Breaking
VECTICE raises $15.6M seed + Series A led by Sorenson & Crosslink EU AI Act live — auto-documentation demand climbs Integrates with ServiceNow IRM, Sept 2025 IBM watsonx.governance ties cut review cycles ~50% Founded 2020 · San Francisco · ~12 employees Member: NIST AI Consortium & AI Verify Foundation Docs mapped to SR 11-7, ISO 42001, NIST AI RMF VECTICE raises $15.6M seed + Series A led by Sorenson & Crosslink EU AI Act live — auto-documentation demand climbs Integrates with ServiceNow IRM, Sept 2025 IBM watsonx.governance ties cut review cycles ~50% Founded 2020 · San Francisco · ~12 employees Member: NIST AI Consortium & AI Verify Foundation Docs mapped to SR 11-7, ISO 42001, NIST AI RMF
Company Profile — Enterprise AI & Governance San Francisco

Vectice

The company that automated the least glamorous problem in artificial intelligence: proving your models to an auditor.

Founded2020
HQSan Francisco
Funding$15.6M
CategoryRegulatory MLOps
Vectice logo - a yellow cube mark beside the wordmark 'vectice' on navy
The vectice mark: a cube assembled from parts, on the navy the company uses everywhere. Look closely and it reads as building blocks - which is roughly what documentation is, one piece of evidence stacked on the next.
Share this profile LinkedIn Twitter / X Facebook Instagram

Here is a fact about artificial intelligence that nobody puts on a conference slide: a bank cannot deploy a machine-learning model just because it works. It has to prove the thing works, and keep proving it, in a stack of documents thick enough to satisfy a regulator who was hired specifically to be unsatisfied. Vectice is a company built entirely inside that gap.

Vectice, photographed as a logo on its native navy. The cube is the tell - four faces you can see, a couple you have to infer. Which is a fair portrait of the product, too: the model is the part everyone looks at; the evidence underneath is the part that keeps the whole thing standing.

Let's start with the premise, because the premise is the whole company. When you build a machine-learning model at a bank or an insurer, the model is maybe a quarter of the job. The other three quarters is documentation: where did the data come from, who touched it, what version of the code ran, how was it validated, why should anyone trust the number it spits out. In banking this discipline has a name - model risk management - and it has had regulatory teeth since a 2011 Federal Reserve letter with the deeply catchy title SR 11-7. The rules predate the current AI boom by more than a decade. The AI boom just made them everyone's problem.

The traditional way to satisfy those rules is that a data scientist, some weeks after finishing a model, opens a Word template and begins the joyless work of reconstructing what they did from memory and screenshots. It is slow, it is error-prone, and - this is the part Vectice noticed - it is exactly the kind of task that software should do and humans should not.

The pitch

Vectice's answer is a platform it calls, with a straight face, the first "Regulatory MLOps" platform. MLOps - machine-learning operations - is the tooling that helps teams ship models. Most of it is about speed. Vectice's version is about proof. You add a lightweight library to your workflow, call something like autolog(), and as you work it quietly captures the lineage: the datasets, the code versions, the metrics, the validation runs. Then it assembles that raw material into the actual documents auditors want - model development documents, model validation documents - mapped to whichever framework applies: SR 11-7, the EU AI Act, ISO/IEC 42001, the NIST AI Risk Management Framework.

The elegant part is the timing. Vectice captures the evidence while the work happens, not in a panicked reconstruction afterward. That is a genuinely different philosophy from the copy-paste status quo, and it is the difference the company sells hardest: govern continuously, not retroactively.

Why now

Vectice was founded in 2020, which in hindsight looks like good timing and at the time looked like a bet. The EU AI Act - the world's first comprehensive AI law, which puts documentation and traceability obligations on "high-risk" AI systems - has since moved from draft to law. Vectice did not need the regulation to justify the product; model risk management already required all this for financial models. But regulation has a way of turning a nice-to-have into a purchase order, and a market that was always latent became visible.

The founders had the scar tissue for it. CEO Cyril Brignone previously ran Arrayent, an IoT company, and co-founded the startup-media outlet Vator. CTO Gregory Haardt spent two decades in enterprise software, including a stint as CTO of Lattice Engines - a predictive-analytics firm acquired by Dun & Bradstreet - plus time at Salesforce. Neither is a first-timer, and it shows in the choice of problem. First-time founders chase the exciting demo. People who have sold to banks chase the workflow that never gets funded and always gets needed.

The unglamorous moat

What makes Vectice interesting is precisely how boring its problem sounds. "Automated documentation for model governance" does not trend. But boredom is a feature here. The tasks nobody wants to do are the ones with the least competition and the most durable demand, and compliance paperwork is about as durable as demand gets - it is legally mandated and it recurs forever. A model isn't documented once; it's re-documented every time it changes, every audit cycle, every new regulation. Vectice is selling relief from a chore that regenerates.

This is only the first step in Vectice's mission to revolutionize the way data science teams work and collaborate within their organization.

