Hebbia
Enterprise AI platform built for institutional finance. Its Max agent and Matrix grid reason across thousands of documents and market data feeds, then deliver memos, models, decks and dashboards with every answer traced back to its source sentence.
What is Hebbia?
Hebbia is an enterprise AI platform built by Hebbia, Inc. of New York for the way institutional finance actually works. The site sums up the ambition in two words, Institutional Intelligence, and the product is aimed squarely at investors, bankers, lawyers, consultants and Fortune 500 finance teams rather than at a general audience.
Two products sit at its core. Max behaves like an always-on analyst: it researches a topic, spreads comps, recaps an earnings call, drafts a presentation or builds an Excel model, and it can be reached by email as well as in the application. Matrix takes a different shape, a grid where rows are documents or companies and columns are the questions you want answered across all of them, so that hundreds of filings, transcripts or credit agreements can be interrogated in one pass. Around them, Projects gives people and agents a shared workspace with private history and published outputs, while Skills & Agents turns a firm's own procedures into reusable routines, from buyer diligence questionnaires to ESG disclosure benchmarking and sector M&A tracking.
The platform's value depends on what it can reach. Hebbia connects private document stores such as SharePoint, OneDrive, Box, Dropbox, Egnyte, AWS and IntraLinks, public filings from the SEC, Europe, UK Companies House, Australia and New Zealand, professional data feeds including FactSet, S&P Capital IQ, PitchBook, Preqin, Fitch, ICE Market Data and Daloopa, expert networks like Third Bridge and Guidepoint, and internal systems such as Salesforce, DealCloud, Snowflake and Databricks.
The recurring promise is verifiability: the Matrix page asks users to check the work as it goes, tracing every answer back to the exact sentence. Hebbia reports customers holding around $30 trillion in assets under management, roughly 200,000 prompts a day, 1.5 billion pages processed and more than a thousand production use cases. Certifications listed include ISO/IEC 42001:2023 and SOC 2 Type II, with AES-256 encryption at rest and TLS 1.2+ in transit, and an explicit commitment not to train on user data. Access is granted only after a demo.
What it does
- Reason across thousands of documents at once, beyond any single context window
- Trace every answer back to the exact sentence it came from
- Produce client-ready decks, memos, emails and dashboards from the analysis
- Build financial models with a citation inside every cell, in the firm's own format
- Screen a data room, a target list or a covenant library against defined criteria
- Codify a firm process into an agent that then runs continuously on its own
- Answer, draft or brief straight from the inbox, without opening the application
When to use Hebbia / When not to
A quick filter to help you decide if Hebbia is the right fit.
When to use Hebbia
- Investment bankers running sell-side coverage, precedent transactions and pitch material
- Private equity, private credit and hedge fund investment teams screening VDRs, CIMs and covenants
- Corporate and restructuring lawyers reviewing contracts, debt documents and diligence files
- Management and strategy consultants assembling client-ready analysis from deal rooms
- In-house corporate development and finance teams building target lists and board material
When not to use Hebbia
- Individuals and small teams: there is no free plan, no trial and no self-service sign-up
- Anyone needing a published price before talking to sales, since the pricing page shows none
- Organisations wanting shared or service accounts, which the Fair Use Policy forbids outright
- Mobile-first or offline users, as the product runs only in the browser with no app or extension
- Teams without subscriptions to the professional data feeds the platform is designed to connect
How to use Hebbia
A typical end-to-end flow, from setup to results.
