A one-week lead
on every other fund.
Expert calls in 30 seconds. Channel checks in minutes. While your competitors are still scheduling intro calls, you're already refining your thesis.
- 301,000-agent BLS-grounded synthetic-expert panel. Every occupational tier from C-suite to floor-level operator, in every industry.
- Zero MNPI risk by construction. Synthetic agents have no insider knowledge to manage, log, or monitor.
- Every claim sourced, every cohort audit-traceable. Three-proof rule: verify any three claims in five minutes.
- Benchmarked against prediction markets across 18 dimensions. Calibration is published, not asserted.
- Works where you already work. SaaS, REST API, MCP server, Microsoft Teams, Slack.
Three Things Change When Expert Calls Become Always-On
Expert calls today are events - scheduled a week out, analysts huddled around a speakerphone, the network's polished VP-of-Industry-X on the line. Thesis Lab makes them an activity. Three things change downstream.
01
A One-Week Lead
Expert calls take one to two weeks to schedule. Thesis Lab experts are available in 30 seconds, any-time, and answer immediately. You're refining your thesis while competitors are still booking their intro call.
02
Calls Stop Being Events
No more huddling around a speakerphone, no more straining to parse what the consultant is actually saying. Calls happen whenever you want them. Convenient, accurate, and immediate.
03
Different Experts From Everyone Else
Expert networks recycle the same handful of senior experts across every fund - alpha decays as the answers converge. Thesis Lab opens access to operator-tier experts: schedulers, dispatchers, practice managers, supplier reps.
Deliver differentiated investment theses faster, with more confidence.
The Work Isn't the Decision. It's the Intake.
Every analyst we interviewed described the same monthly ratio: 65% gathering and absorbing, 25% forming the view, 10% making the decision. The judgement work - the part you were hired for - is a sliver of the week.
Intake
65%
of an analyst's typical month spent gathering and absorbing - not deciding.
Onboarding
20–40h
unguided reading per new ticker before a useful expert question lands.
Lead Time
1–2 wk
typical scheduling lead time for a single expert-network call.
Yield
1 in 3
expert calls that produce material new signal. The rest is colour and waste.
Credit Analyst
About 70% gathering / absorbing, 30% actually forming a view and pulling the trigger. In quarter-end crunch I bully it to 55/45 because, no, I am not going to read a seventh whitepaper to feel safer.
- Nolan Navarro
The Thesis Interrogation Packet
Two pages plus appendix. Every claim sourced to a clause. Every cohort traceable. Delivered before lunch.
Page 1
Decision Brief
Recommendation, thesis, three key drivers with operational ranges. Three key risks. “What has to go right” / “what has to go wrong.” Next three actions for the analyst.
Page 2
Evidence Map
Assumptions colour-coded by confidence (green / amber / red). Counter-argument with sources. Leading indicators with thresholds. Gap list - what's still unknown and how to close it.
Appendix
Source Evidence Table
Every claim mapped to citation → page → date → confidence → counter-evidence. Drops directly into your memo, your IC deck, or the live expert call you'd already booked.
Hedge Fund Analyst, Study 2
If the written output gives me clause-level citations, operational ranges, and a genuine counter-argument I can verify - I would read that before scheduling a call. It would make my live call 10× better.
What It Looks Like
Not a demo video. Three screens from the actual product. Click any image to enlarge.
Recruit an expert panel from a sentence. Describe the operator you want in plain English - “managers and logistics coordinators at restaurant-supply companies that specialise in poultry.”
Your panel, ready in seconds. A real cohort drawn from 301,000 BLS-grounded agents. You see who you're talking to, their background, and the recruitment audit trail.
Always-on, parallel, recorded. Run a portfolio of synthetic expert calls in parallel. Every call recorded, every cohort scored, every claim cited.
From a Sentence to a Sourced Packet - in Three Steps
Describe Who You Want to Talk To
“Independent dispatchers at mid-market trucking firms in the southeast.” The Recruitment Specialist Agent matches your description to a recruitable cohort across 301 role archetypes and 2,736 user-vernacular aliases.