— Cyril Brignone, Co-Founder & CEO, on the Series A
01

By the numbers

$15.6MTotal funding raised
2020Year founded
~12Employees
1Line of code to autolog

Revenue is estimated in the low single-digit millions and the team is small - this is an early-stage company selling into very large, very cautious buyers. In enterprise compliance software, that is a normal shape, not a warning sign.

02

How it works

Four steps, and the point of the design is that a data scientist only really touches the first one. The rest is meant to be invisible.

STEP 01

Autolog

Add one line to your Python, Databricks, or Snowflake workflow. It logs assets as you build.

STEP 02

Capture lineage

Datasets, code versions, metrics, and validation runs are recorded automatically with full history.

STEP 03

Assemble docs

Model development & validation documents are generated from templates mapped to the right regulation.

STEP 04

Review & ship

Teams collaboratively review, then push audit-ready evidence into ServiceNow or watsonx.governance.

What Vectice says it changes

Company-reported figures — treat as vendor claims, not audited results
Documentation time reducedup to 90%
Compliance review cycles cut (watsonx)~50%
Time-to-production accelerated~25%
ROI realization window~45 days
03

What you can do with it

For the people who build models

  • Stop reconstructing what you did - it's logged as you go
  • Auto-generate model cards and reports from one library call
  • Keep a versioned map of data, code, and model lineage
  • Collaborate on reviews without leaving your tools
  • Works with Python, R, Databricks, Snowflake

For the people who answer to auditors

  • Documentation mapped to SR 11-7, EU AI Act, ISO 42001, NIST AI
  • Full audit trail and traceability for every model
  • Run policy reviews and log findings for remediation
  • Push evidence straight into ServiceNow risk records
  • A single source of truth across the whole model estate
04

The founders

Co-Founder & CEO

Cyril Brignone

Former CEO of IoT company Arrayent and co-founder of startup-media outlet Vator. Brought a go-to-market instinct for regulated enterprise buyers and the conviction that governance was a workflow problem, not a dashboard problem.

Co-Founder & CTO

Gregory Haardt

Twenty-plus years in enterprise software. Former CTO of Lattice Engines (acquired by Dun & Bradstreet) and a Salesforce alum. Holds an MS in applied mathematics and computer science from ENSEEIHT in France.

05

Where it sits in the stack

Vectice's 2025 moves tell you how it wants to be understood: not as a destination app you log into, but as plumbing that runs underneath the systems enterprises already use. In September 2025 it announced an integration with ServiceNow, letting it drop model development and validation documents directly into ServiceNow's Integrated Risk Management, with a bi-directional loop that pulls controls in and pushes findings back out.

It also built an integration with IBM watsonx.governance, auto-uploading its reports and syncing model metadata in real time - a combination the company says trimmed compliance review cycles by roughly half at adopting firms.

On the credibility side, Vectice is SOC 2 compliant, a member of the U.S. NIST AI consortium, and part of the AI Verify Foundation, the open-source community for responsible-AI testing. For a buyer whose entire job is managing risk, those memberships are less trophies than table stakes.

The competitive field is crowded but split. Experiment-tracking tools - Weights & Biases, Comet, Neptune, the open-source MLflow - overlap on capturing model runs. Governance specialists like Credo AI and Fiddler, and the incumbent IBM watsonx.governance, overlap on oversight. Vectice's wager is that the specific, unfashionable intersection - regulatory documentation for model risk - is a big enough place to stand.

06

Timeline

2020

Founded in San Francisco

Cyril Brignone and Gregory Haardt start Vectice to automate documentation and governance for data science teams.

2021

Seed funding, first platform

Early capital funds a platform for capturing and documenting data science assets.

Jan 2022

$12.6M Series A

Sorenson Ventures and Crosslink Capital co-lead, bringing total funding to $15.6M.

2023

Auto-documentation for the EU AI Act

Vectice launches an auto-documentation solution for ML projects aligned to emerging EU AI Act requirements.

2025

ServiceNow & IBM watsonx integrations

Deeper push into financial services with integrations into ServiceNow IRM and watsonx.governance.

07

Frequently asked

What does Vectice do?
Vectice is a Regulatory MLOps platform that automatically documents AI/ML models - capturing lineage, versioning, and validation evidence - and generates audit-ready reports mapped to regulations such as SR 11-7 and the EU AI Act.
Who founded Vectice and when?
Vectice was founded around 2020 by Cyril Brignone (CEO) and Gregory Haardt (CTO), both experienced enterprise-software leaders.
How much funding has Vectice raised?
$15.6M in total, including a $12.6M Series A announced in January 2022, co-led by Sorenson Ventures and Crosslink Capital.
Who uses Vectice?
Data science, model risk, and governance teams in regulated industries - especially banks and insurers that must document and validate AI models for compliance.
How is it different from Weights & Biases or MLflow?
Those tools focus on experiment tracking and model development. Vectice focuses on regulatory documentation, governance workflows, and audit readiness for model risk management.