The Augmented VC Index

Running since 2016 (original architect: Henrik Landgren, ex-Spotify VP Analytics). Tracks ~50M companies; logs internal assessments and employee network connections — a decade of captured judgment no competitor can buy. As of EQT's 2024 disclosures, nine investments were described as fully Motherbrain-sourced — including Peakon (acquired by Workday for $700M, 2021 — third-party confirmed via Workday IR), AnyDesk, Handshake, Standard Cognition, Netlify — and, honestly disclosed, one failure: WarDucks, which shut down in January 2024. EQT has said the Peakon exit alone covered the platform's costs. Since 2024, Motherbrain expanded group-wide across EQT's PE and infrastructure business; in late 2025 EQT described an agentic overhaul ("deep agentic assessments of opportunities" — Tech.eu, Nov 2025) and named compute "the main bottleneck." A telling diaspora note: Landgren left to co-found ArK Kapital (AI-driven growth lending), which pivoted and rebranded to Gilion — even Motherbrain's architect could not simply repeat the trick.

Why A: the A tier requires an attribution record (R3) — multiple named machine-attributed investments with dated outcomes, failures and write-downs included where applicable, a stated selection basis, and at least one outcome independently confirmed. EQT is the only firm worldwide that meets it: nine investments described as fully Motherbrain-sourced as of its 2024 disclosures, including WarDucks (shut down 2024), with Peakon's $700M Workday exit third-party confirmed as an exit. The machine-sourcing link itself is EQT's own account, as it is for every firm in this Index.

Funds: €566M (2016), €660M (2019), €1.1B (2022). Sources: eqtgroup.com/about/motherbrain · Workday IR (Peakon) · Tech.eu Nov 20, 2025 · Asymmetrix investor analysis Mar 2024.

Earlybird (founded 1997, €2.5B AUM, Fund VIII €360M closed April 2026) built EagleEye across sourcing, screening, diligence, and portfolio support. The attribution case: by the platform team's own detailed account, EagleEye captured Aleph Alpha from its Handelsregister registration in 2019 and re-prioritized it in late 2020 when online signal data improved its score; Earlybird then led the €23M Series A in July 2021 alongside Lakestar and UVC Partners. Aleph Alpha's later "$500M+" Series B (Nov 2023) needs the standard caveat: the officially disclosed breakdown put pure equity at ~€110M of a ~€470M package including research funding and order commitments. EagleEye's own performance numbers — "200% team efficiency," "95% coverage of relevant opportunities," "10M+ companies assessed" — are marketing figures without published methodology. The best process quote in the field, from the platform's builder: "You only get one chance to change the workflows of your investment team."

Why B+: a single named sourcing case (Aleph Alpha), told in one first-person account, with a Series A co-led alongside Lakestar and UVC Partners (July 2021) and the Series B equity-composition caveat. Under the consolidated attribution rules this is R1 — the same tier as InReach's Oberlo case, which additionally carries an SEC-confirmed exit. Single-deal attribution does not reach the A tier (R3).

Sources: earlybird.com/eagle-eye · the team's Nov 2023 Medium account of the Aleph Alpha sourcing · Tech.eu (Series A, July 2021) · The Decoder (Series B breakdown).

The largest firm built AI-first from inception (2013, Chris Farmer ex-Bessemer; CTO Ilya Kirnos, early Google). ~$3B AUM after a $1B+ fund closed April 2025 — including CalPERS' first-ever commitment to the firm ($100M). Beacon spans sourcing, portfolio go-to-market, and talent; self-reported scale: 80M companies, 650M people tracked, roughly half a trillion data points (the firm's own pages round it differently in different places — treat as marketing-scale). Cost is the credible part: ~$80M over 12 years (PitchBook); in the first $50M fund, "AWS costs equaled half our management fee," and one credit-card dataset cost more than the entire fee (Farmer). The grade ceiling: SignalFire names no deals sourced by Beacon and says it deliberately runs no attribution system. Best externally reported outcome: 44 of 46 head-to-head term sheets won over 2.5 years (Forbes, 2018). Farmer's line remains the category's mission statement: "We look more like Uber or Bridgewater than anyone on Sand Hill Road."

Sources: Businesswire Apr 7, 2025 · Bloomberg Apr 2025 · TechCrunch Apr 2025 (CalPERS) · Forbes Oct 2018 · signalfire.com/beacon-ai.

Data platform since 2014; ingests credit-card and enterprise-expense data; >$30M/year on data science. Holds the best-documented negative screen in the field: Coatue passed on HeadSpin after Mosaic showed usage weaker than claimed — the founder was later charged with fraud. AI's clearest documented "save" is a deal not done.