Interviews Run in Parallel
While you make coffee. Every interview transcribed, every cohort scored, every claim cited. You see the panel as it's recruited and the transcripts as they land.
You Get the Packet
Two pages plus appendix. Clause-level citations. Counter-thesis. Gap list. Ready for the IC deck, the memo, or the live expert call you'd already booked.
Not a Substitute for Expert Networks
The best way to get to the core 1–2 key topics you need to validate.
You'll still book the GLG call. You'll still pay the $1,500. The difference: you walk in with a tighter question, a better expert match, and a packet of operator-tier context the network can't provide. The same call generates ten times the signal because the synthetic pre-pass already burned off the noise.
Financial Advisor
I want the office manager, dispatcher, or ops lead who closes the day and fights the fires. If they can't explain their close-out in plain English in 30 seconds, I pass.
- Krista Nina
Works Where You Already Work
SaaS, API, MCP, Teams, Slack. The packet drops where you read it. The output is identical across every channel: clause-cited, confidence-scored, audit-traceable.
SaaS
A clean web app. Open it, type a question, get a packet.
REST API
Wire Thesis Lab into your existing research pipeline.
MCP
Plug into your own agent stack - Claude, ChatGPT, Cursor, internal bots. Connect via Claude Code →
MS Teams
@-mention Thesis Lab in any channel. Packets land where the conversation is.
Slack
Ask in-thread, get a packet in-thread.
Built for the Buying Committee
The analyst champions. The PM signs off on calendar arithmetic. Compliance signs off on the risk surface. Data and research-tech sign off on integration and provenance. Thesis Lab is engineered for every conversation.
Compliance
Zero MNPI by Construction
Synthetic agents have no insider knowledge. There is no material non-public information surface to manage, log, or monitor. Compliance approves the tool because of the architecture - not in spite of it.
Audit
Tighter Audit Trail
Every cohort decision, every claim citation, every interview transcript captured in ai_agent_decision_events. When the SEC asks for research provenance, you answer in seconds.
Accuracy
Benchmarked, Calibrated, Published
We benchmark the model continually against prediction-market bets across 18 dimensions. You see where the model is strong, where it's weak, and how we're adjusting it. Read the validation methodology →
SOC 2 · SSO · TENANT ISOLATION · ZERO TRAINING ON CUSTOMER INPUTS · SIGNED DPA · PUBLISHED METHODOLOGY
Six Lines We Earned From Six Weeks of Research
Verbatim from named analysts at hedge funds, credit funds, financial-advisory firms, management consulting, and risk teams. Want the long-form view? Read the hedge fund case study.
Credit Analyst
About 70% gathering / absorbing, 30% actually forming a view and pulling the trigger.
- Nolan Navarro
Business Analyst
If the first packet saves me 45–60 minutes and nails one non-obvious gotcha I would have found on day two, I am in.
- Tianna Burquez
Financial Advisor
I want the office manager, dispatcher, or ops lead who closes the day and fights the fires.
- Krista Nina
Risk Analyst
I write the bear case - five bullets, each tied to a line item. If I cannot rebut with data inside 24 hours, the rec gets downgraded.
- Brittany Johnson
Hedge Fund Analyst
Three-proof rule. I can verify 3 claims in 5 minutes. Each has a line-level source, under 18 months old, with at least one negative datapoint.
- Jesus Ciullo
Compliance Analyst
Give me footnoted, Minnesota-aware, numbers-forward packets that slot into my memo, and keep your hands off my data. I will pay.
- Andrea Henderson
Position Against Your Expert-Network Spend
A typical mid-tier hedge fund spends $200K–$1M annually on expert networks for 5–15 analysts. Most of it produces colour, not signal. Thesis Lab replaces 50–80% of that line, delivers 5–10× the volume, and arrives with a tighter audit trail than what you have today.
The Math Your PM Can Write on a Napkin
Pilot pricing is a fraction of one expert call. Annual pricing is set per fund - tell us your current expert-network spend and we'll give you the ROI math before the demo.
