Use AI to match invoice charges to your agreed rates. Inspect differences with the evidence beside them, record your decision and keep a clear review report.
Watch a recorded run of the working invoice workspace. AI proposes a description match; you confirm the agreement, inspect the difference and export your decision.
FREIGHTVEYRA / PRODUCT WALKTHROUGH
Actual product screen capture, recorded 4 November 2026. Playback is slowed for readability. No customer records, recovered savings or payment approvals are represented. No audio.
01
Enter the source
A synthetic diesel surcharge is billed at $450. The supplied fixed rate is $330, with a test agreement reference.
02
Inspect the AI match
The on-device embedding model proposes the fuel rate with 0.844 similarity in this run. Similarity is a ranking signal, not a probability of correctness.
03
Confirm and compare
The reviewer confirms the rate. Fixed-rate arithmetic flags a $120 potential discrepancy and displays the supplied source beside it.
04
Decide and export
The reviewer selects follow-up and exports a CSV containing the billed amount, rate, difference, evidence and decision. No money moves.
REPRODUCIBLE PRODUCT TEST
Check the result yourself.
Use the same synthetic inputs in the free workspace. The expected fixed-rate difference is $120. Similarity and run time can vary by model runtime and device. This checks a specific path; it is not an accuracy benchmark.
Invoice reference: SYNTHETIC-REVIEW-001. Carrier: Synthetic test — not a customer. Currency: USD. Charge L1: Diesel fuel surcharge, 450. Rate FUEL: Fuel surcharge for diesel — test agreement section 3, 330.
“Suggest rate matches” runs Xenova/all-MiniLM-L6-v2 in the browser. The captured run reports 0.844 similarity and 132 ms of model inference after the model is available. This is one measured run, not a speed guarantee. The user confirms FUEL, then runs the audit. The result shows $450 billed, $330 agreed and $120 potential difference. The user selects “Accept for follow-up”, then “Export CSV”. The downloaded row records the decision as accepted.
AI ranks description matches. Deterministic arithmetic calculates differences. People validate agreement applicability and decide what to do next. Current input is entered text or CSV; this recording does not demonstrate PDF extraction, an Azure deployment, automatic dispute handling or payment execution. The pretrained model is not a FreightVeyra-trained proprietary model.
PUBLISHED INDUSTRY CONTEXT
A practical problem. A documented use for AI.
These independent customer stories explain why freight-invoice analysis and billing exceptions matter. They support the use case; they do not validate FreightVeyra’s performance.
MICROSOFT CUSTOMER STORY / 19 NOV 2024
Dow: reviewing freight invoices with AI
Microsoft describes Dow using Copilot to compare freight invoices with negotiated rates and identify discrepancies. Its story includes invoice-analysis and PDF-processing agents, with projected savings.
Relevant to our workflow: connect charges to agreement evidence, surface differences and support a review decision. FreightVeyra’s current input is text or CSV; it does not replicate Dow’s email and PDF automation.
AWS describes an Amazon Bedrock solution for pricing variances, mismatched line items and unexpected accessorial charges. The workflow includes human approvals. AWS reports a 75% reduction in active exceptions for Worldwide Express.
Relevant to our workflow: organise exceptions around source evidence and human review. FreightVeyra does not claim that result or the Bedrock and transport-system integrations in this case.
Dow and Worldwide Express are not presented as FreightVeyra customers, partners or endorsers. Their results belong to their respective implementations. FreightVeyra’s own evidence above is a synthetic product test.
LIVE PRODUCT / EARLY-ACCESS MVP
Working software. Ready for your own data.
Customers run AI matching, fixed-rate checks and report exports themselves. The same application serves every team; founder support is optional.
On your device: try the audit without an account. Saved reports stay in that browser.
Export and retain: download a backup or CSV for your own records. Cloud reports use a separate workspace and sign-in; they are not synchronized with this browser.
AI and cloud infrastructure
The invoice matcher uses MiniLM-family sentence embeddings through Transformers.js and Microsoft's open-source ONNX Runtime Web. The sentence-transformer model is a community-trained model, not a Microsoft endorsement. Matching runs on your device. Public website delivery uses Netlify; the separate cloud workspace uses Cloudflare Workers and R2. Azure cloud services are not connected.
A focused software product for finance and logistics teams. Try the same review steps before discussing a pilot.
01 / THE INPUT
Invoice + agreed rates
Enter your charge descriptions and fixed rates, or import CSV. Use permissioned, non-sensitive examples for your first evaluation.
