Sensible Instruct
Sensible Instruct is a document extraction API for structured data from PDFs, images, and spreadsheets.
Last verified:
What is Sensible Instruct?
Sensible Instruct is a developer-first platform for extracting structured data from documents using a hybrid approach that combines LLM parsing with layout-based rules. The tool transforms unstructured and semi-structured data from business forms in PDF format, email bodies and attachments, spreadsheets, and PNG/JPEG document images into structured JSON output. It is built for mission-critical document processing where accuracy matters, solving the problem where pure AI extraction gets 80% of the way but breaks in production with hallucinated fields and silent model drift.
Key features include hybrid extraction combining LLM flexibility with deterministic guardrails, schema enforcement that validates every extraction against defined output schemas, and SenseML (Sensible's document-specific query language) for config-as-code that can be version-controlled and deployed through CI/CD pipelines. The platform offers 150+ pre-built configurations for common document types like bank statements, pay stubs, tax forms, ACORD forms, policies, invoices, and bills of lading. Additional capabilities include confidence scores for every extracted field, source coordinates for audit trails, automatic OCR, computer vision-enhanced table detection, free text summarization, invoice recognition, human review functionality, and document splitting. The platform is SOC 2 Type II certified and HIPAA compliant with full audit trails showing exactly where every extracted field came from.
Sensible is designed for engineers and developers who ship document extraction features to production, particularly vertical SaaS product builders who need to embed document-automation features into their products. It serves businesses of all sizes from startups to enterprises, including insurance companies (for extracting data from quotes, applications, certificates of insurance, loss runs, EOBs, policies), finance/lending companies (for bank statements, pay stubs, tax forms, offering memos, closing disclosures, rent rolls), and logistics companies (for rate confirmations, bills of lading, invoices). The platform is ideal for teams who want to focus on creating unique value instead of fighting to scale document processing efforts.
Sensible Instruct pricing
Pricing model: Freemium
Sensible offers transparent, predictable per-document pricing with 3 tiers, all starting with a 14-day free trial on the Growth Plan. Growth Plan: $499/month ($449/month billed yearly with 10% discount) includes 750 document extractions, 1 API call per second, unlimited document layouts, unlimited data fields, unlimited reference documents, access to onboarding and integration services, automatic/selective OCR, computer vision-enhanced table detection, free text summarization, invoice recognition, human review, and document splitting. Overage: $0.50 per additional document. Scale Plan: $1,499/month ($1,349/month billed yearly with 10% discount) includes 3,200 document extractions, 10 API calls per second, all Growth features plus dedicated customer success manager. Overage: $0.50 per additional document. Enterprise Plan: Custom pricing with volume tiers from 10,000 to millions of documents, 25+ API calls per second, everything in Scale plus white-glove customer enablement, HIPAA compliance, custom data retention policy, custom region availability, and custom SLA. All plans include unlimited document layouts, unlimited data fields, and unlimited reference documents. No per-page billing - billed per document regardless of length.
Sensible Instruct pros
- Hybrid extraction combines LLM flexibility with deterministic layout-based rules in single configuration
- Schema enforcement validates every extraction against defined output schema, failing fast not silently
- Config-as-code with SenseML enables version control and CI/CD pipeline deployment
- SOC 2 Type II certified and HIPAA compliant for enterprise security requirements
- 150+ pre-built configurations for common document types including bank statements and tax forms
- Per-document pricing (not per page) with no surprise overages regardless of document length
- Confidence scores and source coordinates on every extracted field for audit readiness
- Full observability with extraction logs, confidence scores, and complete audit trails
- RESTful APIs, SDKs for Python and other languages, and webhook support for seamless integration
- Automatic and selective OCR with computer vision-enhanced table detection
- Free text summarization included without additional cost
- Human review feature allows manual oversight of document extraction process
- Multimodal Engine extracts data from non-text images including charts, graphs, and handwriting
- Email-driven document extraction by forwarding emails with attachments to Sensible
- Scales from 750 documents per month to millions with linear per-document pricing
- Dedicated customer success engineers provide personalized onboarding and SenseML guidance
- Multi-region AWS infrastructure ensures reliable high-volume document processing
- No token-based cost volatility unlike pure AI extraction platforms
Sensible Instruct cons
- No free tier advertised publicly - only 14-day free trial on Growth Plan
- Starting price of $499/month may be expensive for small startups or solo developers
- Requires development knowledge for API integration and SenseML configuration
- API calls per second limited to 1 on Growth plan, 10 on Scale plan
- 50¢ overage per document after plan allotment can add up for unpredictable volumes
- Small company (2-10 employees) may raise concerns about long-term support capacity
- Relies on GPT-4 and LLMs which still have development limitations despite guardrails
- No offline mode - requires internet connection for all document processing
Frequently asked questions about Sensible Instruct
How does Sensible pricing work?
