What this replaces.
AI patent tools help you search faster. We tell you what will happen when you file.
Two kinds of patent intelligence.
Most AI patent platforms are workflow tools. They take a task you already do, searching for prior art, drafting a claim chart, classifying a portfolio, and make it faster. You upload a patent. You click a button. You get a document back. The intelligence stays at the surface: text similarity, keyword matching, document scoring.
That is useful. It is not what we do.
We built a connected intelligence layer. Every patent is wired to every case it appeared in, every judge who ruled on it, every attorney who litigated it, every firm that represented the parties, every corporate entity in the ownership chain. That connectivity is the difference between “here are similar patents” and “here is what will happen.”
A search tool gives you 50 cases about SEPs.
This platform tells you: if you enforce patent X against company Y in court Z, the most likely path is a Markman hearing at month 8, settlement negotiation at month 14, with 63% probability of settlement between $2M and $8M, but only if you file the MSJ early, because 82% of high-value settlements in this judge's history include an early MSJ.
That is the difference.
What each approach delivers.
Upload a patent. Get a list of potentially similar products ranked by text similarity. Review the list yourself.
Ranked enforcement targets with settlement probability, litigation history, counsel track record, and financial exposure, scored across 200M+ connected entities. Not "this product might infringe." Instead: "this defendant settles 73% of the time, has been sued 12 times before, and their preferred counsel loses in this court."
How the graph connects entities→AI-assisted mapping of claim elements to evidence. Paste evidence in, get a formatted chart out.
Element-by-element mapping with source code analysis, technical documentation, and prior art cross-referenced against the full patent family graph. Evidence is connected to the entities and cases it has appeared in before, so you know whether it has survived challenge.
How we build claim charts→Keyword and semantic search across patent databases. You read the results and assess relevance.
Graph-connected prior art. Not just text similarity but structural relationships: same inventors, same assignees, same classification codes, same citation chains. PTAB survival rates for patents with this profile. Which IPR petitioners succeed against patents like yours, and which arguments they use.
Prior art research service→Scoring algorithms rank your assets high to low. You decide what the scores mean. Auto-classify into a taxonomy. Filter and sort.
Seven independent valuation methods applied to your portfolio, cross-referenced against 8M+ litigation outcomes, judge ruling patterns, and jurisdiction-specific success rates. Not "this patent scores 82." Instead: "this patent is worth $4.2M–$7.8M against these three defendants, in this court, with this judge, given comparable verdicts."
Questions you can ask about your portfolio→Not offered. The tool shows you data. Strategy is your job.
Jurisdiction-sequenced enforcement approaches ranked by predicted outcome. If you file in Mannheim first, here is month 14. If you settle with defendant one before filing against defendant two, here is the leverage shift. Each scenario backed by a frozen seed you can reproduce.
See this in Nokia v OPPO→Not offered. Some platforms show "likelihood scores" without explaining what they mean or how they were derived.
Settlement, trial, or dismissal, with confidence ranges that narrow as the case progresses. Duration windows. Damages ranges. Based on the actual behaviour of the specific judge, defendant, and jurisdiction. Not a score. A probability distribution derived from 8M+ historical cases where the outcome is already known.
Four verified predictions→Not offered. You hire an expert at £50K–£100K per engagement.
Seven independent valuation methods that cross-check each other. Comparable licence analysis. Georgia-Pacific factor scoring. Reasonable royalty computation. Lost profits modelling. Each method produces a range. Where three or more methods converge, that is your defensible figure.
FRAND rate simulation: Ericsson v D-Link→Alerts when a new patent publishes or a competitor files. Static notifications.
Continuous tracking of litigation, verdicts, settlements, and regulatory actions across all tracked targets. Predictions tracked against reality in real time. When a prediction was wrong, the platform knows it and recalibrates.
Services overview→Four engines, one stack.
A fully connected map of the legal system. Every patent, case, judge, attorney, firm, party, and outcome, wired into a single structure. Updated daily from 30+ public sources.
A proprietary foundation model trained on the shape of the legal system rather than the words inside it. Given a patent and a defendant, it finds the cases that look most like yours and tells you how those cases ended. Anyone could rebuild the architecture. Nobody else can train it, because nobody else has the graph.
Technical expertise behind the model→Runs counterfactual rollouts of an entire enforcement campaign across multiple jurisdictions. If you file in Mannheim first, what does month 14 look like? If you settle with one defendant, what does that do to leverage against the next? It plays the campaign forward, thousands of times, before you commit a single filing fee.
Nokia v OPPO: 6/6 predictions correct→Every number in every report, reproducible by anyone, on any machine. A deterministic engine that takes a JSON seed and produces a byte-identical output. We ship the engine with every report. Opposing counsel can rerun it. There is no model temperature. There is no black box.
See the verified scorecards→The platform predicts. Then we check.
Every prediction the platform makes is logged before the outcome is known. When the outcome arrives, the platform compares what it said to what happened. These are published results.
The platform does not just analyse what has happened.
It predicts what will happen, and publishes those predictions before the outcome is known. Then it checks.
Intelligence products, not dashboards.
We do not sell software seats. We deliver finished intelligence.
Project
A target dossier. A portfolio valuation. A FRAND analysis. A campaign plan. Fixed scope, fixed fee, delivered in 1–4 weeks.
Service engagements→Retainer
Continuous access to the platform through our analysts. Monthly intelligence updates on tracked targets, portfolios, or markets.
Discuss a retainer→Joint Venture
For larger enforcement campaigns. We bring the platform, the analysis, and the strategic direction. Returns split against agreed milestones.
Talk to us about JV structures→Structure, not search.
Search tools give you a faster way to find documents.
This platform tells you who to enforce against, which strategy wins, and what will happen when you file.
See the difference on your portfolio.
Send us a patent or a target. We run the analysis, you decide.
Book a Call