Short answer: Insurance AI systems can be structured to avoid the EU AI Act's high-risk classification by limiting decision scope and maintaining human-in-the-loop architecture. Wednesday scopes this in a one-week discovery sprint before any code is written.
Your legal team has classified your claims triage AI as a high-risk system under the EU AI Act. Your product team built it without the conformity assessment your compliance officer now needs.
Retroactive conformity assessments are more expensive and slower than ones built into the development process. The architecture changes required after a conformity assessment flags gaps are typically larger than the ones that would have been needed if compliance was designed in from the start.
What decisions determine whether this project ships in 6 weeks or 18 months?
Four decisions determine whether the conformity assessment your compliance officer needs takes 6 weeks or 6 months.
Risk classification confirmation. Not all insurance AI is high-risk under the Act. AI that influences access to essential services or evaluates creditworthiness is explicitly listed. Claims triage AI that assists a human adjuster rather than making the final decision may fall into a lower risk tier. Getting your legal team to confirm the classification before the conformity assessment work begins could change the scope of that work significantly - and the cost of it.
Human oversight architecture. High-risk systems require a human in the loop who can understand, monitor, and override the AI output. The override mechanism has to be designed into the app from the start. A claims triage feature that presents AI results as final and buries the override in a settings menu doesn't meet the standard. The oversight UI has to be designed for the adjuster's workflow, not for the compliance reviewer's checklist.
Technical documentation. The Act requires documentation of training data sources, model architecture, accuracy metrics, and testing methodology. If you're using an open-source model with published documentation, that documentation exists and you need to reference it. If you fine-tuned on your own claims data, you're generating the documentation from scratch - and it needs to meet the standard the notified body will apply.
Audit logging. The Act requires logging sufficient to enable post-hoc review of AI-assisted decisions. The logging architecture has to produce records in a format a regulator can query - not just internal QA logs. The log has to capture the input, the model output, the human override decision if one was made, and the final claims outcome. Designing this before the first sprint prevents a mid-project rework that could delay your go-live.
Most teams spend 4-6 months discovering these decisions by building the wrong version first. A team that has shipped this before compresses that to 1 week.
On-Device AI vs. Cloud AI: What's the Real Difference?
| Factor | On-Device AI | Cloud AI |
|---|---|---|
| Data transmission | None — data never leaves the device | All inputs sent to external server |
| Compliance | No BAA/DPA required for inference step | Requires BAA (HIPAA) or DPA (GDPR) |
| Latency | Under 100ms on Neural Engine | 300ms–2s (network + server queue) |
| Cost at scale | Fixed — one-time integration | Variable — $0.001–$0.01 per query |
| Offline capability | Full functionality, no connectivity needed | Requires active internet connection |
| Model size | 1B–7B parameters (quantized) | Unlimited (GPT-4, Claude 3, etc.) |
| Data sovereignty | Device-local, no cross-border transfer | Depends on server region and DPA chain |
The right choice depends on your compliance constraints, query volume, and task complexity. Wednesday scopes this in the first week — before any code is written.
Why is Wednesday the right team for on-device AI?
We built Off Grid because we hit every one of these problems in production. Off Grid is the fastest-growing on-device AI application in the world, with 50,000+ users running it today.
It's open source, with 1,650+ stars on GitHub and contributors from across the world. It has been cited in peer-reviewed clinical research on offline mobile edge AI.
Every decision named above - model choice, platform, server boundary, compliance posture - we have made before, at scale, for real deployments.
How long does the integration take, and what does it cost?
The engagement is four sprints. Each sprint is fixed-price. Each sprint has a named deliverable your team can put on a roadmap.
Discovery (Week 1, $5K): We resolve the four decisions - model, platform, server boundary, compliance posture. Deliverable: a 1-page architecture doc your CTO can take to the board and your Privacy Officer can take to Legal.
Integration (Weeks 2-3, $5K-$10K): We ship the on-device model into your app behind a feature flag. Deliverable: a working build your QA team can test against real workflows.
Optimization (Weeks 4-5, $5K-$10K): We hit the performance and compliance targets from the discovery doc. Deliverable: benchmarks signed off by your team.
Production hardening (Week 6, $5K): Edge cases, OS version coverage, app store and compliance review readiness. Deliverable: shippable build.
4-6 weeks total. $20K-$30K total.
Money back if we don't hit the benchmarks. We have not had to refund.
"Wednesday Solutions' team is very methodical in their approach. They have a unique style of working. They score very well in terms of the scalability, stability, and security of what they build." - Sachin Gaikwad, Founder & CEO, Buildd
Is on-device AI right for your organization?
Worth 30 minutes? We'll walk you through what your version of the four decisions looks like, what a realistic scope and timeline would be for your app, and what your compliance posture and on-device target mean in practice.
You'll leave with enough to run a planning meeting next week. No pitch deck.
If we're not the right team, we'll tell you who is.
Book a call with the Wednesday team
Frequently Asked Questions
Q: Does the EU AI Act classify insurance AI as high-risk?
It depends on the decision scope. EU AI Act Annex III lists specific use cases that qualify as high-risk. Systems making or materially influencing consequential individual decisions fall under high-risk requirements. Systems structured as decision-support tools with mandatory human review can often avoid the classification.
Q: What technical requirements does the EU AI Act impose on on-device AI?
High-risk systems require: risk management, data governance, technical documentation, operational logging, user transparency, human oversight, and accuracy standards. On-device AI satisfies data sovereignty requirements more cleanly than cloud, but the other requirements apply regardless of deployment mode.
Q: How long does it take to ship an EU AI Act-compliant insurance AI app?
4–6 weeks for technical integration. Compliance documentation — risk management, technical docs, conformity assessment — adds 2–4 weeks in parallel if you don't have a compliance team already familiar with the Act.
Q: What does EU AI Act-compliant on-device AI cost?
$20K–$30K for technical integration across four fixed-price sprints. Compliance documentation scope varies by system classification.
Q: Can an on-device AI system avoid EU AI Act registration?
General-purpose AI models deployed as product components are subject to transparency obligations but may not require conformity assessment if the overall system is not high-risk. Classification depends on use case, decision scope, and affected population — resolved in the discovery sprint.
Top comments (0)