02 / THE INTELLIGENCE
AI suggests a match
A sentence-embedding model ranks descriptions by meaning. You confirm each match. Uncertain or conflicting matches need your attention.
03 / THE OUTPUT
A review you can explain
Inspect billed and agreed amounts, the difference and your supplied source evidence. Record a decision and export the review.
A PILOT WITH A PURPOSE
Measure the workflow with your team.
Agree a permissioned sample and success measures before starting: matching usefulness, unresolved charges, review time and the quality of the exported record. Potential differences are not confirmed savings.
Prepare payroll alongside your freight-cost review. A separate working workspace for fixed and hourly pay, clear adjustment notes and a reviewable register.
PREPARE
Use your own pay data
Start an empty pay run or import CSV. Define the period, one currency and your worker references.
REVIEW
Explain every adjustment
Calculate base pay, allowances and authorised deductions. Check each person and record your review.
HAND OVER
Export a clear register
Download CSV or a backup, print your review, or explicitly save a draft on your private device.
Payroll preparation is available now. Transfers, tax calculation and statutory filing are outside this release.
This illustrative invoice is already matched to fixed agreement rates. Edit the linehaul or fuel amounts and recalculate. To match your own descriptions with real AI, open the full workspace.
01 / A clearer way to review
Read the charge. Review the reason.
A practical review desk for finance and logistics teams: connect each charge to its rate, investigate differences, and record the decision.
01 / BRING THE DOCUMENTS
Start with the source.
Review a structured invoice and its agreed rates. Free neural matching connects invoice descriptions with agreement rates, directly in your browser.
02 / INVESTIGATE
See what changed.
Inspect billed amounts, agreed rates and exact differences. Unsupported charges stay unresolved until evidence arrives.
03 / TAKE CONTROL
Make the call.
Accept a finding for follow-up or dismiss it. Export a review report and draft carrier questions for your own approval.
02 / Built for accountable AI
Match by meaning. Approve with evidence.
A sentence-embedding model ranks possible agreement matches. You confirm the relationship; the audit calculates fixed-rate differences and preserves the supporting text.
MVP scope: fixed amounts per charge code in USD, EUR, GBP or NGN. Conditional surcharges, quantity-based rates and complex clauses need manual review. Free AI matching runs locally. PDF/OCR and generative document extraction are not enabled in this release.
COMPANY / AT A GLANCE
Building FreightVeyra.
$200,000Funding raised over three years
25Team members
Company figures provided by founder Bethel Clement · October 2026.
03 / Founder-led
Bethel Clement
Founder & CEO, FreightVeyra AI
FreightVeyra AI is led by founder and CEO Bethel Clement, with a 25-person team focused on evidence-led freight cost review. The MVP combines local neural matching, fixed-rate comparisons, and a clear human review step. The next step is a measured pilot with finance and logistics teams using permissioned invoice examples.
FreightVeyra helps finance and operations teams at shippers and freight operators check carrier or supplier invoices against fixed rate agreements. It brings charge matching, potential discrepancy checks and review records into one workflow.
What does the AI actually do?
A real transformer model compares the meaning of invoice and agreement descriptions. Explicit charge terms can support the match. You review and confirm suggestions before fixed-rate arithmetic calculates differences. Uncertain and conflicting matches stay unresolved.
What can I use in the MVP today?
Enter charges in editable tables or import CSV, choose USD, EUR, GBP or NGN, run AI matching, inspect evidence, record decisions and export a review. Private cloud reports are available in the separate signed-in workspace. Each invoice uses one currency.
Can I upload PDFs or use complex freight contracts?
The current release uses CSV or entered data and fixed amounts per charge code. PDF/OCR import and active generative extraction are planned. Conditional surcharges, quantities, tax interpretation and complex contract clauses require manual review.
Do I need an account, and where is my data stored?
The device-local workspace needs no account: matching, calculations and saved reports stay in that browser. In the cloud workspace, invoice fields are sent to the backend for the audit, and saving a private report requires sign-in. Model downloads involve normal network requests to the model and runtime hosts. Read our privacy notice for details.
Are flagged differences confirmed overcharges or savings?
No. They are potential discrepancies against the rates you supplied. Verify the agreement applies to the shipment before following up. Missing evidence and repeated charge codes stay unresolved. FreightVeyra does not approve payments or send disputes automatically.
What does it cost, and how do I start a pilot?
