Bdeo

Without the computer vision there is no product, only a photo upload form. Underwriting inspection and claims triage run on the same detection and severity models, so a weakness there carries into both.

UK material calls the product an added layer of visual verification, while the founding thesis rests on assessing about 70 percent of minor claims without a person. What the system decides alone, what routes to an assessor and who reviews a severity figure before settlement are never stated.

Damage detection and severity have a measurable accuracy against expert assessment, and none is published, nor any false positive rate or benchmark against adjusters. Drift as vehicle models and capture devices change is unaddressed, and investor remarks on reliability are not measurement.

Reale, Mapfre and Generali are among named carriers across more than 50 insurers in 25 countries, and Mutua Madrileña co created an underwriting automation system. No automation rate, accuracy or cost saving is published, and the Spanish motor penetration figures date from mid 2023.

Vision models learn from labeled damage imagery, and nothing published says whether client claim images train models or inform capability sold to other insurers. Pre existing damage checks re query images captured for earlier claims, a repurposing no published document bounds.

Scene photos and video routinely capture plates, home interiors and people, personal data under European law by any reading. No privacy policy content, retention schedule or subprocessor list says how long images are kept or what is redacted.

Security goes undocumented: no trust center, attestation report or certification appears on the site or in press coverage. A smaller vendor gives a buyer less public corroboration before a confidentiality agreement.

A European Innovation Council Seal of Excellence and public grants recognize innovation, not supervision of outputs. How the product supports carriers' claims handling obligations across more than 25 jurisdictions is not addressed, and the company claims no insurance license.

Motor and household damage assessment sits outside the EU AI Act's high risk schedule, which lowers compliance exposure without settling fairness. Vision accuracy varies with image quality, lighting and device, which track claimants' circumstances, and no testing of that is published.

A severity figure sets what a claim is worth, and a pre existing damage finding can defeat a claim outright, acting as an accusation as much as an assessment. No published process gives the policyholder notice or a route to human reassessment, and nothing allocates liability between vendor and insurer for a wrong finding.

Proprietary vision dating from 2017 and 65 percent of staff on the technology point to internal models, though no statement confirms it. Training data assembly, any third party vision service and change notification are undisclosed.

Integration reaches insurers, policyholders capturing at the moment of loss and repair workshops whose estimates are validated against image evidence. In Germany, telematics providers detect an incident and start capture automatically, though no interface documentation or partner list appears.

Hosting provider, processing location and residency options are undocumented for a customer base spanning Europe, Latin America and South Africa. A historical image library implies persistent storage of claim imagery whose location is undescribed.

A demo request is the only route, and reduced operating cost is claimed repeatedly without a figure or unit. Implementation by use case suggests per inspection charging, but whether images, claims or a subscription set the price is unstated.

Deployments run from Spain and Portugal to Mexico and South Africa, and buyers include fleet operators, leasing and rental firms alongside insurers, with workshops in the workflow. Only motor and household property are covered, leaving commercial, specialty and life lines out.