Contract clauses and insurance claims
Clause-level extraction where the legal meaning of a phrase decides the field, and the claims documents built on top of it.
A contract clause is not a form field. On an invoice, “total” is a number in a box and the only question is whether you read it correctly. In a limitation of liability clause the cap may be a multiple of fees that have not been invoiced yet, sitting next to a list of carve-outs that are not capped at all; in a force majeure clause the words that decide a borderline event are the eleven words after “including but not limited to”, which is exactly the part a summariser deletes. Getting the characters right and the meaning wrong is the normal failure here, and it does not look like a failure in any confidence score.
These pages take one clause or one claims document each and work out what its language commits you to before deciding what the schema should hold. Several of them end in arithmetic, because that is what makes an extraction checkable: a declarations page has premium lines that must reconcile to a total, a loss run has claim rows that must foot to a stated aggregate, a beneficiary designation must sum to 100% within each tier. Where the check fails for a legitimate reason — a tax that is not premium, a claim valued at a different date — the page says why, because a validator that does not know the exception will reject correct data all day.
Extracting Fields From a Debit Note
Why a debit note and a credit memo are the same layout with opposite signs, and how to make the direction of the adjustment an extracted field rather than an assumption.
9 min read
Extracting Terms From a Non-Solicitation Clause
One sentence usually contains two different restrictions with different objects, and collapsing them into a single field destroys the only distinction that matters.
9 min read
Extracting Limitation of Liability Caps From a Contract
The cap is usually a formula over a fee schedule rather than a number, and the carve-outs that escape it are half the clause.
10 min read
Extracting Force Majeure Trigger Events From a Contract Clause
The clause is an enumerated list plus a catch-all read against that list, so the catch-all has to survive extraction verbatim.
10 min read
Extracting Assignment and Change-of-Control Provisions From a Contract
Whether a change of control counts as an assignment is decided by specific words, and those words are usually split across two clauses in different parts of the document.
10 min read
Extracting Warranty Disclaimers From a Contract's Boilerplate Section
Under the UCC a disclaimer of implied warranties must be conspicuous, so the formatting of the span is operative data and a plain-text pipeline throws it away.
10 min read
Extracting Delivery and Acceptance Criteria From a Purchase Contract
Deemed acceptance turns silence into a legal event, so the most important field is a date that appears nowhere in the document.
10 min read
Extracting Data Processing Terms From a Data Processing Addendum
GDPR Article 28(3) fixes most of what a DPA must contain, which turns the extraction into a conformance check against a published list — except that the substance lives in exhibits.
11 min read
Extracting Named Insureds and Additional Insureds From a Certificate of Insurance
Three different parties appear on one certificate and only one of them is the policyholder; separating them is the whole extraction.
10 min read
Extracting Loss Run History From an Insurance Loss Run Report
Per-claim rows across several policy years that must foot to a stated aggregate, where the amounts themselves change every time the report is run.
11 min read
Extracting Premium Breakdown From an Insurance Policy Declarations Page
Coverage premiums do not sum to the amount due, because taxes, fees and a separately disclosed terrorism premium sit between two different totals.
10 min read
Extracting Beneficiary Designations From a Life Insurance Policy
Designations are tiered and the shares within each tier must sum to 100%, with distribution instructions that are not shares at all.
10 min read
Extracting Property Details From a Homeowners Insurance Policy
The declarations page carries several lettered coverages with separate limits, some printed only as a percentage of another, and conflating them loses the structure.
10 min read
Extracting Claim Status History From an Insurance Claims System Export
The export is an event log, and the question people ask of it — how long a claim spent in each status — is not a column in it.
10 min read
Extracting Structured Data From a Subrogation Demand Letter
How to pull the demand amount, the loss date and the two different policy numbers out of a subrogation letter without mixing up which carrier owns which.
9 min read
Extracting Party Information From a Power of Attorney Document
Why scope of authority is the field that governs every other one, and how a statutory short form encodes grants and withholdings as marks rather than as text.
9 min read
Extracting Terms From a Living Will or Advance Directive
How to represent treatment-preference checkboxes so that an unmarked option is recorded as unstated rather than silently defaulted in either direction.
9 min read
Extracting Provisions From a Last Will and Testament
Why specific bequests and residuary shares cannot share a column, and how to keep distribution modifiers attached to the beneficiaries they govern.
9 min read
Extracting Trust Terms From a Revocable Living Trust Document
How to model the successor trustee chain as an ordered, conditional sequence instead of a list of names, and why a restatement is not an amendment.
9 min read
Extracting Deposition Testimony Into a Structured Fact Table
A pipeline that turns transcript testimony into sortable fact rows, each carrying the page-and-line citation that makes it checkable against the source.
11 min read
Extracting Exhibit Lists From a Trial Filing
How to build an exhibit-number-to-description table and cross-check it against the exhibits actually cited in the body text.
9 min read
Extracting Interrogatory Questions and Answers Into Paired Records
How to pair numbered interrogatories with their responses, and how to tell an objection-only response from one that actually answers.
9 min read
Extracting Structured Fields From a Subpoena
Which fields a subpoena actually carries, and how to turn a relative compliance period into candidate absolute dates from the date of service.
9 min read
Extracting Structured Charges From a Criminal Complaint or Indictment
Why every count needs its own record with its own statute citation, alleged date range and defendant list, rather than one block of charge text.
