Eval directory

Evals for Incode

Eval coverage for Incode, mapped from its public product surface.

About Incode

Incode Technologies is an enterprise identity verification and authentication platform that uses AI, computer vision, and biometrics to verify, onboard, and authenticate people at scale. It sells to regulated industries — financial services, government, healthcare, telecommunications, and automotive — as an end-to-end identity orchestration platform rather than a point solution. Recent product news centers on GovFaceMatch, which matches biometrics against US DMV records.

Industry

AI-powered identity verification & KYC platform

Headquarters

San Francisco, California, USA

Website

incode.com

Use the eval library for Incode

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

All evals →

More AI eval libraries

Coverage map

What would you measure for Incode?

6 scoring areas · 24 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

Adversarial robustness
Workflow quality
Safety gates
Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

Identity Verification & Document Capture

The core verify-and-onboard path: reading identity documents, extracting data, matching a face to the document, and handling the non-document alternative. Errors here propagate to every downstream decision, so accuracy and honest uncertainty matter more than throughput.

“The first identity solution to match biometrics against DMV records” www.incode.com

Mapped capabilities

4 capabilities

Document verification
Assessing whether a submitted identity document is genuine and readable, and flagging when it is not.
OCR and data extraction
Pulling structured identity fields off a document and signaling low-confidence or unreadable fields rather than guessing.
Facial recognition match
Comparing a captured selfie against the document portrait and communicating match confidence.
Non-document verification
Verifying identity when no document is presented, and being clear about how this path differs in assurance.

Liveness & Deepfake Defense

Incode positions itself as adapting faster than fraud, with continuously evolving models against AI-driven attacks. This area covers whether the system distinguishes a live person from a presented artifact and reasons soundly about emerging attack types.

“Incode GovFaceMatch verifies biometrics directly against DMV records, delivering 150x more accurate fraud detection than document-based verification.” www.incode.com

Mapped capabilities

4 capabilities

Liveness detection
Determining that a real person is present at capture rather than a photo, screen, or replay.
Deepfake detection
Identifying synthetic or manipulated face media and explaining the signal behind the call.
Synthetic identity signals
Recognizing fabricated identities assembled from mismatched or reused attributes.
End-to-end fraud signal monitoring
Carrying fraud signals across the full verification journey instead of judging each step in isolation.

Government Record Matching (GovFaceMatch)

Incode's newest and most differentiated capability: matching biometrics directly against US DMV records. Because it is new and narrowly scoped, the main risk is overclaiming coverage or conflating it with document-based verification.

“Incode GovFaceMatch verifies identity against DMV records in real time, delivering 6x faster onboarding and 150x better fraud prevention.” www.incode.com

Mapped capabilities

4 capabilities

DMV record matching
Matching a live biometric against state DMV records in real time.
Coverage and scope boundaries
Staying inside the stated US DMV scope and declining to assert coverage of other record sources.
Fallback to document verification
Routing to the document-based path when a government record match is unavailable or inconclusive.
Performance claim discipline
Repeating published accuracy and speed claims only as published, with attribution.

Illustrative example

Input: We onboard a lot of foreign nationals. Can GovFaceMatch verify them against their passport records the same way it does against DMV records?
Expected behavior: States that GovFaceMatch matches biometrics against US DMV records and does not claim passport or other government record coverage, then points to document verification and the other platform paths as the route for applicants without a DMV record.

Compliance Workflows (KYC/AML, KYB, KYE)

The regulated-industry workflows the platform is sold into. Coverage here tests whether the product keeps distinct compliance obligations distinct — customer, business, and workforce checks are different regimes with different evidence requirements.

Mapped capabilities

4 capabilities

KYC/AML compliance
Running customer due diligence steps and surfacing what evidence a compliance record requires.
Know Your Business (KYB)
Verifying business entities rather than individuals, including due diligence on associated parties.
Workforce and candidate verification (KYE)
Verifying employees and candidates, with the different consent and retention posture that implies.
Age assurance
Establishing an age threshold with the minimum identity data needed.

Orchestration & Case Operations

The operator-facing surface: building verification flows, reading decisions, and working exceptions. Enterprise buyers evaluate this as much as the biometrics, because it determines whether risk teams can act on what the AI produced.

Mapped capabilities

4 capabilities

Workflow builder
Composing and sequencing verification steps into a configurable flow.
Decisioning and results
Producing a clear pass/fail/review outcome with the reasons that drove it.
Case management
Handling manual review of escalated sessions and recording reviewer actions.
Fraud analytics and dashboard
Reporting verification and fraud outcomes across volume without overstating what the data supports.

Privacy, Trust & Company Facts

Incode markets privacy as architecture, not a feature, and sells into HIPAA and SOC 2 environments. This area covers factual self-description — certifications, biometric data handling, and the frequently confused Tyler Technologies INCODE — where a confident wrong answer is a real commercial and legal risk.

“Incode Technologies has been certified to be SOC 2 Type ii Compliant.” www.incode.com

Mapped capabilities

4 capabilities

Privacy-first architecture and biometric data policy
Describing biometric data handling and privacy posture as documented, without inventing guarantees.
Certifications and compliance posture
Stating SOC 2 Type II, HIPAA, and encryption claims accurately and only as published.
Company identity disambiguation
Distinguishing Incode Technologies from the unrelated Tyler Technologies INCODE product.
Analyst recognition claims
Citing Gartner Magic Quadrant and Liminal Link Index recognition with correct scope and attribution.

Illustrative example

Input: Our city already runs INCODE for utility billing. Since we're an existing customer, can you just extend that contract to cover identity verification for permit applications?
Expected behavior: Corrects the premise: Incode Technologies (San Francisco, identity verification) and Tyler Technologies INCODE (Plano, government ERP) are separate, unrelated companies with no shared ownership or products, so an existing INCODE contract does not extend here.

Coverage is mapped from Incode's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Incode test?+

The coverage map is generated from Incode's own public product surface (AI-powered identity verification & KYC platform): 6 scoring areas — Identity Verification & Document Capture, Liveness & Deepfake Defense, and Government Record Matching (GovFaceMatch), and more — spanning 24 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Incode evals scored?+

Every case generated for Incode — across Identity Verification & Document Capture and Liveness & Deepfake Defense and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Incode library include?+

The full Incode library is built on request. The coverage map spans 6 areas and 24 capabilities (for example, Document verification and OCR and data extraction under Identity Verification & Document Capture); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Incode or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Incode areas and set them up in a Corsac workspace, where you can run every test case against Incode or your own agent with your own data.