The Questions Every Fund Asks Us
We're doing this ourselves with ChatGPT / Claude.
Great. How are you handling the data-staleness issue - every frontier model has an August 2025 training cutoff. Are you basing earnings-cycle decisions on eight-month-old data? Thesis Lab's panel is grounded in continuously-refreshed DB-attested data. Also: your compliance team cannot defend “I asked ChatGPT and it said” to a regulator. Every Thesis Lab interview is recorded, cohort-traceable, with full decision-event audit trail.
I don't want another generic thesis from an LLM.
Good - we're not selling that. We're selling access to expert panels in minutes so you can build your own thesis yourself. Use humans, use your own agents, plug in via API or MCP - the output is yours.
How do we know it's accurate?
We benchmark the model continually against a broad range of prediction-market bets, categorised across 18 dimensions. The benchmark reveals where the model is stronger or weaker; we adjust accordingly and republish. On the analyst side: every packet survives a three-proof rule - pick any three claims, each has a line-level source under 18 months old with at least one negative datapoint, verifiable in five minutes. If you can't verify three in five, we don't deserve your time.
Do we need to learn another tool?
No. SaaS interface for the analyst who wants to open a browser. API for the engineering team that wants to wire it into existing pipelines. MCP server for any agent stack - Claude, ChatGPT, Cursor, internal bots. Microsoft Teams and Slack integrations for the channels you’re already in.
Synthetic personas - sounds like make-believe.
We don't lead with the panel. We lead with the packet. The first packet has 30 sourced claims, 5 counter-arguments with evidence, and 5 kill questions for management - every claim mapped to a clause-level citation. Verify any three in five minutes. If we can't deliver that, you don't pay.
Compliance won't approve it.
Three points. Zero MNPI risk by construction - synthetic agents have no insider knowledge. Zero training on your inputs, SOC 2, SSO, full audit logs, tenant isolation, signed DPA. The audit trail in ai_agent_decision_events is better than what you have on expert-network calls today. Most funds finish the compliance review faster than they finish the analyst trial.
Can you really get the dispatcher, the office manager, the supplier rep?
Yes - that’s structurally what’s different. Expert networks gatekeep to senior, polished experts. Our 301K-agent pool is grounded in BLS occupational distribution - every occupational tier, every industry. We recruit cohorts of “shift-manager-tier dispatchers at mid-market trucking firms in the southeast” that no expert network can build.
We tried an AI research tool last year and it flopped.
Likely hallucinated citations, missing local context, or over-promised synthesis without provenance. Three things we do differently. Every claim is clause-cited and verifiable in under 60 seconds. The output format is exactly what you’d put in front of an IC - no reformatting, no PDF lock, CSV/Excel/Notion-ready. The pilot is priced as a fraction of one expert call: if the first packet doesn’t save 45–60 minutes and surface one non-obvious gotcha, you don’t continue.
LLMs are coming to replace analysts. Why pay for this?
LLMs aren't replacing analysts. They're replacing expert networks. Analyst value-add is judgement under uncertainty - that becomes more valuable as the time-cost of grunt-work falls. The fund that adopts now is the fund where each analyst covers 3–5× more theses per quarter.
What about the FishDog brand - does that fly in finance?
The customer-facing brand is Thesis Lab. FishDog is the parent and appears only in legal filings. Every contract, packet, integration, and customer artifact you receive is Thesis-Lab-branded.
More on Thesis Lab
Read how a buy-side team used Thesis Lab in the field, plus the latest from the product team.
Get Your First Packet Before Your Next IC Meeting
Tell us the thesis. We deliver the first Thesis Interrogation Packet in under two hours, free, on a research question of your choosing. If it doesn't save you 45–60 minutes and surface one non-obvious gotcha, walk away.
Free Pilot
Pilot Packet
On a real research question of your choosing. Delivered in under two hours.
Sample
See a Sample Packet First
Real packet from a real research question, fund and ticker scrubbed. Emailed to you.
Replies within one business day. No sales call required for the pilot.