9 min read
Extracting License and Royalty Terms From an IP Licensing Agreement
How to extract a tiered royalty structure so that the arithmetic is reproducible, including whether the tiers are marginal or apply to the whole amount.
10 min read
Extracting Trademark Registration Details From a Trademark Certificate
How to read registration and serial numbers, per-class recitations and registration dates off a certificate, and derive the maintenance windows that follow from them.
9 min read
Extracting Claims Data From a Workers' Compensation First Report of Injury
How to get coded injury data out of a first report when one free-text narrative describes several body parts, and which of the form's dates is which.
9 min read
Extracting OSHA Recordable Fields From an Incident Report
Which facts an incident report has to yield for a recordability decision, and how the day counts and column selection follow arithmetically from them.
10 min read
Other topics
- LLM fundamentals & architecture
- Tokens, tokenization & context windows
- Prompt engineering
- Reasoning models & test-time compute
- Multimodal AI: vision, audio, video
- RAG & retrieval
- Embeddings & vector search
- AI agents & tool use
- Structured output & function calling
- Fine-tuning & post-training
- Local inference errors, string by string
- Running local models day to day
- Testing code that calls an LLM
- Snapshot and property testing for model output
- Regression suites for prompts
- Eval gates in CI
- Flaky tests against a model
- Determinism and the cost of testing
- Contract and streaming tests
- Testing tool calls and retrieval
- Inference, serving & latency
- Rolling out a prompt change
- Testing AI systems in practice
- Forecasting a time series
- Machine learning on tabular data
- Geospatial data and models
- Understanding audio that is not speech
- Understanding video
- Core computer vision tasks
- Machine learning on graphs
- Point clouds and 3D
- Evaluation, benchmarks & LLM-as-judge
- Sensor and IoT data
- Logs and event streams
- Models over biological sequences
- Machine learning on molecules
- Embedding and searching code
- Extracting invoices and purchase orders
- Receipts, statements and tax forms
- Insurance policies and contracts
- Deeds, court filings and patents
- Extracting from medical records
- Observability & LLMOps
- CVs, certificates and identity documents
- Shipping, customs and technical documents
- Meetings, email, chat and filled-in forms
- Building an extraction pipeline
- Business, property and inspection documents
- Regulated and compliance documents
- Consumer, travel and closing documents
- Mapping one chat API onto another
- SDK and framework migrations
- Hallucination & failure modes
- Re-embedding and model deprecation
- Cutting over between providers
- Parity gaps, shims and legacy endpoints
- Moving between model versions
- Migrating vector stores and caches
- Mapping capabilities and parameters
- Migrating pipelines and agents
- Contracts, runbooks and rollback
- Auditing a codebase before a cutover
- Compliance and fine-tune migration
- LLM cost engineering
- Routing, cost tracking and multi-tenancy
- What a migration does to your prompts
- AI security & prompt injection
- Privacy, compliance & data residency
- AI governance, policy & society
- Building reliable AI applications
- AI hardware, GPUs & compute
- Open-weight models & local inference
- AI for developers & coding agents
- AI in industry: vertical playbooks
- AGI, superintelligence, alignment & the long future
- Machine learning foundations
- NLP fundamentals & classical tasks
- Data engineering for AI
- Synthetic data & dataset curation
- AI product design & UX
- Search, ranking & recommendation
- Enterprise adoption & change management
- AI careers, skills & teams
- Reading AI research
- AI in science & discovery
- Robotics & embodied AI
- AI economics, markets & business models
- AI myths, hype & media literacy
- Context engineering
- Shipping AI features: patterns & anti-patterns
- Build it: end-to-end AI tutorials
- Python for AI: hands-on recipes
- TypeScript, React and the web
- Frameworks and SDKs
- Errors and troubleshooting
- AI facts, numbers and statistics
- The history of AI
- The maths behind AI
- Architectures beyond the transformer
- Reinforcement learning
- Diffusion and generative media
- Speech, audio and voice engineering
- Benchmarks, one at a time
- AI search visibility
- Infrastructure and operations
- Databases and storage for AI
- Knowledge graphs and structured knowledge
- Classical ML in production
- Regulation, jurisdiction by jurisdiction
- Prompt recipes and pattern library
- AI for people who do not write code
- Writing, media and creative work
- Edge and on-device AI
- Interpretability and model internals
- Field notes
- OpenAI model behaviour
- Claude model behaviour
- Gemini model behaviour
- Llama model behaviour
- Mistral model behaviour
- Qwen model behaviour
- DeepSeek model behaviour
- Cohere model behaviour
- Grok model behaviour
- Small model behaviour
- Hybrid model architectures
- Token cost by language and script
- Transliteration, romanization and script handling
- Locale-correct output
- Multilingual generation quality
- Multilingual pipelines
- The EU AI Act, article by article
- AI under the GDPR and EU data law
- US AI regulation, state and sector
- International AI governance and standards
- AI litigation and enforcement
- Running AI workloads on AWS
- Running AI workloads on Google Cloud
- Running AI workloads on Azure
- AI at the edge: Workers, Vercel and Netlify
- Serving models on Kubernetes
- Operating AI infrastructure
- Quantization formats and what they cost
- llama.cpp, flag by flag
- Ollama and the desktop local-model runtimes
- Local models on Apple Silicon
- Hardware for local inference
- Running speech and embedding models locally
- Model files, adapters and conversion
- VRAM arithmetic for local models