- Request a demo from the site, enabling the targeted advertising cookies the form needs
- Meet one of Hebbia's AI strategists, who maps the platform onto your firm's workflows
- Sign an Order Form under the Main Services Agreement, with named Core User Licences
- Log in to the platform at search.hebbia.ai
- Connect your document stores, public filings, data providers, CRM and data warehouse
- Open a Project, upload the relevant files and invite the deal team into the shared space
- Ask Max in plain language, for example to spread comps or flag guidance changes in a call
- Build a Matrix with one row per document or company and one column per question
- Check each answer by following the citation back to the source sentence
- Turn the validated process into a Skill or agent, then publish the deck, memo or model
Pros & Cons
Pros
- Every answer traces back to the exact source sentence, which makes review possible
- Handles document volumes far beyond what a single model context window can absorb
- Outputs are finished deliverables, not chat: memos, decks, models and dashboards
- Unusually deep integration with professional financial data feeds and internal systems
- Serious compliance posture: ISO/IEC 42001:2023, SOC 2 Type II, published DPA and sub-processor list
- Named EU and UK Article 27 representatives, plus a designated Data Protection Officer
- Contractual 99.9% monthly uptime SLA with service credits, and weekday phone and email support
Cons
- No public pricing at all: the pricing page shows a demo button and nothing else
- No free plan and no free trial are mentioned anywhere on the site
- Access requires a demo and a negotiated contract, so evaluation is slow
- No public API documentation, although the Fair Use Policy governs API and programmatic access
- Browser only: no mobile app, desktop app or browser extension was found
- Named-seat licensing with usage audits, throttling and possible suspension for shared accounts
- No interface language other than English is documented
Pricing & Plans
Hebbia publishes no prices. The pricing page carries no tiers, no amounts and no free plan or free trial, only an invitation to book a demo. Commercial terms are set individually in an Order Form attached to the Main Services Agreement, and fees are payable in US dollars unless the Order Form states otherwise. The Fair Use Policy describes the underlying unit as a Core User Licence assigned to a single named individual, and mentions the possibility of moving to a higher-tier plan when usage exceeds what was agreed, without naming or pricing it.
- a named seat assigned to one individual
- priced privately in an Order Form
- referenced by the Fair Use Policy for excessive usage
- neither named nor priced
- negotiated for unusually high or specialised use
- on request
Data, GDPR & hosting
A consolidated view of how Hebbia handles your data.
GDPR overview
GDPR compliance is claimed on the homepage and security page, and backed by concrete mechanisms. A Data Protection Officer is appointed and reachable at privacy@hebbia.ai. An Article 27 representative is named for the EU (Osano International Compliance Services Limited, Dublin) and a separate one for the UK (Osano UK Compliance LTD, Belfast). Transfers rely on the European Commission's 2021 standard contractual clauses and, for the UK, the International Data Transfer Addendum, since the United States holds no adequacy decision. A Data Processing Agreement was published on 27 April 2026, carrying the SCCs, a ten-day notice period on new sub-processors and a named sub-processor annex. Eight data subject rights are listed, with a public request form. Non-essential cookies rely on Article 6(1)(a) consent.
Who owns the data?
The Main Services Agreement is explicit: the customer owns all right, title and interest in its Customer Materials, while Hebbia owns the underlying Hebbia Technology. The customer grants Hebbia a limited, non-exclusive licence to use those materials to run and support the platform. Hebbia may use them to improve machine learning models reserved solely for that customer, but may never license those models or make them available to any third party. Usage data can be reused to improve the product only once aggregated or de-identified, and the customer cannot be named as a source without prior written approval. Suggestions sent to Hebbia, by contrast, become freely usable and sub-licensable by the vendor.
Reuse rights
Customers receive a limited, non-exclusive licence for their authorised users to access the platform for internal business purposes. Because the customer keeps ownership of its own materials, the outputs built from them can be reused internally without asking Hebbia for permission. The restrictions bite elsewhere: no reverse engineering, no derivative works, no reselling, renting, sub-licensing or time-sharing of the platform, no use for the benefit of a third party, and no removal of proprietary notices. The Fair Use Policy adds that reselling outputs requires Hebbia's prior written consent, and that high-volume automation or batch processing, whether through the interface or through APIs, falls outside fair use. Hebbia may audit usage, gives ten business days to remedy a breach, and can then require extra licences or suspend access. Beta features are supplied as-is with no liability attached.
