OLI-Intel — the Idea Landscape, automated.
A five-workflow pipeline that takes an inventor's submission, researches the patent landscape across millions of publicly available records, analyzes what it finds, and produces a publication-quality PDF — automatically.
The Problem: Inventors Fly Blind Right Before They Spend the Most Money.
An inventor with a new idea faces an expensive fork. To understand whether anyone has already filed something similar, they either pay a patent attorney or a search firm four figures for a prior-art search, or they guess. Most guess — and then spend thousands more on legal fees and consulting built on top of that guess.
OLI-Intel removes the guess. It gives inventors a clear picture of their Idea Landscape — what already exists in the public patent record around their concept — before they invest $1,000s in legal fees or consulting. Not a legal opinion. Not a patentability verdict. An honest, structured map of the publicly available landscape, so the next dollar they spend is an informed one.
The system takes a submitted idea and runs it through five chained workflows: intake and decomposition, multi-source research, USPTO retrieval across millions of patents, analysis and filtering, and report generation. The output is a tiered PDF report that walks the inventor through their entire Idea Landscape.
The pillar that makes it usable: a 100% confidential pipeline
An automated research product is only trustworthy if the inventor's idea is safe inside it. OLI-Intel's confidentiality isn't a promise bolted on — it's a structural consequence of how the pipeline works.
The raw idea never leaves in one piece. At intake, the idea is dissected into discrete search parameters — industry, CPC classification, general category terms, component concepts. Only these decomposed, non-reconstructable fragments are ever sent to third-party services (USPTO, Google). No external party receives an assembled, reconstructable idea. Reasoning that requires the whole idea happens on in-house LLMs that never transmit it outward.
That's the safeguard that makes an automated Idea Landscape report trustworthy: the inventor's concept is decomposed before any external call, and no third party can reassemble it from what it receives.
Five Workflows. One Submission. One Tracking Key End to End.
Each stage is a distinct n8n workflow. A single tracking key (the golden thread) ties every stage back to one submission, so every record downstream is traceable to the request that created it.
Intake & Session Gate
A payment-gated submission posts to the WF1 webhook. WF1 creates the intake record in PostgreSQL and initializes the tracking key that binds every downstream stage to that one submission.
This is also where the idea is dissected. The raw concept is broken into smaller search parameters — industry, CPC classification, general category terms, component concepts — that downstream workflows use for third-party calls. This decomposition is what delivers the confidentiality guarantee: from this point forward, only non-reconstructable fragments travel outside the pipeline. The whole idea stays in-house.
Google Research & CPC Extraction
Runs the decomposed parameters — never the raw idea — against Google APIs to surface existing products, competitors, and market context. Extracts and refines CPC classification pivots that target the next stage's patent retrieval.
USPTO Patent Retrieval
Queries the USPTO Open Data Portal (ODP) API across millions of publicly available patent records — the official source, not scraped web results. Retrieves the candidate set (typically 50–200 patents per run depending on the idea's breadth) into the patent queue for analysis.
Patent Analysis & Filtering
Analyzes the retrieved set, scores relevance against the decomposed idea parameters, filters noise, and builds the strategic analysis layer (density, assignee activity, landscape clustering).
This stage also scales its resource use to the chosen report tier — a higher tier triggers deeper analysis passes; a lighter tier runs an efficient subset. Cost and compute track the deliverable the inventor actually purchased.
Report Generation & Delivery
Assembles the report at the depth the chosen tier specifies and renders it through Gotenberg into a thorough, visually polished PDF that meticulously walks the inventor through their entire Idea Landscape. Delivers the finished report to the customer.
Before delivery, the final report is human-verified — the 5% that keeps "95% automated" honest.
What a Built-and-Running Research Pipeline Actually Is.
This is an architecture and economics story, not a vanity-metrics one. The numbers below are structural and operational — nothing here is a projection.
The pipeline is the product. Five chained workflows turn an unstructured idea submission into a publication-quality landscape report with zero manual research labor between intake and the human-verification step. 65 Postgres tables hold the research, retrieval, analysis, and report state — every stage's output is persisted and traceable to the originating submission via the golden-thread key. When a report is questioned, the trail shows exactly which parameters were searched, which patents were retrieved, and how they were scored.
The Gotenberg decision is the cost story. Report rendering originally depended on a paid third-party PDF-generation SaaS billed per document. OLI-Intel replaced it with self-hosted Gotenberg running on existing infrastructure. The result: publication-quality PDF output at effectively zero marginal cost per report, no per-document SaaS metering, and full control over the rendered document. At any volume, the rendering layer doesn't add a variable bill.
Tier-scaled compute keeps unit economics sane. Because WF4 scales analysis depth to the purchased tier, a $19.95 report doesn't consume $99-tier compute. Cost tracks the deliverable. The architecture was built so that the cheapest tier stays economically viable rather than subsidized.
Confidentiality is architectural, not contractual. The decomposition step in WF1 means the confidentiality claim doesn't rest on a third-party processor's privacy policy — it rests on the fact that no third party ever receives a reconstructable idea. That's a structural property of the pipeline, verifiable in the stage design.
Production-Tested Decisions.
What building and running this for real taught us.
Decomposition is the confidentiality architecture.
The strongest privacy guarantee isn't a clause — it's a pipeline that can't leak the whole idea because the whole idea never travels. Breaking the concept into non-reconstructable search parameters before any external call turned confidentiality from a promise into a structural property.
USPTO ODP beats scraping.
Querying the official Open Data Portal across millions of records produces a defensible, source-citable patent set. Web-scraped results can't carry the same provenance. For a report someone makes decisions on, the source has to be the source.
Tier-aware compute is what makes the cheap tier real.
Scaling analysis depth to the purchased tier is the difference between a $19.95 report being a loss leader and being a viable product. Resource use has to track the deliverable, or the unit economics quietly break.
Self-hosting the render layer removes a variable bill.
Gotenberg replacing a metered PDF SaaS removed per-document cost from the model entirely. At low volume it's a rounding error; at scale it's the difference between margin and no margin. Own the layers that bill per unit.
The 5% human step is the trust layer.
"95% automated" is the honest number. A human verifies the final report before it reaches the inventor. That last pass is not inefficiency — it's the reason the output can be trusted, and it's why the claim is 95% and not 100%.
Need a Research-to-Report Pipeline
for Your Operation?
OLI-Intel is the patent-intelligence instance of a general pattern Gold Root Solutions builds: a structured request comes in, the system researches it across authoritative sources, analyzes what it finds, and delivers a publication-quality branded PDF — with confidentiality built into the architecture, not promised in a contract. The same pattern fits legal research memos, due-diligence packets, and compliance reporting. Scoped as custom purpose-built work.
30-minute discovery call. We'll map whether the pattern fits your operation and tell you honestly what it costs to build.