The public demo is free. A guided 30-day pilot is $149, with scope and suitability confirmed before payment. Broader evaluations receive a written quote. Compare plans and pilot details. Email bethel@freightveyra.site with your invoice volume, file formats and the freight-review problem you want to solve.
Let’s build the right workflow.
A better freight review starts with a real invoice.
We’re inviting finance and freight teams to test the review workflow with permissioned examples.
Sample data only. Findings are potential discrepancies, not confirmed savings.
Invoice total
—
Potential discrepancy
—
Charges to review
—
Loading sample…
Charge
Billed
Agreed
Difference
Status
Draft carrier questions
Review before sending. Nothing is sent automatically.
Start a new audit
Enter charges and agreed rates, review AI suggestions, then inspect the differences.
Input format and supported scope
Use currency, invoice, carrier, lines and rates. Each line needs id, code and amountCents. Each rate needs code, amountCents and source. Amounts are integer minor units. Match one fixed rate per charge code; do not use this prototype for conditional or quantity-based pricing.
Advanced document extraction is not included in this drag-and-drop edition. Use Free AI matching for real local model inference. Review every extracted field before running the audit. PDF/OCR upload is not included in this release; paste the document text.
Advanced extraction is disabled in this edition. Text entered here is not submitted to an AI provider.
Processing your review…
Reports on this device
Your saved reports appear here on this device.
Human review required · Evidence references supplied with your data · No automatic payments or disputes
EARLY ACCESS / B2B PRICING
Start small. Review with confidence.
Use the working MVP for free. When your team is ready, run a guided pilot using a defined set of freight invoices and agreed rates.
Transparent pilot scopeHuman review built inNo annual commitment
EXPLORE THE WORKFLOW
Free workspace
Use AI matching and evidence-led review with your own entered data or CSV.
Integrations, team roles and document extraction are scoped separately as future development.
All prices are in USD, excluding applicable taxes. Paid pilots begin after we agree scope and confirm suitability. Payment is arranged directly; this site does not collect card details or start subscriptions.
Business model: B2B subscription software with organisation and invoice-volume tiers as commercial hosting expands. The current MVP has a free self-service workspace and an optional one-time product pilot. Recurring subscription checkout is not yet active. Read the product and business model.
A PRACTICAL FIRST STEP
From sample to your first review.
Start with one invoice type and a clear agreement. Expand after you know what works.
01
Try the workspace
Test AI matching, inspect a difference and export a review.
02
Agree your pilot
Tell us your volume and file format. We confirm scope, fit and the fee before starting.
03
Measure the workflow
Review findings together and decide whether the workflow is worth expanding.
WHAT YOU CAN USE TODAY
A focused MVP. A clear scope.
The self-service software is available now. The optional pilot adds onboarding and evaluation support.
Available now
AI matching with human confirmation
Entered data and CSV import
Fixed amounts per charge code
USD, EUR, GBP or NGN per review
Evidence, decisions and report exports
Planned capabilities
PDF and OCR invoice import
Generative document extraction
Shared team roles and permissions
ERP and freight-system integrations
These capabilities are outside the current pilot offer.
BEFORE YOU COMMIT
Your pricing questions.
What counts as one invoice review?
One invoice, in one currency, checked against supplied fixed agreement rates. Each review accepts up to 100 invoice lines and 100 rate entries. Quantities, taxes, conditional surcharges and complex contract clauses need manual review.
Is there a subscription or automatic renewal?
No. The guided pilot is a one-time $149 fee for 30 days with an agreed start date. Continuing afterwards requires a new agreement. The free workspace remains available at no charge.
What happens if my invoice volume is higher?
Contact us to discuss a higher-volume software evaluation. We agree the volume, support, timeline and price in writing. No usage overage is billed automatically.
Where are reports stored?
The free local workspace saves reports in the browser you use. A separate signed-in cloud workspace supports private reports, with a current limit of 100 saved reports per account. Local and cloud histories do not automatically sync. See the privacy notice before using sensitive data.
Does the pilot include recovered savings or payment approvals?
The product flags potential differences against the rates you supply. Your team checks applicability and makes every decision. The software supports customer-led review. A pilot measures product usefulness; it does not guarantee recovered money or approve payments.
How do I start?
Open the free workspace or request a guided product pilot. Include your company, approximate invoice volume, file format and the problem you want to test. Bethel will confirm the scope and next steps directly.
OPTIONAL PRODUCT SUPPORT
Let’s review your first use case.
Speak directly with Bethel Clement about your team’s freight-review process.