Sensible offers transparent and predictable pricing for document extraction by billing per document, not per page. This means no matter how long your document is (a 50-page contract costs the same as a 1-page form), you pay the same rate. Each plan includes a set number of document extractions (750 for Growth, 3,200 for Scale), and you pay $0.50 per additional document after your allotment. Enterprise customers get volume pricing tiers for 10,000 to millions of documents. This approach lets you budget confidently without worrying about surprise overages from page-based billing.
What happens if I exceed the usage limits of my plan?
When you exceed your plan's document extraction allotment, your extractions continue to be processed at the associated overage rate of $0.50 per additional document for Growth and Scale plans. For Enterprise customers, overages are billed at tiered volume pricing. You can upgrade your accounts at any time via your account page in the Sensible app to get more included extractions. There are no hard limits that stop your processing - you simply pay for what you use beyond your plan's inclusion.
What document types does Sensible support?
Sensible supports all document types through its hybrid extraction approach, with 150+ pre-built configurations for common documents. Out-of-the-box supported types include bank statements, pay stubs, tax forms, driver's licenses, utility bills, insurance documents (quotes, applications, certificates of insurance, loss runs, ACORD forms, policies, EOBs, payer confirmations), real estate documents (offering memos, closing disclosures, rent rolls), logistics documents (rate confirmations, bills of lading, invoices), and business forms in PDF format. You can also create custom document types by uploading your own example documents and configuring extraction prompts with SenseML.
What is SenseML and how does it work?
SenseML is Sensible's document-specific query language that combines LLM techniques with visual layout-based rules to extract structured data. With SenseML, you can preprocess documents (correct layout metadata, remove unwanted pages), use methods to extract primitives like rows, columns, tables, boxes, and checkboxes, handle document variations with conditional logic, and post-process extracted data (validate customer ID is 9 digits, concatenate/split fields, transform output schema with JsonLogic). SenseML configs live in your repo for version control, can be tested against sample documents, and deployed through CI/CD pipelines. The output is structured JSON with typed values.
How does Sensible ensure accuracy in production?
Sensible uses a hybrid approach combining LLM parsing with layout-based rules wrapped in deterministic guardrails. Every extraction includes schema enforcement that validates against your defined output schema - mismatches fail fast rather than silently entering your database. Each field comes with confidence scores so you know what to trust, and source coordinates showing exactly where in the document the data came from for full audit trails. The platform includes validation checks, extraction coverage scores, and a human review feature for manual oversight. Regular audits, penetration testing, and full observability with extraction logs help you monitor and verify data quality in production.
Is Sensible compliant with security standards?
Yes, Sensible is SOC 2 Type II certified and HIPAA compliant, making it suitable for handling sensitive data in regulated industries. The platform includes robust security measures with regular audits and penetration testing, configurable data retention policies, and full audit trails showing exactly where every extracted field came from. Enterprise plans offer additional compliance features including custom data retention policies and custom region availability to meet specific regulatory requirements.
What integration options does Sensible offer?
Sensible offers RESTful APIs with bearer token authentication, SDKs for Python and other languages, webhooks for real-time updates, and a quick-extract UI for manual testing. You can integrate by making API calls to endpoints like https://api.sensible.so/v0/extract/<TYPE> with your API key as bearer authorization. The platform supports email-driven extraction by forwarding emails with attachments to Sensible, and includes pre-built integrations for common workflows. API calls per second vary by plan (1 on Growth, 10 on Scale, 25+ on Enterprise). Custom integrations with existing systems are available through the customer success team.
What is the difference between Sensible and pure AI extraction?
Pure AI extraction gets you 80% of the way but breaks in production with outputs that vary between runs, hallucinated fields, and silent model drift when models update. Sensible solves this with a hybrid approach: LLM flexibility wrapped in deterministic guardrails. The layout-based rules provide consistency while LLMs handle document variation. Schema enforcement ensures every extraction matches your defined output format, confidence signals flag uncertain extractions for review, and full audit trails provide traceability. This combination gives you the flexibility of AI with the reliability needed for mission-critical production workflows.
Can I test Sensible before committing to a paid plan?
Yes, all trials start on the Growth Plan for 14 days with no credit card required upfront. During the trial, you get access to 750 document extractions, all Growth plan features including onboarding and integration services, automatic/selective OCR, computer vision-enhanced table detection, free text summarization, invoice recognition, human review, and document splitting. You can get an account at sensible.so, copy your API key from your account page, and start extracting data immediately. The quickstart guide includes example documents and code samples to help you get a 'hello world' API response quickly.
What support does Sensible provide?
Sensible provides dedicated support from customer success engineers committed to your success from onboarding through ongoing optimization. All plans include access to onboarding and integration services. Growth and Scale plans include personalized onboarding and setup assistance, expert guidance on SenseML configurations, and ongoing technical support with best practices advice. Enterprise plans include white-glove customer enablement and a dedicated customer success manager. The team offers help with custom integrations, SenseML configuration challenges, and tackling any document automation use case. You can reach support at [email protected] or [email protected].