Most conversations about AI in property and casualty insurance start with the same promise: “faster, cheaper, smarter.” But in practice, the real question is where AI is being used.
Is it just doing the same work a little quicker… or is it changing the way underwriting, claims, and loss control actually run?
One of the cleanest ways to explain that difference is to borrow a framework from education technology: the SAMR model—Substitution, Augmentation, Modification, Redefinition—originally articulated by Ruben Puentedura. In SAMR, the first two levels are typically “enhancement” and the latter two are “transformation,” because they represent meaningful redesign (or reinvention) of the work itself.
In this post, I’ll map SAMR to the kinds of operational and strategic value AI can create across P&C insurance services (intake, underwriting, claims, fraud, and risk/loss services), staying away from customer chatbots and focusing instead on business process change that actually moves KPIs. Along the way, I’ll flag where AI and Insurance leaders tend to underestimate the “operating model” work required to reach the top of the SAMR ladder.
SAMR as a lens for AI in insurance operations
Here’s the core idea: SAMR isn’t a tech maturity model. It’s a work design model. It helps you tell the difference between:
- AI that replaces a step
- AI that improves a step
- AI that rebuilds the workflow
- AI that enables a new kind of service altogether
That matters in P&C because so many outcomes (loss ratio, expense ratio, cycle time, leakage, litigation rate, customer friction) are created not by a single model—but by how decisions move through a system.
A quick translation of SAMR into insurance terms:
| SAMR level | What changes | The “value type” you can credibly expect |
|---|---|---|
| Substitution | Same workflow, AI swaps in for manual effort | Cost takeout, capacity, fewer clerical errors |
| Augmentation | Same workflow, AI improves decision quality | Better triage, better consistency, reduced leakage |
| Modification | Workflow is redesigned around AI + humans | Cycle-time collapse, straight-through segments, fewer handoffs |
| Redefinition | New workflows/services become possible | New products, proactive risk services, new settlement models |
Puentedura’s original definitions are concise: substitution is “direct tool substitute,” augmentation adds “functional improvement,” modification allows “significant task redesign,” and redefinition enables tasks “previously inconceivable.”
Substitution: AI replaces manual steps, without changing the process
Substitution value is easy to spot because it looks like automation. The workflow stays the same; humans just stop doing the most repetitive parts.
In P&C services, substitution often shows up where the work is document-heavy and rules-heavy:
- Submission ingestion and data entry: extracting fields from ACORD forms, broker emails, loss runs, property schedules, and attachments into underwriting systems.
- FNOL and claim file intake: pulling structured data out of PDFs, photos, and inbound correspondence so adjusters start with fewer blank screens.
- Coverage and policy artifact handling: classification, indexing, and retrieval of endorsements, declarations, and forms so downstream work isn’t spent “searching.”
This is not the glamorous side of AI, but it’s where many carriers first find real ROI because it reduces friction across everything else.
A practical way to describe substitution value in a single sentence: it buys time back—especially in roles where experienced people are spending too much of the day doing clerical work that doesn’t require judgment.
It’s not that this is bad, or even undesirable, I’ve talked about how automating the happy path can have massive benefits, but this isn’t fundamentally changing the way people work.
Augmentation: AI improves decisions inside the existing workflow
Augmentation is where AI stops being a “robot typist” and becomes a decision quality amplifier. The process remains recognizable, but decisions become faster, more consistent, and more defensible.
In P&C services, augmentation value commonly shows up as:
Claims triage and segmentation
AI can score severity, complexity, fraud likelihood, and routing priority so the “right claim goes to the right lane” earlier. This is also where AI-based claims solutions often summarize documents and recommend next actions, reducing adjuster cognitive load.
Fraud detection and network signals
AI can surface suspicious patterns across entities and behaviors (not just single-claim anomalies), giving SIU teams better leads without requiring a total process rebuild.
Underwriting decision support
Instead of forcing underwriters to manually reconcile scattered facts, AI can highlight what matters most: appetite mismatches, missing exposure details, inconsistent statements, and unpriced hazards—without taking the decision away from the underwriter.
