Document AI / From PDF to data
Document AI and OCR for invoices, contracts and PDFs
Friday Works combines Document AI, OCR, extraction models and validation rules to turn invoices, contracts and forms into structured data. Low-confidence fields are routed to a person before approved data enters a CRM, ERP or operational workflow.
- Indicative budget
- 15–35 million VNDPilot: up to two templates, fifteen fields each and 100 test pages
- Pilot timeline
- 3–5 weeksProduction-scale operation is quoted after the pilot data
- Control
- People review doubtful fieldsFields below the confidence threshold are never recorded automatically
- Output
- Excel or APIInto accounting, CRM, ERP or approval workflows within scope
Common problems
What businesses usually ask us to solve
Situation 01
The finance team keys in hundreds of invoices and receipts every month.
How Friday Works responds
The system reads PDFs or images and extracts supplier, document number, dates, line items and totals; reviewers only check flagged fields.
Situation 02
Contracts come in many templates, making dates and terms hard to track.
How Friday Works responds
Parties, terms, values and tracked fields become searchable, filterable data.
Situation 03
Free OCR shifted columns, misread figures and lost Vietnamese diacritics.
How Friday Works responds
OCR is only the first step. Document AI also understands layout, normalises data and cross-checks it, for example whether line items add up to the total.
Situation 04
You do not dare push data into accounting software automatically.
How Friday Works responds
Only data that passes validation rules and human review moves on, and every record links back to its source page.
Project types
Projects we typically take on
Invoice and document OCR
Extract supplier, document number, dates, line items and totals to reduce entry and support reconciliation.
Contracts and records
Identify parties, dates, clauses or tracking metadata across varied document templates.
Operational forms and PDFs
Turn unstructured content into records that can be searched, classified or used to trigger workflows.
When is Document AI worth deploying?
A good fit when
- Document volume is steady and repeats across a limited number of templates.
- You know which fields to extract and where the data must go.
- Someone can review uncertain fields during the early phase.
Not a fit when
- Documents are mostly handwritten or every document has a completely different layout.
- You require perfect accuracy without accepting a human review step.
How we work
Step by step: who does what and what you receive
Example schedule for a pilot with up to two templates, fifteen fields each and 100 test pages (package A03, 3–5 weeks). Production-scale operation is quoted once the pilot's accuracy data exists.
- 01Before signing
Conversation & document choice
Pick a high-volume document type with clear fields for the pilot.
Friday Works does
- Understand the document types, monthly volume and target system.
- Review a few masked samples to judge difficulty.
- Propose the templates, target fields and quality criteria for the pilot.
Your part
Send a few masked samples and explain where the data must go.
You receive
- A pilot proposal: templates, fields and quality criteria
- Schedule and quotation
CheckpointSign the proposal and pay the first instalment (40%).↗︎
- 02Week 1
Samples & schema
Test with poor documents too, not only clean ones.
Friday Works does
- Assemble a representative sample: many suppliers, varied scan quality and hard cases.
- Define the schema (field list): required fields, formats and normalisation rules.
- Agree cross-checks and the acceptable error level.
Your part
Provide masked samples and confirm the correct answers for the test set.
You receive
- A sample set and an answer-keyed test set
- Schema and validation rules
CheckpointAgree the schema and quality thresholds.↗︎
- 03Weeks 1–3
Pipeline & accuracy measurement
Measure field by field, because averages hide important errors.
Friday Works does
- Build intake, OCR, extraction, normalisation and validation.
- Measure accuracy per field and per document group.
- Set confidence thresholds that decide which fields go to a reviewer.
Your part
Review results on the test set and settle disputed cases.
You receive
- A pipeline running on the test set
- A field-level accuracy report
CheckpointResults demo and the second instalment (40%).↗︎
- 04Weeks 3–4
Review screen & data export
Reviewers see the source page right next to the extracted data.
Friday Works does
- Build the review screen: source page beside the data, doubtful fields flagged.
- Allow editing, approval and Excel export.
- Connect an API or target system if it is in scope.
Your part
Name reviewers and try the review screen.
You receive
- A review screen
- Excel output or an API connection
- 05Weeks 4–5
Live batch & pilot acceptance
Measure handling time and correction rate on a real batch.
Friday Works does
- Users process a real batch through the system.
- Measure handling time, correction rate and remaining errors.
- Propose the next step: more templates, integrations or volume.
Your part
Run the real batch and confirm the figures.
You receive
- A pilot report
- A production-scale proposal
CheckpointPilot acceptance and the final instalment (20%).↗︎
- 06After the pilot
Handover & quality monitoring
Quality is watched as new templates or poor scans appear.
Friday Works does
- Hand over source code, configuration and operating documentation.
- Show how to read quality metrics and handle new templates.
- Fix defects during the warranty period.
Your part
Tell us when new templates appear or forms change.
You receive
- Operating documentation
- Three months of defect fixes for Friday Works' implementation
Deliverables
What you receive
Samples & schema
- Representative samples and test setAnswer-keyed, used to measure objectively before and after every change.
- Data schemaFields, formats and normalisation rules for dates, amounts and tax codes.
- Quality thresholdsAcceptable error levels and the threshold for human review.
Extraction pipeline
- PDF and image intakeDirect upload, or from email or folders within scope.
