Summary in 5 points
- Harvey AI is a vertical legal-AI SaaS on OpenAI + Azure; Amaii is a horizontal private-AI platform running inside the customer's own infrastructure in the EU.
- Amaii is model-agnostic (Llama, Mistral, Mixtral, EU-commercial); Harvey is built around OpenAI's GPT-4-class models.
- Harvey ships pre-built legal workflows (drafting, redlining, research); Amaii ships a configurable private company brain usable across legal, finance, healthcare, consultancy and public sector.
- Amaii provides full prompt/answer-level audit logging and document-level RBAC inside the customer environment; Harvey offers SSO, SOC 2 Type 2, ISO 27001 and contractual GDPR alignment on Azure.
- Harvey is priced per lawyer through annual enterprise contracts (figures not public); Amaii uses transparent licence-per-group plus infrastructure pricing — independent of lawyer headcount.
Comparison table
| Criterion | Amaii | Harvey AI |
|---|---|---|
| Scope | Horizontal — legal, finance, healthcare, consultancy, public sector | Vertical — legal & professional services |
| Data processing location | Customer infrastructure (on-premise or private cloud in EU) | Microsoft Azure (EU residency available; OpenAI processor chain) |
| Processing party | No third parties — all inside customer environment | Harvey + OpenAI + Microsoft Azure |
| Choice of language model | Open-source (Llama, Mistral, Mixtral) or EU-commercial | Fixed — OpenAI GPT-4-class and partner models |
| Pre-built legal workflows | Configurable; not pre-shipped | Yes — drafting, redlining, research, due diligence |
| GDPR-compliant | Yes, by design (no external processor) | Yes, contractually — DPA, EU residency on Azure |
| NEN 7510 / DORA / BIO ready | Yes | Not the target market |
| Per-action audit logging | Full — user, document, prompt, answer (inside customer env) | Audit logging via Harvey platform |
| Document-level RBAC | Yes, native, independent of source-system permissions | Workspace + matter-level controls |
| Vendor lock-in | None — model and infrastructure replaceable | Tied to Harvey + OpenAI + Azure stack |
| Pricing transparency | Transparent — per user group + infra | Not publicly disclosed; per-lawyer annual contract |
| Time-to-PoC | 4–6 weeks | Weeks to months (sales + procurement cycle) |
How Harvey AI processes your data
Harvey AI is a SaaS layer built on top of OpenAI's models, hosted on Microsoft Azure. Harvey publishes information about its security posture (SOC 2 Type 2, ISO 27001, GDPR DPA, EU residency on Azure) and contractually excludes training on customer data. Documents uploaded into Harvey are processed by OpenAI-class models running on Azure infrastructure inside the EU Data Boundary or other selected regions.
For many international law firms that posture is sufficient. For firms working under strict client-privilege requirements, conflict-of-interest controls and national-bar rules around third-party processing, the dependency on OpenAI as a generative processor — and on Microsoft Azure as critical ICT provider — can be the limiting factor in a DPIA.
How Amaii processes your data
Amaii installs a private AI stack inside your own environment: a Retrieval-Augmented Generation (RAG) pipeline, a vector database and an LLM of your choice. Every step — retrieval, embedding, generation — stays inside the customer infrastructure. There is no external API call to OpenAI, Harvey, Microsoft, Google or any other third-party processor.
Documents can be synchronised from a document-management system (iManage, NetDocuments, SharePoint), Confluence, file shares or internal APIs — but once they enter Amaii they are processed by a locally running language model. That makes Amaii suitable for firms where every external data flow must first be approved by the General Counsel, MLRO or DPO.
Scope: legal-only versus firm-wide
Harvey AI is purpose-built for legal and professional-services workflows: contract drafting, redlining, legal research, due diligence and matter summaries. The product, the prompts and the integrations are designed around how lawyers work. For pure legal-team productivity that is a real advantage.
Amaii is horizontal: it is a configurable private company brain used by law firms, but also by hospitals (NEN 7510), banks (DORA), accountants, consultants, R&D teams and government bodies. Many firms run both — Harvey inside the lawyer team for drafting, and Amaii as the firm-wide knowledge layer covering finance/back-office, HR, business intelligence and the parts of legal work where data must not leave the firm.
