Independent governance for sign language AI procurement
Novara Consulting Group helps governments, healthcare systems, universities, employers, and public institutions evaluate sign language AI before procurement or deployment, under the SLAT Index™: a published trust framework whose ratings are computed from evidence, not negotiated from marketing.
Analysis & Briefings
Start where you are
Read the Framework
What the SLAT Index measures, how trust ratings work, and why the method can be trusted: domains, criteria, ratings, and governance, in five tabs.
Commission an Assessment
Five engagement types, from a public-evidence Trust Snapshot to a full Trust Assessment with a computed rating for your setting. No license needed to engage.
Get the Volumes
The framework is published as licensed volumes: the instruments in force, the public record, the assessor guide, and the workbook. Read them, or license them to use.
Procurement is outrunning the evidence
Sign language AI is being purchased and deployed faster than the evidence for it can be established. Vendor demonstrations and conformance documents certify that signed output exists, not that it conveys meaning to Deaf users. Institutions are left approving systems they cannot independently verify, and when a system fails, the failure lands on Deaf and hard-of-hearing people first and the deploying institution second.
Evidence before adoption. Trust before deployment.
The SLAT Index provides the independent layer: criterion-based evaluation of claims, evidence, governance, accessibility risk, and institutional accountability, so the decision rests on what has been demonstrated.
Because your existing diligence cannot answer the question that matters
Conformance is not comprehension
An accessibility conformance report certifies that an interface works with assistive technology. It cannot say whether a Deaf person understands the output, and no standard procurement artifact can.
The evidence vacuum is real
Peer-reviewed literature documents that accuracy claims in this market frequently omit what was measured, against what reference, with which signers, under what conditions. SLAT grades the evidence and treats its absence as a finding, never a pass.
Failure lands in high-stakes settings
Hospital intake, emergency notification, classrooms, courtrooms: a failed output is a safety event and a civil rights exposure. Since 2018, the World Federation of the Deaf and WASLI have cautioned against signing avatars as replacements for qualified human interpreters in live, complex, or high-stakes information.
A decision you can defend
Findings are criterion-based, Deaf-reviewed, and documented, written for the people who must defend the purchase to a board, an auditor, or a court.
Five questions every deployment must answer
Linguistic Trust
Disclosed, validated, examined as language
Transparency Trust
Disclosure before it is asked for
Governance Trust
Accountable authority and data
Accessibility Trust
Real access, with a human answerable
Deployment Trust
The right system for this setting
Engineered so that supplying less evidence can never earn a better rating than supplying more, and a located defect is never confused with an undetermined record.
How the framework works →From request to finding
The full path →Request & scope
The vendor, the setting, the decision. We fix the engagement type and what the finding will cover.
Evidence & review
The record is gathered and worked through the criteria, with Deaf reviewer determinations wherever signed output is examined.
A defensible finding
Documented, criterion-based, checked before it publishes: written for the people who must defend the decision.
Evaluate before you deploy
The cost of a structured assessment is small. The cost of an access failure in a hospital, courtroom, classroom, or emergency is not.
Request an Assessment Browse the volumes →