This is the level where #DataScience begins to pay off because the model output is tied to a human decision point. You’ll still measure classic outcomes (accuracy, precision/recall, drift), but the operational metric is usually something like touch time per file, reopen rate, referral rate, or leakage.
Modification: AI enables real workflow redesign
Modification is where the “same process, faster” story stops working—because you begin to redesign the process around what AI makes possible.
In P&C services, modification often looks like segmented operating models, taking our earlier “happy path” example and turbocharging it by making sure we get the right claims to the right place as soon as possible while still capturing the saved time:
- a lane where humans handle complex judgment
- a lane where AI + lightweight oversight handles routine volume
- a lane where exceptions are routed quickly and cleanly
Concrete examples:
Touchless / straight-through segments in claims
Computer vision can assess vehicle (and increasingly property) damage from images and produce estimates quickly enough to change the claim’s first decision moment. Tractable, for example, describes AI-generated damage assessments and repair estimates that can support settlement in minutes for appropriate cases.
The key point isn’t the model—it’s the redesigned workflow: fewer inspections, fewer handoffs, faster settlement on the right inventory.
Human-in-the-loop claim handling at scale
One company describes their tool as a generative AI tool built to automate and streamline claims handling while keeping human experts in control of decision-making. That’s a modification story: the work is re-allocated, not merely accelerated.
Underwriting intake that reshapes the submission pipeline
If AI extracts, validates, and pre-populates submission data—and routes it based on appetite and completeness—you can redesign underwriting around “exception handling” rather than “data chasing.” This is the operational logic behind newer underwriting workflow tools and assistants being introduced into the market.
When carriers get this right, the measurable value tends to move from “minutes saved” to cycle time collapse, improved indemnity control, and better risk selection, because the system is no longer bottlenecked by avoidable manual steps.
Redefinition: AI enables services that were previously impractical
Redefinition is where AI stops being a tool inside the carrier and starts changing what the carrier can offer.
In P&C insurance services, redefinition-level value usually involves at least one of these shifts:
From reactive claims to proactive loss prevention
Instead of waiting for a loss, insurers can deliver risk signals (and actions) continuously using external data, imagery, IoT, and predictive models—especially for commercial insureds. The “service” becomes: reduce frequency and severity, not just pay claims.
From periodic underwriting to continuous risk monitoring
When risk characteristics can be refreshed from updated property data, hazard signals, and exposure changes, underwriting becomes an ongoing process rather than an annual event. That opens the door to new policy structures and new service commitments.
From traditional indemnity to alternative settlement models
Parametric triggers, faster catastrophe response, automated severity estimation, and new forms of coverage transparency can create experiences that are genuinely difficult to deliver without AI-enabled workflows.
At this level, you’re no longer asking, “How do we process the same work faster?” You’re asking, “What work do we do now that we couldn’t do before?”
The hidden requirement: governance grows as you move up SAMR
There’s an important reason many organizations stall at augmentation: transformation requires trust.
Regulators and industry bodies are explicit that insurers should emphasize fairness, accountability, compliance, transparency, and safe/secure/robust systems in the development and use of AI. That expectation becomes more operationally demanding as you move from “assist” to “reshape” to “reinvent.”
This is where risk management stops being a parallel workstream and becomes a design constraint:
- You can’t redesign claims workflows around AI without strong monitoring, auditability, and escalation paths.
- You can’t deliver redefinition-level services without clear decision rights, data provenance, and controls that stand up under scrutiny.
Conclusion: SAMR turns “AI hype” into an operating model conversation
If you want a grounded explanation of AI value in P&C insurance services, SAMR gives you a simple, durable story:
- Substitution buys capacity.
- Augmentation improves decisions.
- Modification redesigns workflows.
- Redefinition changes what insurance services can be.
The most transformative outcomes aren’t created by a single model or a shiny interface. They’re created when AI is paired with redesigned workflows, clear decision rights, and governance strong enough to earn trust.
If you’re trying to prioritize investments, start by mapping one end-to-end service (submission intake → underwriting decision, or FNOL → settlement) to SAMR. The fastest path to value usually isn’t “do everything with AI.” It’s “pick the one bottleneck where a redesign changes the whole system.”
If you’re interested in how you can move your company to using AI to redefine your workflows, please let me know!
