Sandstone

What it is

Sandstone is an AI-native front door for in-house legal departments. Requests are captured where the business already asks — email, Slack, Microsoft Teams, Salesforce, Jira, procurement portals — and agents read the intent, pull the surrounding context, and route the work. The context is the point: an arriving request lands with the counterparty’s history, the prior decisions on that relationship, the linked documents, and the business signals attached, rather than as a ticket with a free-text field. Agents then finish the routine end of the queue — first-pass redlining, replying to comments, accepting and rejecting changes, drafting from the team’s own playbooks with multi-source citations — and hand the judgment calls to a lawyer. Founded by Nick Fleisher (previously head of McKinsey’s legal technology practice), Jarryd Strydom (McKinsey legal ops, formerly an in-house technology attorney), and Liam Germain, Sandstone came out of stealth in October 2025, raised a $10M seed led by Sequoia in January 2026, and a $30M Series A led by Lightspeed on 9 June 2026 — $40M total. Named production customers: Wayfair, Mercury, Grindr, ElevenLabs, MasterClass, Cox Media, and Hypertherm.

Why it shows up in Legal Ops stacks

• The data model is the relationship, not the document. Fleisher’s stated argument against the field is that CLMs are repositories that do not understand relationships and intake tools route tickets they do not comprehend. Sandstone’s repository holds companies, people, documents, and past decisions as one graph, and the vendor calls the category Legal Relationship Management rather than CLM. That is a real architectural difference from Ironclad, where the contract record is the object everything else hangs off.
• Time-to-live is the wedge. The vendor’s claim is that a team stands up intake, triage, first-pass redline, and drafting agents in under 10 minutes, against the multi-month implementations that legal software is known for. Lightspeed reports accounts signing up and going live the same day. If that survives your own evaluation, it is the strongest reason to look here before Checkbox, whose configuration canvas is more capable and takes weeks to shape.
• It reports on the queue, not just the contract. Aggregate cycle time, risk exposure, capacity, and deviation rate are first-class outputs, which is what a GC needs to argue for headcount. See legal intake and matter management for the model underneath.
• More than 50 integrations, and a document add-in. Google Docs, Microsoft Word, Google Drive, OneDrive, SharePoint, Dropbox, Salesforce, HubSpot, Zendesk, Workday, Coupa, Zip, Notion, Jira, Asana, ServiceNow, Outlook, and Ironclad are named on the vendor’s own list, with a Google Workspace Marketplace add-on for drafting inside the document.

Pricing reality

Sandstone publishes nothing. There is no pricing page — sandstone.com/pricing returns a 404 — and every route ends at a booked demo or team@sandstone.com. No usable Sandstone-specific contract values are public as of 8 September 2026, so the only honest anchor is the segment: off-the-shelf legal AI platforms sold to in-house departments land roughly $20,000 to $200,000 per year, with enterprise seats reported at $280 to $550 per user per year and premium research tools like Harvey starting near $500 per seat per month. Sandstone sells to mid-market and enterprise legal departments and is quoted by legal ops, not self-served by an individual lawyer, so plan the business case at the low-to-middle end of that band for a 5-to-15-lawyer department and treat anything above it as a negotiation, not a list price.

Best for

A GC or legal ops manager at a 500-to-5,000-person company where legal is the bottleneck on commercial velocity and the queue arrives in five channels at once. The specific trigger: sales is waiting on NDA and MSA redlines, procurement pastes vendor paper into Slack, nobody can say how many requests came in last quarter or from which business unit, and the team has neither the headcount nor the appetite for a nine-month CLM rollout. Sandstone is the pick when you want the front door working this month and the reporting to justify the next hire.

Versus the alternatives

Ironclad is the incumbent to beat when the executed contract is the object of record and obligation tracking is the actual purchase — intake is a stage of the CLM there, and Sandstone integrates with it rather than replacing it. Onit wins when outside-counsel spend and matter management drive the budget. Checkbox is the right call when compliance, HR, finance, and procurement want the same configured front door, because one engine across four departments beats a legal-only tool on total cost. Streamline AI is the closest purpose-built rival on intake and workflow and the safer pick if you need a longer production track record than 11 months. LawVu is the choice when you want matters, contracts, spend, and knowledge in one legal workspace and are willing to trade agent depth for coverage. Wordsmith is the fastest-growing entrant in the segment — $114M raised and more than 500 companies — and beats Sandstone on programmability, since its MCP server makes legal review callable from another team’s agent. Harvey and Legora are a different purchase: research and drafting depth for lawyers, not a queue for the business. If none fit, run legal intake as a service-desk project in the Jira or ServiceNow instance you already pay for and add a triage rota — that fixes measurement for two weeks of admin, and it is the correct answer if you cannot name your top three request types today.

Watch-outs

• The traction number is real growth on a base near zero. Revenue up more than 40x in 90 days is a vendor-published figure from a company that was in stealth 11 months ago, and no independent audit exists. Guard: cap the first term at 12 months, refuse multi-year prepay, and get a contractual data-export right for the relationship graph before signing — the graph is the switching cost.
• No published price means no published floor. The quote for a 6-lawyer team is not the quote after sales ops, procurement, and two more entities are on the platform, and expansion is the vendor’s growth motion. Guard: ask for year-one and year-three pricing in the same document, priced at double the seats and triple the request volume.
• No public API reference and no MCP server. Sandstone ships more than 50 prebuilt integrations, but there is no developer documentation site and nothing MCP-callable, so it cannot currently be a step inside agents your own team builds. Guard: if that matters, request the API reference and confirm webhook coverage for request-created and status-changed events during evaluation rather than assuming it exists.
• It is not the CLM of record, and the naming invites confusion. The vendor markets a “CLM+” repository, but the obligation calendar, clause library governance, and executed-contract archive that a real CLM owns are not the product’s centre of gravity. Guard: write down which system holds the executed contract and the renewal dates before go-live; if the answer is Sandstone, get a written scope of what it does on obligation expiry.
• Strong security posture, undisclosed residency. SOC 2 Type 2, AES-256 at rest, TLS 1.3 in transit, SAML SSO, RBAC, audit logs, annual third-party penetration tests, zero-data-retention agreements with model providers, and an explicit no-training-on-customer-data commitment are all published. Hosting region, data residency options, named model providers, and the subprocessor list are not. Guard: for EU entities, get residency and the subprocessor list written into the DPA before privileged material passes through the front door — non-negotiable under GDPR.