Sources: blakeir.com deep dive · Domino Data Lab case study.

Europe's original AI-first VC (2015, Roberto Bonanzinga ex-Balderton; more engineers than investors from the start; >€3M spent on DIG by 2019). By the firm's May 2025 account: "75% of all investments at InReach Ventures have been sourced through our unique technology platform." The attribution case, documented end to end: InReach's platform-driven cold outreach to the Vilnius founders of Oberlo is the firm's own account (Bonanzinga, in Q&A quotes within TechCrunch, Feb 2019); the outcome is independently confirmed — Shopify acquired Oberlo in May 2017 for $17.2M in cash (Shopify Q2 2017 SEC filing). That makes InReach one of two firms (with EQT) holding an R2 attribution: a named machine-attributed deal plus a confirmed exit. Fund II investing as of 2025 (size undisclosed). Bonanzinga has also written the category's sharpest provocation (Feb 2026): sourcing, screening, and diligence get automated, firms shift to "intellectual leverage over organizational mass," a "$1B solo-GP fund" is coming — "The agents are coming. And they will not ask for carry!"

Sources: TechCrunch Feb 2019 · Bonanzinga on Medium, May 20, 2025 · VCWire Feb 11, 2026.

Founded 2021 by Mattias Ljungman (Atomico co-founder) as "a technology company that does venture capital"; ~10 people, more engineers than investors, engineering led by GP Mike Arpaia (ex-Facebook, creator of osquery). Reviews up to 50,000 companies per week by its own estimate. The generational bridge datapoint of the whole index: wiring GPT-4 into their venture-scale classifier took an afternoon, improved results ~20%, and outperformed 50–70% of the custom models they had built over years — the moment rented intelligence beat owned intelligence. Funds: $60M (2021), $115M (2023, 95% LP re-up); Fund III filed with the SEC May 2026, size not yet public. Attributed sourcing: LiveFlow.

Sources: Sifted May 2023 · Moonfire engineering blog · SEC Form D/A May 2026.

$1.6B AUM; Nokia-backed. Q has run since ~2019: an internal platform for discovery, monitoring, and portfolio support, tracking millions of startups and producing ~500 recommendations a month (firm figures). The best-documented engineering account in its tier: a November 2, 2022 firm post describes the ML models and the operating philosophy on the record — "Q is an aide, not a decision-maker... we would never make an investment based solely on Q." Current site: 70% of pipeline originates from Q (undated, denominator unstated). Externally covered since 2020 (then ~30% of dealflow). No named Q-sourced deal appears in any source we found (as of August 2026) — the sharpest example of a strong platform stopping below rung three.

Sources: ngpcap.com/q-ngp-dealflow-engine · ngpcap.com/insights/three-years-of-q (Nov 2, 2022) · Sifted, Aug 13, 2020.

Deeptech growth; ~€1B+ cumulative funds (Fund IV €371M; Fund V €600M first close, Dec 2025, EIF-anchored). Ninja, developed since 2016: deep learning/NLP over 5M companies (one added per minute, firm figures), 7.8M patents, proprietary Deep Tech/IP/Team scores. Claims: "75%+ of deals in Jolt's portfolio are sourced by Ninja"; "100% of Jolt's deals are managed through Ninja." What makes Jolt unusual: external institutional corroboration — Québec's Fonds de solidarité FTQ signed a licensing partnership (press release, May 7, 2026) giving its partner funds access to Ninja, and Fund V is marketed on the platform. A counterparty institution putting its name behind the system is the strongest external validation of any platform in this Index. No individual deal is named as Ninja-sourced in any source we found.

Sources: jolt-capital.com/ninja · fondsftq.com press release, May 7, 2026 · Tech Funding News, Dec 1, 2025.

The first mega-brand with first-party deployment documentation: "Meet the Agents" (blog.usv.com, Spencer Yen, March 25, 2026) describes a custom internal database and named agents — meeting-prep, deal analysis, people/company records from email and calendar context, organizational memory — built on Tasklet (a USV portfolio company, disclosed in the post). The post is unusually honest: placeholder data in screenshots, renamed agents, AI-written agent introductions, all flagged. USV posted a dedicated AI Lead role in 2025 ($150–300k). No outcomes or attribution claimed. Taxonomy note: USV is an agent-adopting established firm, not Agent-Native — founding era and operating architecture are classified separately.

Source: blog.usv.com/meet-the-agents (Mar 25, 2026).