- OCR and field extractionUnderstands table layout, line items and Vietnamese diacritics.
- Cross-checksFlags totals that do not add up, malformed dates or missing required fields.
Review & output
- Review screenSource page beside the data, with fast correction of flagged fields.
- Excel exportIn the structure the finance or operations team already uses.
- API or integrationSends approved data to the target system when in scope.
Measurement & handover
- Field-level accuracy reportShows which fields can be trusted and which need review.
- Operating figuresHandling time and correction rate on a real batch.
- Source code, documentation and warrantyHanded over under the contract with three months of fixes.
Technology and tools we commonly use
The exact stack is chosen for each project and recorded in the proposal rather than forcing one platform onto every problem.
- OCR and document modelsCloud services such as Google Document AI or vision-capable AI models, chosen for Vietnamese quality, cost and data requirements.
- Validation rules in codeChecks totals, formats, required fields and duplicates.
- Web-based review screenReviewers work in the browser with nothing to install.
- Excel export and APIsFeed approved data into existing workflows.
Collaboration
What both sides prepare
What you prepare
- A masked, representative sample set that includes poor scans.
- The list of fields to extract and the desired formats.
- Correct answers for the test set, or someone who confirms them.
- Reviewers who take part in the pilot.
- The target system's import template or API documentation.
What we check before handover
- Accuracy is measured per field, not only as an average.
- Fields below the confidence threshold always go to a reviewer.
- Cross-checks catch mismatched totals and malformed dates or tax codes.
- Every record links back to its source page.
- Reprocessing the same file does not create duplicates.
- Sensitive documents are stored and deleted within the agreed period.
Pricing
Indicative packages and how the budget is set
Related starting packages
Version 20 October 2026 · Unit: million VNDDocument extraction
15–35million VND / pilot
Estimated: 3–5 weeks
Baseline scope. Up to two document templates, fifteen fields per template and 100 test pages; PDF/image intake, extraction, review screen and Excel export.
Not included: Universal templates or handwriting by default; production-scale operation is quoted after the pilot.
These are indicative budgets for new engagements, not fixed quotations or prices from completed contracts. Scope, schedule and cost are confirmed for each project before work begins.
Typical timelines
- Document extraction pilot (A03)
- Often 3–5 weeks for up to two templates and fifteen fields each when representative samples are available.
- Several document types
- Split into stages according to template variation and target systems.
- Continuous tuning
- Quality is monitored as new templates, poor scans and form changes appear.
The budget depends on
- The number of document types, templates, languages and file quality levels.
- The number of fields and cross-validation rules.
- Document volume, latency and retention requirements.
- Review interface, system integrations and operational support.
Proposed payment schedule
- 40%Project start
- 40%Demo & core features
- 20%Acceptance & handover
Before a quote is final
- Agree the page, feature, data and acceptance-criteria list before implementation.
- The schedule starts after scope approval, receipt of required inputs and the first payment.
- The proposal includes three months of defect correction for Friday Works' implementation after acceptance.
- Domain, hosting, email, AI/API, data, licences, third-party services and applicable tax are shown separately.
After handover
Warranty, maintenance and next steps
Three months of defect fixes
Defects in the work Friday Works delivered are fixed for three months after acceptance and are not charged as maintenance.
Monthly maintenance
2–5 million VND / month
Up to 6 hours/month · One application; log review, compatibility updates and technical support.
Support is provided during business hours, not as a 24/7 service. Infrastructure and work above the included allowance are quoted separately. Warranty defects are not charged as maintenance.Frequently asked questions
Before we begin
01Can OCR be perfectly accurate?
That should not be assumed. Quality depends on templates, scan quality, language and target fields. The system needs a representative evaluation set and human review for uncertain fields.
02What determines invoice, contract or PDF OCR cost?
Invoice, contract or PDF OCR cost depends on document types and volume, scan quality, target fields, languages, validation rules, the expected human-review rate and destination-system integrations. A representative pilot measures field-level accuracy and real processing cost before scale-up.
03How is Document AI different from basic OCR?
OCR turns images into characters. Document AI also classifies documents, understands layout, extracts fields and applies validation rules to create data that a workflow can use.
04Can extracted data be integrated with an ERP or CRM?
Extracted document data can be integrated with an ERP or CRM when the target provides an API or suitable import mechanism. Mapping, approval state, idempotency and reconciliation should be designed with the document pipeline.
05Do e-invoices need OCR?
E-invoices do not need OCR: an XML file or source data should be read directly, which is more accurate than OCR. OCR and Document AI are for scanned PDFs, photos and paper documents.
06How are operating costs calculated after the pilot?
They depend on monthly page volume, per-call OCR or AI fees, the review rate and integrations. The pilot provides real figures for an estimate before the business decides to scale up.
Talk to Friday Works
Turn documents into actionable data
Share a sanitised sample set and the fields you need. We will help establish a baseline and a practical pilot scope.
Describe your projectAfter you send a request
- 01Send a short brief
Your goal, current situation and preferred timing; attachments are welcome.
- 02A reply within one business day
Friday Works sends clarifying questions or proposes an online call.
- 03Discuss the scope
Review the goal, risks and an approach that fits the budget.
- 04Receive a written proposal
Scope, milestone schedule, cost and acceptance criteria.