Data retention & training
Hosting summary
Hebbia is headquartered in the United States and states that information it collects is processed there, noting that the United States has neither sought nor received an EU adequacy decision. Storage locations are restricted to the geographic regions set out in the Data Processing Agreement. Annex III names Amazon Web Services as the cloud provider, operating in the AWS US-East-1, US-West-1, EU-West-1 and EU-Central-1 regions; several other sub-processors are listed as US-based with EU regional processing available, and one as regional based on tenant. The DPA also allows processing in any country where Hebbia, its affiliates, partners or authorised sub-processors maintain facilities, subject to applicable data protection law. All processing takes place in physical data centres run by those sub-processors. Data is encrypted at rest with AES-256 through AWS transparent disk encryption, with keys managed by AWS. Transfers out of the EEA rely on the 2021 standard contractual clauses, and UK transfers on the International Data Transfer Addendum.
Things to keep in mind
Risks and trade-offs to weigh before adopting Hebbia.
- Client-ready output invites over-trust: IC memos, models and decks feed real investment decisions, and the platform itself tells users to check the work along the way
- A citation proves where a sentence came from, not that the reasoning built on it is sound
- Junior analysts who never read the primary documents may not build the judgement the job requires
- Highly sensitive material, from data rooms and CIMs to credit agreements and litigation files, is entrusted to a third party
- The contract lets Hebbia improve models from customer materials, keep aggregated usage data, use suggestions freely and put the client's name and logo in its marketing
- Processing defaults to the United States with no EU adequacy decision, behind a long chain of fourteen named sub-processors including several model providers
- Deep integrations, codified agents and named-seat licences make leaving expensive once the platform is embedded
Setup & Integrations
Technical difficulty
Getting in is a commercial exercise rather than a technical one: a demo, then an Order Form under the Main Services Agreement. Daily use is simple, since the platform runs in the browser. The real work is connecting sources, including SharePoint, Box, Dropbox, Egnyte, IntraLinks, Salesforce, DealCloud, Snowflake and Databricks, with the Salesforce and Outlook connectors routed through a third-party provider. Professional data feeds assume existing subscriptions. Hebbia fields AI strategists and forward-deployed engineers to help, and codifying firm processes into agents takes business time more than technical skill.
Deployment
Integrations
Behind Hebbia
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What does Hebbia actually do?
What is the difference between Max and Matrix?
How much does Hebbia cost?
Is there a free plan or a free trial?
Does Hebbia train its models on customer data?
Where is the data hosted?
Is there a DPA and a list of sub-processors?
Which sources can be connected?
Is there a mobile app?
What support and availability are committed?
Should you pick Hebbia?
Hebbia knows exactly who it is for, and makes no attempt to be for anyone else. It targets institutional finance, banks, funds, law firms, consultancies and large corporate finance teams, and everything about the product follows from that choice. The strongest argument in its favour is verifiability: in a domain where being slightly wrong costs as much as being completely wrong, a system that traces every answer back to the exact sentence it came from is more useful than one that merely sounds confident. The second is scale, since Matrix is designed for the hundreds or thousands of documents that overwhelm a single context window, and the third is the depth of its connections to professional data feeds and internal systems.
The compliance file is unusually complete for a company of this age: ISO/IEC 42001:2023 and SOC 2 Type II, a published Data Processing Agreement naming its sub-processors, appointed EU and UK representatives, and a contractual uptime commitment.
The drawbacks are equally clear. Nothing about pricing is public, there is no free plan, no trial and no self-service entry, and evaluation therefore begins with a sales conversation. Named-seat licensing is enforced by audit. The product runs only in the browser. And while the Fair Use Policy governs API and programmatic access, no public API documentation exists.
Founded in 2020, backed by a $130 million Series B led by Andreessen Horowitz and now hiring across two offices, Hebbia is a maturing company rather than an experiment. If your team already drowns in deal documents and pays for professional data it cannot fully exploit, it deserves a demo. If you are an individual, a small firm or anyone who needs to see a price first, look elsewhere.
- Choosing a selection results in a full page refresh.
- Opens in a new window.