The original quant VC (2010). Pure co-investment: never leads, no board seats; a predictive model scores every deal against a database of (claimed) virtually every US venture financing outcome; decisions in days, always under two weeks. Fund III $130M (June 2023), ~$500M AUM, 400+ portfolio companies. Sixteen years in: returns never published, no Fund IV announced as of August 2026, low visibility since. The fair reading: the model survived — it just didn't take over.

Source: PRNewswire Jun 27, 2023.

The most concrete AI-analyst deployment anywhere. "Viki," deployed April 2025: first-pass analysis of market, competition, technology, traction; due diligence cut from ~30 days to ~3 (by mid-2026, reportedly 24 hours); 87.5% agreement with human analysts' conclusions (firm's figure, Feb 2026). Decision stays with founder-turned-partners — a deliberate hybrid. February 2026: first Korean VC to partner simultaneously with OpenAI, Google, and Anthropic for portfolio credits. Caveats: self-reported metrics, small fund, friendly trade press.

Sources: Unicorn Factory Apr 2026 · KoreaTechDesk Feb 2026 · HelloT Jul 2026.

The best-documented Workflow Adopter globally — proof that the buy-don't-build path can be done seriously. $2B FUM; published, named internal agent skills (Board Meeting Prep reads board packs, searches Slack, checks competitor news, tracks commitments; Board Summary: 30 minutes → 2; market research: 2 days → minutes) wired via MCP into Notion, Slack, Affinity, Harmonic, Granola. Partner Jackie Vullinghs supplies the index's thesis in one sentence: "The value of AI in a workflow like ours isn't replacing judgment; it's in protecting the time."

Source: Claude customer story (claude.com/customers/airtree).

Two named systems built over 12+ years: Searchlight (sourcing; 7M+ companies tracked; credited discoveries: Segment, SEMrush, Bitwarden — notably all US) and Deepdive (diligence on Stripe/Chargebee payment data, opened free to founders). 17 engineers and data scientists on the platform team — disclosed, which is rarer than it should be.

Source: Headline's own platform blog.

Founded 2014 by Nnamdi Okike (ex-Insight) and Aaron Holiday (ex-Goldman engineer); Voyager does metrics-based sourcing/screening; ~$555M AUM after $347M raised in Dec 2022; outcomes: FiscalNote (public 2022), Panther Labs.

Source: TechCrunch Dec 6, 2022.

The purest Machine Screener still operating: founders interact with Wendal for ~20 minutes, no pitch deck; 92% get decisions within 3 business days. 25,000 founders assessed (own figure). The most interesting claim in the whole index on bias: 46% of portfolio CEOs are women or minorities — roughly 3x the industry average — attributed to removing humans from the first screen. Social Capital's 2017 CaaS screen reported a related but distinct figure (42% of funded-company CEOs were female): the two measures use different populations and definitions — portfolio CEOs "women or minorities" vs funded-company CEOs "female" — and stand as two separate results. October 2024: launched $VCAFX, a Nasdaq-listed interval fund opening VC to non-accredited investors.

Sources: conneticventures.com · LINK nky Oct 3, 2024.

$2B+ raised (Dawn V $620M + Opportunities III $80M, Sept 2023). Rolodex, disclosed 2024: a data-and-AI relationship-intelligence platform in three modules — Intel (automated meeting insights), Coverage (emerging-investor tracking), Access (an LLM over the firm's relationship graph for warm paths to founders). Dedicated hires: data & AI strategy lead (Jan 2024), data engineer (Jul 2024). The evidence is capability-level and self-sourced (the firm's own lead, via a Forbes contributor column); one access win is described — a term sheet after cold outreach failed — but anonymized. Analytically important: Dawn's platform aims at access, not discovery — a reminder that sourcing is not the only path AI takes into the process.

Sources: dawncapital.com 2024-in-review · Forbes (contributor), Jul 30, 2024 · CNBC Sept 25, 2023.

An internal data platform (company lookup app plus ML-driven dashboards blending external and proprietary data) is described in a vendor case study with a named executive (Sarah Guemouri, Head of Insights), including the claim that "1 in 3 net-new opportunities" originate from the data engine. Vendor-published, firm-supplied figure, no named deals.

Source: Crunchbase customer story (Atomico).

Small US pre-seed fund (~$9.5M Fund I; Yohei Nakajima, also known for BabyAGI, and Jessica Jackley). A March 31, 2025 first-person operating account (itself AI-drafted from a talk transcript and author-edited — disclosed) documents deployed workflows: inbound deal information extracted into a pipeline, internal pitch review against a custom GPT, AI-assisted diligence, portfolio updates extracted into Airtable — with experiments explicitly separated from tools in regular use. Independently profiled (GeekWire, Aug 27, 2025). The grade rests on the documented workflows, not on open-source fame. Distinct entity from Untapped Ventures.

