Solve Intelligence

What it is

Solve Intelligence is a browser-based AI platform for patent work: drafting applications, prosecuting them through office actions, and — since December 2025 — charting claims for litigation and licensing. Sanj Ahilan, Angus Parsonson, and Chris Parsonson founded it in London in June 2023 and took it through Y Combinator’s S23 batch. The editor is the product. It behaves like a shared document with a chat panel beside it, and the patent-specific machinery — figures, chemical structures, biological sequences, antecedent-basis review — lives on the same page instead of in an attachment workflow.

It is the mirror image of Patlytics. Patlytics started at the litigation end, where the deliverable is a claim chart, and worked backwards into drafting. Solve started with the attorney writing claims and worked forwards into charts. Both now describe themselves as full-lifecycle. Both are still weighted where they started, and that weighting is the actual buying decision.

Funding runs $12M in April 2025 with Microsoft’s M12 and Thomson Reuters Ventures investing, then a $40M Series B on 9 December 2025 co-led by Visionaries Club and 20VC, with Thomson Reuters and Y Combinator increasing their stakes and Operator Collective joining — $55M total. On 31 March 2026 Solve acquired Palito.ai, a Munich patent-litigation AI startup, and opened a Munich office with part of the Palito team. That deal added German, EPO, and UPC validity workflows, obviousness charting under 35 U.S.C. 103 for US litigation, and sentence-level citations.

Why it shows up in Legal Ops stacks

• Drafting is where the volume actually is. The vendor counts 433K+ patent applications drafted on the platform against 1,100+ claim charts created. Both are vendor figures, and the ratio between them is the honest read on the product: drafting is mature and load-bearing, Charts is nine months old and still proving itself. If your bottleneck is applications per attorney per quarter, this is the half of the market Solve owns.
• Review catches the defects that cost office actions. The Review pass flags inconsistent terminology and antecedent-basis problems inline with one-click corrections. Those are the errors a junior associate burns an afternoon on and an examiner finds anyway.
• Regional processing is an answer a European IP department can use. SOC 2 Type II and ISO 27001 certified, GDPR and CCPA, AES-256 at rest and TLS 1.3 in transit on AWS, US or EU processing selectable per customer, zero-data-retention agreements with the third-party model providers, and no training on customer data. Combined with the Munich office and UPC workflows, Solve is now the entrant that survives an EU procurement review rather than dying in it.
• Scale signals. 700+ IP teams across six continents, more than 60% of them US-based, with Intel, DLA Piper, Finnegan, Perkins Coie, Siemens, and Amgen named as customers. Vendor numbers, unaudited.

Pricing reality

There is no published price, no tier list, and no free trial. solveintelligence.com/pricing does not exist; every deal goes through /book-demo.

The only per-seat figure in circulation is roughly $775 per user per month, about $9,300 a year — and it is an inference, not a quote. A third-party analysis backed it out of a NAPP member benefit offering 15% off an annual subscription with the discount valued near $1,400. One data point, reverse-engineered. Treat it as an estimate and do not budget on it.

The segment anchors are firmer. Spellbook’s 2026 legal-AI pricing survey puts firm-side enterprise tools at about $1,000–2,000+ per user per month at the Harvey end, $300–800 at Legora, and $225–650 for research AI like CoCounsel and Lexis+ Protégé. The inferred Solve number lands between the research tier and the Harvey tier, which is where a specialist platform sold to patent attorneys belongs.

Drafting volume moves the quote as much as seat count does. Four attorneys filing 15 applications a quarter each is a different contract from four attorneys filing three. Put four things in the order form: the applications-per-month entitlement, the overage rate above it, whether Charts is bundled or a separate SKU — it launched as its own product — and whether EU regional processing carries a premium.

Best for

In-house IP counsel and law-firm patent prosecutors whose workload is dominated by filing volume: teams drafting and prosecuting dozens of applications a quarter across mechanical, electrical, life-science, or software art units, where one platform across the lifecycle beats a point tool — and especially where EPO and UPC matters are a real share of the docket.

Not for you if your attorneys will not leave Microsoft Word. Solve is a browser editor with no Word add-in, and that is a workflow fight you lose. Not for you if charting volume dominates the practice: Charts is the newest surface with the thinnest record, and Patlytics has been building that deliverable since 2024. Not for you if you need docketing, annuities, or IP spend management — Anaqua stays on the invoice. And not for you if you want to drive patent work from Claude Code or another MCP client: Solve publishes no MCP server and no public API, so the work happens in the browser or not at all.

Alternatives — and when to pick them instead

• Patlytics — the closest full-lifecycle rival, $40M Series B in April 2026 led by SignalFire for roughly $65M total, plus a read-only MCP server. Pick it when evidence-of-use charting is the recurring cost center, or when you want prior-art lookup inside Claude Code.
• DeepIP — the fastest-growing entrant on the prosecution side: $15M Series A in March 2025, $25M more in March 2026, and 400+ law firms and corporate IP teams. It drafts inside Word and connects to IPMS and docketing tools. Pick it when the Word requirement is non-negotiable or docketing integration is a hard dependency.
• PatSnap, Clarivate Derwent, and Questel — the incumbent patent data and analytics layer, and the two or three names already on your shortlist by market share. Pick one when the job is landscape analysis, search coverage, and a defensible data license rather than drafting work product.
• Harvey or Legora — generalist legal AI the firm probably already licenses. Pick it when patents are one slice of a mixed docket and consolidating vendors beats best fit on any single workflow.

Watch-outs

• Charts is the newest thing on the page and the number behind it is small. 1,100+ charts against 433K+ applications describes a product still finding its footing, and the Palito capabilities landed five months ago. Guard: buy Solve on the drafting and prosecution case. Run Charts as a scoped POC against a matter you have already charted by hand, diff it limitation by limitation, and only then let it become a contractual entitlement.
• ISO 42001 is stated two ways on the vendor’s own site. The homepage lists it with the certifications; the security page says it is in progress. Guard: ask for the certificate and the auditor name in writing before your AI-governance questionnaire cites it. SOC 2 Type II and ISO 27001 are the two you can rely on today.
• Output quality tracks prompt quality, and it tracks unevenly by domain. Independent review reports that mechanical, medical-device, and chemistry-heavy applications need more prompting than software and electrical work. Guard: run the POC in your weakest art unit rather than your strongest, and write a house template library during the trial. Solve’s reusable templates exist for this, but somebody on your side has to author them.
• Browser-only, with no MCP server and no public API. Nothing about the platform can be automated from outside it, and drafts leave as documents rather than as data. Guard: get export format and bulk-export rights into the contract, and keep the IPMS as system of record so a vendor change never strands the portfolio.
• Nearly every proof point above is vendor-supplied. 700+ teams, the 60–80% drafting time savings, the 265% growth in weekly actions per user — all Solve’s own numbers, none independently audited. Guard: measure attorney-hours per application across ten filings before the trial and ten during it. That is the only figure that survives a renewal conversation.