Speculative Design / Agentic AI
When AI Acts Autonomously
Designing Human Oversight for Agentic Claims Processing
Most AI systems in insurance today operate in an advisory capacity: they surface recommendations that a human then acts on. This speculative design explores the next step: what happens when the AI acts autonomously, approving routine payouts, flagging potential fraud, and routing exceptions to human reviewers, all without requiring human intervention for straightforward cases. The design challenge is not the AI itself. It is the oversight architecture that maintains trust, accountability, and regulatory compliance when no human reviewed the decision.
My Role
This is speculative design work, but it is grounded entirely in domain experience from production systems I have designed at multiple Fortune 500 financial institutions. I conducted the workflow research, developed the autonomy zone model, designed all three oversight interfaces, and validated the concepts with insurance operations leaders.
Responsibilities:
- Current-state workflow mapping and bottleneck analysis
- Three-zone autonomy model design and validation
- Interface design for agent dashboard, exception queue, and audit trail
- Concept testing with claims operations stakeholders
The Problem
A mid-size insurer processes roughly 50,000 claims a month. 70% are straightforward — water damage, minor collisions, routine medical — yet adjusters spend most of their time on exactly these cases, leaving too little attention for the complex claims that actually need expert judgment.
Mapping the current workflow (7 manual steps, 7-10 days to resolution) showed three natural zones of AI autonomy: Autonomous (~70% of claims, AI handles it end-to-end), Assisted (~20%, AI flags one decision point for a human), Escalated (~10%, AI organizes the data, human decides). The harder design question wasn't the AI — it was how an adjuster monitors 500 autonomous decisions a day without drifting into rubber-stamping, and what happens when the AI gets one wrong.
"Users don't need to watch every AI decision. They need confidence the system is functioning correctly, and immediate access when something goes wrong."
Process
Workflow Mapping
Three Autonomy Zones
Agent Dashboard
Design Strategy
Progressive Disclosure
Three interconnected interfaces, each for a different cognitive mode:
The Agent Dashboard — a control-room view of what the AI is doing in real time, closer to air traffic control than a claims spreadsheet.
The Exception Queue — when the AI isn't confident, it routes to a human with its analysis and the specific source of uncertainty already highlighted, not raw data to investigate from scratch. Cuts average review time from 45 minutes to under 4.
The Audit Trail — every autonomous decision logged with full reasoning chains, so any payout traces back to its inputs and logic. Auditability isn't a feature added on top; it's the architecture.
The hardest problem wasn't any of these interfaces — it was automation complacency: once AI handles 85% of claims, humans reviewing the rest start rubber-stamping. The fix was 'attention anchors' — a source-document confirmation checkbox plus randomized verification prompts, just enough friction to interrupt autopilot without hurting throughput.
Design Artifacts
Exception Queue
Audit Trail
Impact & Outcomes
85%
Auto-Resolution
Projected autonomous processing rate for routine claims
4 min
Review Time
Average human review time for AI-escalated claims (vs. 45 min today)
99.2%
Accuracy Target
Required accuracy threshold before autonomous processing activates
100%
Auditability
Every AI decision traceable to data inputs and model reasoning
What I Learned
Design the failure mode before the happy path
Before designing how the system processes claims successfully, I designed what happens when it gets one wrong. How does the system detect the error? How is the policyholder made whole? How does the audit trail preserve accountability? Starting with failure modes produced a fundamentally safer architecture than starting with the ideal flow and patching failures afterward.
Autonomy is a spectrum, not a binary
The most consequential design decision was rejecting the on-off framing that stakeholders initially proposed. Autonomy operates on a continuum that should flex based on claim complexity, adjuster experience level, regulatory requirements, and the model's own confidence calibration. The three-zone model makes this flexibility structural rather than ad hoc.
What I'd Do Differently
None of this is fictional — every pattern traces back to problems I solved in production (trust calibration from the AI underwriting work, progressive disclosure from the risk assessment platform, data density management from a bank analyst-tools project). What I'd do differently: bring claims adjusters into validation earlier, not just operations leaders — the people who process claims daily would have pressure-tested the three-zone model harder.
Read the Full Story
Discovery, in full
I mapped the current claims workflow end to end: seven manual steps, multiple specialist handoffs, averaging 7 to 10 business days from first notice to resolution. The inefficiency was structural, not individual.
From this mapping, I identified three distinct zones where AI could operate at graduated levels of autonomy:
Zone 1: Autonomous. AI handles the claim end-to-end. The human sees a summary dashboard but does not intervene. Approximately 70% of claims.
Zone 2: Assisted. AI performs 80% of the analysis, but flags a specific decision point that requires human judgment. Approximately 20% of claims.
Zone 3: Escalated. AI gathers and organizes all relevant data, presents its findings, but the human makes the final determination. Approximately 10% of claims.
The three interfaces, in full
The solution consists of three interconnected interfaces, each designed for a fundamentally different cognitive mode:
1. The Agent Dashboard. A control-room view that shows what the AI is doing in real time. Claims flowing through processing pipelines, auto-resolved cases accumulating, exceptions being flagged for review. The mental model is closer to air traffic control than to a traditional claims management spreadsheet.
2. The Exception Queue. When the AI is not sufficiently confident, it routes the claim to a human reviewer. But rather than presenting raw data and asking the adjuster to investigate from scratch, it presents its analysis, highlights the specific source of uncertainty, and suggests resolution options. This reduces average review time from 45 minutes to under 4 minutes.
3. The Audit Trail. Every autonomous decision is logged with full reasoning chains. Regulators can trace any payout back to the specific data inputs, confidence calculations, and decision logic that produced it. Auditability is not a feature layered on top of the system. It is the architecture the system is built from.
Automation complacency, in full
The most difficult design problem in this concept was not the AI interface itself. It was automation complacency.
When AI handles 85% of claims autonomously, the humans reviewing the remaining 15% begin to rubber-stamp. They trust the AI's pre-analysis too readily. Attention degrades precisely when it matters most, on the cases the AI flagged because it was uncertain.
I designed what I am calling 'attention anchors,' deliberate friction points that require genuine engagement with flagged cases. A confirmation checkbox stating 'I reviewed the source documents' paired with randomized verification prompts. The friction is minimal enough to preserve workflow efficiency, but sufficient to interrupt the autopilot pattern that high automation rates inevitably produce.
More of what I learned
- Regulators belong in the persona map: I included regulatory auditors as a primary user persona alongside claims adjusters and policyholders. Their needs (full auditability, decision chain transparency, override authority) shaped the information architecture as profoundly as the adjusters' workflow needs did. Treating compliance as a constraint rather than a user need would have produced a fundamentally different and weaker system.
- Speculative design draws from real production problems: Every pattern in this concept traces back to something I encountered in production: trust calibration from the AI underwriting work, progressive disclosure from the risk assessment platform redesign, data density management from an analyst tools project at a major bank. The value of speculative design is not invention. It is synthesis.
Next Case Study
AI-Driven Underwriting Assistant
From AI recommendations to autonomous action. See the stepping stone that got us here.