Sources: yoheinakajima.com/building-a-personalized-vc-copilot (Mar 31, 2025) · GeekWire, Aug 27, 2025.

Founded ~2019 (Ziad Makkawi); systematic private-market investing via an "EQUIAM Systematic Ranking" engine. The only cohort-level performance disclosure in this Index: Private Tech 30 Fund I (vintage 2019) — DPI 1.86x, TVPI 1.93x, net IRR 27.5%, as of December 31, 2025, published on equiam.com (public since at least June 12, 2026). Held below the A tier and outside the rung-five vacancy claim for three stated reasons: (1) Fund I was 100% secondary-market purchases — not primary venture investing (and 2019-vintage late-stage secondaries reach DPI structurally faster, which flatters the mixed benchmark comparison); (2) the figures are manager-published with no audit qualifier, against a pooled buyout+growth+VC universe dated one quarter earlier than the fund's own mark, with the "top 1%" claim selected on DPI alone (its TVPI sits at the upper-quartile line of the same cohort); (3) EQUIAM's own current method description says "systematic" and does not claim AI or machine learning. Included — like Coatue — as a labeled comparison entry, because the Index cannot exclude a quant secondaries fund for being quant while including Correlation.

Sources: equiam.com/benchmarks · equiam.com/approach · Business Wire, Jun 10, 2021.

The Social Capital quant lineage (founded 2018 by Sethi/Hsu/Maidenberg; the "Magic 8-Ball" PMF framework; Hsu's honest framing: "We use data like accountants; it's not a magical AI machine"). 2024: engine productized as Termina — quantitative diligence sold to pensions, sovereigns, PE; ~1,500 companies' transaction data; supported Kraken's $1.5B NinjaTrader acquisition diligence. Qualifier: Sethi is now co-CEO of Kraken; leaked 5x Fund I claims were never audited. The strategic reading: Tribe concluded the quant engine is worth more as a product than as an edge — a data point against the proprietary-moat thesis.

The biggest new story in the category: Nik Storonsky's systematic fund ("limited human involvement") doubled from $250M (May 2025) to an oversubscribed $500M Fund II (Aug 2026); 27 companies, 5 claimed unicorns; Harvard Business School already teaches the case. Grade capped hard: earliest positions are under three years old. LPs are buying Storonsky's brand plus a promise — DPI will grade this fund, not press releases.

At 2017 launch: ML engine surfacing 1,000+ companies weekly, ~60% of deals from cold sourcing. €80M Fund III (Dec 2024, oversubscribed) — but the engine has vanished from recent messaging in favor of "technical founders/deeptech" (Wayve the flagship). A quiet, ungraded retreat worth asking them about.

Anand Rajaraman & Venky Harinarayan (Junglee, early Facebook); 100% outbound ML sourcing over "50M+ startups"; Fund III $125M first close 2023; heavy emerging-markets allocation — the model's real distinction: "Data helps us invest beyond geographies." Unicorn claims self-reported.

$25M AUM but methodologically the most transparent firm in the field: publishes algorithms on GitHub; its team lead-authored VCBench, the first public LLM startup-prediction benchmark.

Builds software facing founders (Signal network graph; BriefLink; 9-day FAST decisions) — augmenting deal flow rather than deal judgment.

Finch describes an internal sourcing platform using NLP and deep learning across 30 languages, integrated with its CRM. In February 2021, it reported that Flowrence had sourced 20% of shortlisted deals over the preceding six months — shortlisted opportunities, not completed investments. Documented historical capability; metrics self-reported and current operating status unverified. Source-date note: the firm-hosted "Flowrence, Sourcing Reinvented" post is dated January 26, 2021, with the launch reported as 2020; the sourcing figure comes from the firm's February 9, 2021 fund announcement. No named machine-attributed investment.

Sources: finchcapital.com, "Flowrence, Sourcing Reinvented" (Jan 26, 2021) · Finch Capital fund announcement, Feb 9, 2021.

Pan-European seed firm. A July 22, 2020 first-person account by a Speedinvest team member (Alex Zhigarev) documents a deployed, low-code RPA sourcing system built with four named investment-team members: ~2,500 opportunities sourced in three months, 324 added to the CRM, ~15% progressing through early dealflow steps, and a 27% progression rate for the best-performing source — self-reported operating results, not investment returns, with human review kept explicitly in the loop. Separately, an undated vendor case study describes internal AI work at Speedinvest (CRM and ERP connections, dealflow workflows, email and meeting context, an internal AI lead); it is vendor-published, its date and its relationship to the 2020 system are unestablished, and it is not the basis of this grade — the two accounts are kept separate until reconciled. Grade: provisional B–, on the documented 2020 workflow; current deployment requires confirmation. No named machine-attributed investment.

