The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.
A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that automates returns, warranty claims, and refund processing including eligibility determination and approval. Includes automated claim adjudication and fraud screening; distinct from claims assessment in finance which handles insurance claims rather than product returns.
Returns, warranty, and claims processing automation has crossed from experiment to production at forward-leaning retailers and insurers, but the vast majority of organisations have not yet deployed it. That gap defines the practice's leading-edge position. Purpose-built platforms now process tens of millions of items annually, and the warranty management market is projected at $6.67B in 2026, yet most supply chain executives still report inventory trapped in returns pipelines and an inability to resell eligible goods quickly. The ROI case is proven for early movers -- vendors document 700x cost reductions on claims processing ($0.07 vs. $50 per claim), 68% processing-time reductions, and measurable revenue retention through exchange automation. Scaling remains the problem. Insurance claims adoption has accelerated sharply (8% to 34% full AI adoption YoY 2024–2025), driven by labor constraints and cost pressure, and regulatory frameworks (NAIC Model Bulletin adopted in 24 US jurisdictions; EU Right to Repair Directive effective July 31, 2026) are now accelerating rather than blocking deployment. Yet fraud is escalating faster than defences mature: deepfake fraud attempts surged 60x in three years (0.1%→6.5%), and only a small fraction of retailers consider themselves prepared for AI-enabled threats. Agentic orchestration systems (multi-layer decision architectures with human governance) are emerging as the production pattern, replacing legacy rules-based and single-function automation. However, critical adoption barriers persist: 95% of organizations report zero measurable ROI from general AI initiatives despite doubled enterprise spending; measurement rigor and governance frameworks remain immature; and warehouse-centric cost structures resist process-layer optimization. E-commerce returns platforms demonstrate 98% accuracy and sub-48-hour deployment, but the defining tension is now between proven unit economics and organizational readiness—implementation challenges, governance complexity, and fraud escalation keep most of the market in pilots rather than sustained production.
The vendor ecosystem has consolidated around a handful of scaled platforms deploying agentic orchestration. Optoro, now part of Blue Yonder, processes 25M items annually for 20 of the largest US retailers, achieving 60% waste reduction. Loop has processed 55M+ returns across thousands of Shopify brands, retaining $2B+ in merchant revenue through exchange automation. Narvar's IRIS engine handles 42B interactions for enterprise clients, with Shield fraud-prevention reaching general availability (40% revenue retention, 50% call centre reduction). Insurance claims automation is accelerating with multi-agent FNOL triage and routing systems: Pacific Northwest Mutual deployed coordinated AI agents achieving 68% FNOL-to-assignment time reduction and $2.4M annual savings; Taktile's Series C funding ($110M led by Goldman Sachs) positions one unnamed top-tier global insurer at $90M projected claims-processing cost efficiency. B2B warranty automation emerged as expanding segment: Continuum platform deployment achieves 70% labor cost reduction and <90-day ROI for distributors and manufacturers, validating warranty claim automation ROI in adjacent verticals beyond e-commerce.
Retail policy automation is emerging as a new deployment pattern. Major retailers (H&M, ASOS, Zara, Next, Oh Polly) now enforce tiered return fees and serial-returner detection rules through automated eligibility systems, reducing £6.6B in annual serial-return fraud while maintaining customer experience. E-commerce return rates remain 15.8% of sales ($849.9B market), with Shopify merchants reporting 85% AI adoption for fraud detection; refund automation achieves 70-80% touchless resolution and 92% cost reduction per claim ($0.62 vs. $7.40). Warranty automation vendors continue scaling: Dyrect processes across 300+ brands with 60% auto-approval rates; Happy Returns' product verification exceeds 99% accuracy.
Yet adoption barriers are hardening. Critical negative signals: 95% of organizations report zero measurable ROI from enterprise AI initiatives; deepfake-fraud production surged 60x in three years (0.1%→6.5% of attempts), with only 24.5% human detection accuracy vs. 50% chance; 64% of organizations report higher fraud losses YoY despite 98% integrating AI into workflows. August 2026 fraud data confirms acceleration outpacing defenses: Signifyd reports 33% YoY fraud increase, 78% surge in account takeover attacks, 175% spike in card-testing, 65% increase in BOPIS fraud—revealing AI-enabled sophistication undermining automation confidence. Regulatory acceleration (EU Right to Repair Directive effective July 31, 2026) creates new compliance automation requirements. The core tension: insurance claims adoption reached 34% (8% YoY), and vendor platforms demonstrate 68-75% cycle-time reductions—yet 42% of organizations with AI deployments still lack outcome measurement rigor, 40% report outcomes fall short of expectations, and governance complexity keeps deployment confined to early movers and regulated enterprises. Orchestration maturity is rising but inconsistent: leading deployments feature three-layer decision architectures (rules→AI→human review) with preserved human governance; many organizations remain on legacy task automation with poor integration and high failure rates.
— Critical negative signal: Signifyd Commerce Network (950M unique wallets) documents 33% YoY fraud acceleration, 78% account takeover surge, 175% card-testing spike, 65% BOPIS fraud increase; AI-enabled sophistication outpacing traditional controls, revealing adoption barrier to automation scaling.
— J.D. Power/BCG/McKinsey benchmarks: AI-enabled insurers cut claim resolution 75% (30 days→7.5 days), reduced cost-per-claim 30-40% ($25-36 vs $40-60), achieved 70-90% STP for routine claims vs 10-15% baseline; McKinsey finds only 7% of insurers scaled past pilot phase.
— Independent auto repair chain deployed 4-stage AI agent increasing warranty claim filing from 60% to 94%, recovering $191-239K annually; workflow automates eligibility detection, pre-population, submission, and reconciliation with 60-second manual review time per claim.
— B2B returns/warranty automation platform GA with named distributors/manufacturers reporting 70% labor cost reduction, 30% profitability increase, 8% churn reduction, 15-20 min to <3 min per warranty claim; sub-90-day ROI validates expanding automation scope beyond e-commerce returns.
— Appriss analysis across 1,000+ retail deployments quantifies $100B (14.2%) preventable returns loss; 'Warn and Approve' three-tier AI system reduces abusive returns 90% while protecting high-value customers (top 1% generate 50% revenue, return 8% less).
— Specialty insurer deployed AI achieving >90% intake automation, >10x speed reduction (15 days to ~30 min), $360K annual savings at 568% ROI; production deployment with full AI-gathered context for human escalation on complex cases, demonstrating orchestration ROI at scale.
— Deepfake fraud attempts surged from 0.1% to 6.5% over 3 years; production volume jumped 500K→8M (2023-2025). Only 0.1% of consumers correctly identify deepfakes. Critical detection barrier: human accuracy 24.5% vs 50% chance.
— Quantified adoption surge: full-scale AI adoption by insurers grew from 8% to 34% between 2024-2025; 50-75% cycle time reduction, 65% fraud detection improvement, STP automation handles 80-85% of health claims.