inside Claude and ChatGPT means the roadmap partly belongs to someone else.

---

Final Thoughts

SuperMe is making a bet that the scarcest resource in the AI era is not compute or model weights, it is credible, attributable human judgment. If the company can keep experts contributing and keep their private context genuinely private, it becomes the trust layer that general chatbots structurally cannot provide. The next 12 months will reveal whether this is a durable network business or a very well-funded feature that the labs absorb, and the answer will probably depend more on expert incentives than on model quality.

FAQ

What exactly did SuperMe raise and from whom?

SuperMe raised $20 million led by Kirsten Green at Forerunner, with Greylock Partners participating, announced on October 29, 2026.

How does SuperMe actually work?

You type a question, the platform matches it to the most relevant expert AI profiles, and it creates a group chat where those profiles answer together. You can also message a single expert profile directly or return to a group for follow-ups.

Who are the experts on SuperMe?

The network includes longtime VCs and operators such as Mercedes Bent, Sarah Tavel, and Josh Elman, with profiles built from public content on LinkedIn, YouTube, Substack, and blogs, plus optionally connected tools like Google Drive, Notion, Gmail, and Granola.

Is SuperMe free, and when does that change?

It is free right now. The company plans to cap the number of free questions and introduce paid tiers, but has not announced a date or price.

What are the biggest risks?

Three stand out: whether experts actually get paid enough to keep contributing, whether private data ingested from Gmail and Notion truly never surfaces in answers, and whether users over-trust advice that sounds like a named person but was generated by a model.

Can I use SuperMe inside ChatGPT or Claude?

Yes. SuperMe supports MCP (Model Context Protocol), so you can query its expert network from within chatbots like Claude and ChatGPT instead of only on the SuperMe website.