Source: Zhigarev, "How We Use Robotic Process Automation to Source Companies" (Medium, Jul 22, 2020).

Founded 2020 (Joe Aaron, Fred Campbell, ex-CrowdSmart); >$80M AUM as reported to press; $65M invested across ~120 companies (firm figures); never leads; checks up to ~$2M. Site claims, verbatim: "TRAC uses Machine Learning and Deep Learning algorithms to make every single investment"; "There are no humans on our investment committee"; "<8.5% Loss Ratio"; "Top 5% Graduation rate"; "Top Decile DPI: Fund 1 (2020), Fund 2 (2022)." No numeric performance figure, as-of date, or benchmark methodology is published anywhere we searched (as of August 2026). Press coverage (Business Insider, Sept 2024; Goldfisher, Mar 2025) relays the model as described and simultaneously documents humans throughout the process: founder meetings, sourcing relationships, access work, and an attempted override of the model by a co-founder. TRAC matters because it is the loudest full-authority claim in the field — and the cleanest illustration of the gap between claimed and verified authority.

Sources: trac.vc · Business Insider, Sept 2024 · Goldfisher, Mar 28, 2025.

The Gen-2 exemplar: launched Sept 2025, €30M first close, explicitly designed post-LLM with agents across sourcing, DD, portfolio support. Graded on evidence volume, not age: capability described, nothing attributable yet — the firm to re-grade in v2.

True Ventures-backed ($11.5M, 2025); AI agent interviews founders, tests customer acquisition, deploys $25K checks; 15% perpetual revenue share instead of equity; target "100,000 companies a year." Closest thing to an AI GP — and a governance question mark.

Fund-as-a-service model. SOFIA is described as an in-house "VC data intelligence platform" spanning sourcing, screening, investment analysis, and portfolio support, with stated goals ("invest smarter, reduce bias"). No technical detail, no dates, no users, no metrics, no named deals, no press found; a live login portal (sofia-redstone.vc) is the only independent trace of deployment.

Source: redstone.vc/sofia.

Publicly recruiting "AI-native talent, including researchers and engineers with machine learning backgrounds" (Feb 2026) — a leading indicator, not yet a system.

"$50M can do what used to take $500M" — nothing but claims yet.

$49B fund closed July 2026; "As an AI native investment company, we are leveraging AI in everything we do" (CEO Ahmed Yahia, Jan 2025) — no named platform, no metrics, no engineering disclosure.

Graded C on AI: its real asset — arguably the world's largest proprietary dataset on pre-company founders, one investment every ~22 hours in 2024 — is undocumented as an AI system.

The world's biggest AI allocator with no documented internal AI diligence; the canonical stories (minutes-long gut decisions) run the other way.

Built ~2009; scored deals 0–10, green/yellow/red; evolved into a de facto investment committee with veto power ("It would be foolish to just go out and make gut investments" — Bill Maris, 2013). Shelved because: investors learned to game inputs; software updates retroactively changed portfolio scores; it blocked follow-ons during downturns. Axios' epitaph: data "relegated to its original role as aide, rather than arbiter." Lesson: the failure mode isn't bad predictions — it's giving the model authority without accountability.

The D grade and this entry cover the historical system. GV-the-firm today is ungraded; a current-state search is an open register item.

Oct 2017: automated funding, no pitch meetings; ~3,000 companies evaluated, several dozen funded across 12 countries; 42% of funded-company CEOs were female, the majority nonwhite. The screen demonstrably widened the aperture — but whether it picked well was never tested at scale; the firm collapsed first, of partner exodus and implosion, not model failure. Its team became Tribe Capital. Lesson: the platform doesn't survive the partnership.

CSC Group's Silicon Valley arm; ML model on 30,000+ deals; claimed 40% of model-picked seed deals raised follow-ons vs ~16% industry baseline (McKinsey interview, June 2017); deployed $400M via AngelList. Then: years of litigation with ex-executives, and a reported FBI probe (TechCrunch, Sept 2024) into Chinese-government funding and startup-data sharing. Lesson: the model can be right and the firm still dies — of everything else.

The founding myth of the field, negatively: "appointed an algorithm to its board" — in fact fuzzy logic over 50 parameters, observer status, no vote (HK law requires natural-person directors), "publicity hype" (AI researcher Noel Sharkey, 2014), and no longer in use by 2019 (Fortune). The website has been dormant since 2015. Lesson: the distance between the press release and the capability is the oldest constant in this field.