Compliance: GDPR, client privilege, EU AI Act
Harvey AI complies with GDPR through its DPA, EU residency on Azure and certifications. For client privilege the question is not only legal compliance but also professional-conduct rules: how acceptable is it to send privileged information to a SaaS layer that in turn relies on OpenAI and Microsoft? Some bar associations are increasingly explicit about this; Amaii sidesteps the question by keeping all processing in-house.
- GDPR: with Amaii no external processor; with Harvey there is a chain (Harvey + OpenAI + Microsoft).
- Client privilege: Amaii supports fully isolated processing of privileged case files inside the firm.
- EU AI Act: both support transparency requirements; Amaii additionally gives full control over model choice and logging.
- Conflict-of-interest: Amaii's document-level RBAC lets the firm enforce information barriers independently of DMS permissions.
- Data sovereignty: Amaii operates within EU jurisdiction without a US processor in the chain.
Pricing and transparency
Harvey AI is sold through annual enterprise contracts priced per lawyer. The vendor does not publish list pricing; industry reports cite figures from several hundred to over a thousand euros per lawyer per year, with material differences depending on firm size and contract term. That model scales linearly with headcount.
Amaii uses a licence model per user group plus infrastructure cost (own GPUs or EU cloud). Pricing is published in tiers and is independent of lawyer headcount, which makes total cost of ownership easier to forecast and avoids per-seat lock-in as the firm grows.
Governance, audit logging and RBAC
Both platforms offer audit logging and SSO. Amaii additionally logs every interaction at the level of user, document, prompt and answer — entirely inside the customer environment, so there is no dependency on what the SaaS vendor chooses to expose. For firms under bar-association supervision or operating with strict conflict-of-interest controls, that completeness is often a hard requirement.
Amaii's role-based access control (RBAC) works at document level and is independent of DMS permissions. That allows the firm to enforce information barriers (Chinese walls) at the AI layer even when the underlying DMS permissions are broader.
When do you choose Harvey AI?
Harvey AI is the right choice when your firm wants out-of-the-box legal workflows (drafting, redlining, research), most use cases are inside the lawyer team rather than firm-wide, and your compliance posture allows OpenAI + Microsoft Azure as part of the processing chain. For large international law firms with the budget for per-lawyer SaaS, Harvey is often the fastest route to specialised legal AI productivity.
When do you choose Amaii?
Amaii is the right choice if at least one of these applies: your firm or organisation needs AI beyond the legal team; you require all processing inside your own EU-based infrastructure; you want independent choice of language model; you want transparent pricing not tied to lawyer headcount; or you operate under NEN 7510, DORA or BIO in addition to legal-sector rules.
- Law firms that want firm-wide AI (legal + finance/back-office + HR + operations).
- Notaries and corporate-legal teams with strict client-privilege rules.
- Mid-market and EU-headquartered firms that prefer EU-only processing.
- Multidisciplinary professional-services organisations (legal + accounting + tax).
- Public-sector legal departments under BIO and data-sovereignty rules.
- Any organisation that wants transparent pricing independent of lawyer headcount.
Common misconceptions
"Harvey is built for legal, so it is automatically the safest legal AI." Vertical focus is a product-design choice, not a data-processing guarantee. Harvey still routes data through OpenAI on Azure. For client-privilege risk, what matters is who processes the data and under which jurisdiction — not whether the UI is legal-themed.
"A private LLM cannot match a specialised legal vendor." Amaii does not pretend to ship Harvey-style drafting templates out of the box. Many firms therefore use both: Harvey for templated legal output, Amaii for firm-wide private knowledge work. The two are complementary, not mutually exclusive.
"Per-lawyer pricing is fair because every lawyer uses it." Per-seat pricing is predictable but scales linearly with the firm and ties the cost base to the SaaS vendor's roadmap. A licence-per-group model decouples AI cost from headcount.
Bronnen en achtergrondinformatie
- EU AI Act (Regulation 2024/1689) - EUR-Lex
- GDPR (Regulation 2016/679) - EUR-Lex
- DORA — Digital Operational Resilience Act - EUR-Lex
- Harvey AI — Security & Trust - Harvey
- Microsoft Azure EU Data Boundary - Microsoft Learn
