How it works
I research anything you need. Here's how, on one example.
You give me a list: companies, people, funds or markets. Agents research every row, in the fields you need. I check every value, or leave it open with its reason. You get it where you already work. Below, one VC firm goes through every step. Then a real result.
Step 1 of 6: Your fieldsStep 2 of 6: I split the workStep 3 of 6: SearchStep 4 of 6: Check each valueStep 5 of 6: Settle disagreementsStep 6 of 6: Your data
Your question becomes the fields every company in the dataset must fill.
Research this VC firm.
Your fields
For one VC firm the work splits into tasks: find the fund, its investments, cheque sizes from its past deals, people, and what each portfolio company does.
Research this VC firm.
- Find the fund
- Its investments
- Cheque sizes from past deals
- People
- What each portfolio company does
Paid databases run on your own accounts, so the data rights stay clean.
Sources searched
- Fund websiteOpen web
- Portfolio company websitesOpen web
- News and announcementsOpen web
- Company registriesOpen web
- Your CRM and notesYours
- Paid databasesYour accounts
Separate steps guard against made-up answers: one drafts the value from the sources, one looks for what the draft got wrong, and one keeps the value with its sources or drops it.
- Stage: Seed, Series AKept: the fund's website and two announcements
- Sectors: Fintech, B2B softwareKept: the portfolio page
- People: 4 partnersKept: the fund's team page
- Thesis: Consumer appsDropped: a forum post contradicts the fund
Records that name the same company become one, and where sources give different values, I keep one with every source recorded.
Company names
Cheque size
The data is yours, delivered where you already work.
Northfield Capital
| Field | Value · source | Level |
|---|---|---|
| People | 4 partnersnorthfield.example/team | High |
| Sectors | Fintech, B2B softwarenorthfield.example/portfolio | High |
| Stage | Seed, Series Anorthfield.example/about | High |
| Cheque size | $500k to $2.3Mnorthfield.example/about | Medium |
| Thesis | Infrastructure for finance teamsnorthfield.example/blog | Medium |
| Deals | Ledgerline, led by Maria Lind or Tomas Reyesledgerline.example/news | Open |
A real result: Insight Partners.
Researched from public sources, as of 18 October 2026. Every value has its source or how I counted it, and how sure I am.
Insight Partnersinsightpartners.comActiveGlobal36 cited claims
Profile
- Cheque size
- $5M–$500M+insightpartners.com“Insight’s flagship funds, including Fund XIII, typically invest between $5 million and $500 million+ in a company, with the ability to invest from the earliest institutional check to IPO.”High
- Stages
- From the first institutional cheque to growth and late stage
insightpartners.com: “Insight’s flagship funds, including Fund XIII, typically invest between $5 million and $500 million+ in a company, with the ability to invest from the earliest institutional check to IPO.” High - Strategies
- Venture and growth · Structured equity
insightpartners.com: “Insight Partners is a global software investor partnering with high-growth technology, software, and Internet startup and ScaleUp companies” “The Opportunities Fund provides companies and their investors with alternative financing solutions, typically in the form of structured preferred equity investments” High - Top sectors
- FinTech 40 · Cybersecurity 37 · HealthTech 34 · Data & Analytics 20Counted from the 286 companies I tracked.Medium
- Team
- 11 people I tracked, including 7 Managing Directorsinsightpartners.comThe people I tracked, not the whole team.Medium
Funds
| Fund | Size | Year | Source | Level |
|---|---|---|---|---|
| Opportunities Fund I | $1.56B | 2021 | High | |
| Fund XI | $9.5B | 2020 | High | |
| Fund X | $6.3B | 2018 | High | |
| Fund IX | $3.29B | 2015 | High | |
| Growth-Buyout Coinvestment Fund | $1.46B | 2015 | High | |
| 20/20 Vision Capital Fund II | $118M | MediumNo year stated. |
Track record
- Deals tracked
- 89; led or co-led 72% (55 led, 9 co-led)From 89 deals I tracked, not its full history.Medium
- Median round size
- Series A $34.5M (16 rounds) · B $35M (25) · C $55M (16)Whole rounds with a disclosed size.Medium
- Portfolio
- 286 companies tracked, 24 of them exited. Top countries: US 201 · Israel 20 · UK 15 · Germany 13
insightpartners.com: “Insight Partners has invested in more than 900 companies worldwide and has seen over 55 portfolio companies achieve an IPO.” Insight says 900+ companies; I tracked 286, each with a source.Medium
Recent deals
| Company | Round | Round size | Role | Date | Source | Level |
|---|---|---|---|---|---|---|
| Higharc | Series C | $95M | Lead | 30 Jun 2026 | higharc.com | High |
| ScaleOps | Series C | $130M | Lead | 30 Mar 2026 | insightpartners.com | High |
| Rocketlane | Series C | $60M | Lead | 25 Mar 2026 | rocketlane.com | High |
| Wonderful | Series B | $150M | Lead | 12 Mar 2026 | techcrunch.com | MediumOne news report. |
| Ayar Labs | Series E | $500M | Participant | 3 Mar 2026 | ayarlabs.com | High |
Counts are what I tracked, not the firm's full history. Round sizes are the whole round, not Insight's own cheque.
