COMPARISON · Drafting & review
Harvey vs Spellbook
Drafting and review tools compared on price path and training statements. This page places vendor-published facts side by side. It does not rank the products or state which is better.
| Fact | Harvey | Spellbook |
|---|---|---|
| Category | Drafting & review | Drafting & review |
| Vendor | Harvey | Spellbook |
| Status | Active | Active |
| Price path | Contact sales: No public price was located in the checked vendor source. | Contact sales: No public price was located in the checked vendor source. |
| Confidentiality statement | “We don’t use inputs, outputs, or uploaded documents to train underlying models.” | “Spellbook has negotiated agreements with both OpenAI and Anthropic for zero data retention (ZDR). This means customer data included in requests and responses with these LLMs is not persisted and exists only in memory in order to process a request.” |
| Training or data-use statement | No-training commitment: “Harvey contractually prohibits model providers from training on customer data and only uses your data for processing your requests, not for model improvement.” | No-training commitment: “Zero data retention means our LLM providers never learn from, train on, or store your data.” |
| Firm-size scope | 6-20 | Not stated |
| Named connections | Microsoft Word, Microsoft Outlook, Microsoft SharePoint | Microsoft Word, Google Docs |
| Data location | Choose where your data is stored - EU, UK, US, or Canada. | Not stated |
| Last checked | September 15, 2026 | September 15, 2026 |
Every value above is a dated observation from the vendor’s own published material. Open either tool page for the full evidence ledger and source links.
Editorial assessment, side by side
The rows below are the CounselAtlas editorial desk’s own reading, not vendor statements. They restate what each vendor published and add no claim of its own. Neither column is a rating or a recommendation.
| Assessment | Harvey | Spellbook |
|---|---|---|
| Reading the vendor's wording | The vendor says it encrypts data at rest and in transit, uses access controls, and by default never trains on customer data. It defines customer data as uploaded documents, and defines queries separately as customer content, which is a distinction worth reading closely. | The vendor says it has negotiated zero-data-retention agreements with OpenAI and Anthropic, so customer data in requests and responses is not persisted and exists only in memory to process a request. It also states that its providers never learn from, train on, or store customer data. |
| What adopting it costs | There is no public price, so the first cost is the sales and procurement process itself. Because the platform hosts in Microsoft Azure with EU, Switzerland, and Australia options, a firm with residency requirements should settle the region before signing. | It works inside Word and Google Docs, so there is little to migrate. The document stays where it already lives, which keeps the switching cost low and the lock-in correspondingly low. |
| Closest alternative | Spellbook — The Word-native alternative aimed at a much smaller buyer. The practical difference is whether you can buy the product without a procurement process. | Harvey — The enterprise alternative with a sales-led contract. This one can be adopted inside the tool you already write in. |