- High
- Sources that didn't copy each other agree, or an official record says so.
- Medium
- A good source backs it, and I say why the others differ.
- Open
- No source settles it yet. You see every answer and the reason.
When sources disagree, I find out why.
Often they don't really disagree: they mean a different round, date or company. I check that first. Then the most credible source wins, and I check it against what really happened.
- Fund websiteAbout page, 2023
- $500k to $2M
Used
- NewsWhat it put into four deals, 2024 to 2025
- $750k to $2.3M
Used
- Data providerProfile, last updated 2021
Not used: $100k to $500kOut of date, not used
Combines the fund's own range with what it put into four deals.
- Fund websiteAbout page, northfield.example
- Northfield Capital
Used
- Official company recordsFiled in 2022
- Northfield Capital LP
Used
- NewsArticle about a deal, 2024
- Northfield Cap.
Used
All three names use one website and one address on file.
- LedgerlinePress release, March 2024
- Maria Lind
Disagrees with the news
- NewsArticle about the deal, April 2024
- Tomas Reyes
Disagrees with the press release
- Fund websiteTeam page, 2025
- Both shown as partners
Lists both, not used
Two sources name different partners. No source shows who led.
Details of a deal
Sources often give a round different amounts or dates. First I check if they mean the same thing: announced, closed or filed. Official filings and the people in the deal outrank news and databases.
Company names
One company often goes by several names. I join them only when a registry number, a website or an address matches. If I'm not sure, I keep them apart and tell you.
Other facts
What a fund says and what it does can differ. I check its statements against its real deals, and show both when they differ. When no source settles it, agents reason it through from several sides, and the answer is marked Medium or Open.
Open isn't the end: a second pass, or research on connected facts, often settles it.
Every step is tested, not guessed.
For each step I try several approaches. Agents test them side by side on the same data. I keep the one that wins on measured results. On hard calls, several runs with different models check each other.
Research once, then keep it running.
Companies change every week. I rerun the same research on the same rows. Every change is marked, with its source.
When
- On a scheduleEvery few hours, days or weeks. As often as your data changes.
- On a news signalA funding round or a new hire. I research the companies involved.
- On your signalA new deal in your CRM, a newsletter mention, or any app.
Where
Your inbox, your CRM, Slack, your database, a file, or a page I host for you.
For developers: API or MCP.
Deep research, cheap enough for your whole list.
One example: a pipeline I built to research a VC firm in full.
- 5,000
- VC firms researched in depth.
- $0.73
- to research one firm in full.
- About 37
- agents on one firm.
- Up to 1,000
- agents at once.
Measured on the VC research system I built at ResearchTech. About 437 model calls per firm, 93.9% prompt-cache hit rate, 131,000 portfolio companies behind the 5,000 firms.
Your data stays yours.
No training
AI providers keep none of your data and never train on it.
Hosted by me
It runs on my servers, so you host nothing.
NDA
An NDA before the call, if you ask.
Your accounts
Paid databases run on your own accounts.
Yours to keep
The data is yours. I can hand over your pipeline. My core tools stay mine.
The benchmark against other tools is running.Until it's done, your pilot is the real test:I compare my results with what you use now, on your own list.
