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 monitors shelf stock levels, tracks product placement, and analyses in-store customer movement and behaviour patterns. Includes planogram compliance and dwell-time analysis; distinct from checkout-free retail which automates the transaction rather than analysing the store.
AI-powered shelf monitoring and in-store customer analytics is a proven practice with quantified ROI, production deployments across Fortune 500 retailers, and expanding mid-market adoption. The technology works: autonomous robots and camera-based platforms routinely achieve 99%+ inventory accuracy, while multi-retailer randomized trials and ML models have validated the business case with 4-8% sales uplift from inventory accuracy corrections and demonstrated 60-81% accuracy in predictive models flagging high-risk inventory records. However, organizational readiness remains the true bottleneck. Data fragmentation across disconnected ERP/POS/WMS systems, batch-oriented architectures that rely on stale (hours-old) data, and weak data governance prevent 60% of retail AI initiatives from reaching production scale. Even technically sound deployments fail when integrated into unoptimized workflows: Starbucks' NomadGo system, despite working robotically, was abandoned after 9 months (2025-2026) due to governance, change management, and prerequisites (supply chain consistency, infrastructure, forecasting maturity) that CV accuracy alone cannot overcome. Vendor consolidation (Gemspring's FORM-Trax merger), service-based delivery models (ShelfOptix, Simbe for Merchants), and expanding scope into pricing and markdown automation all signal a practice transitioning from capability validation toward implementation-at-scale. For enterprise retailers with mature data infrastructure, that transition is well underway. For the 65% of mid-market retailers still operating with <70% inventory visibility, organizational transformation and data orchestration remain the binding constraints.
The vendor ecosystem is consolidating around integrated platforms. Gemspring Capital's merger of FORM and Trax Retail's image recognition business created a unified retail execution platform serving 750+ customers, including 30 of the top 50 CPG companies. That consolidation pattern — combining task management with shelf analytics — reflects a market that has moved past point solutions. The market itself is stratified: 60% of enterprise retailers have full store-intelligence deployments operational (up 18 percentage points year-over-year from 2025), while only 42% of mid-market retailers (<$1B revenue) have reached similar maturity. On the robotics side, new entrants continue to prove out the hardware model: Leon's deployed Cypher Robotics' Captis across 300+ stores, achieving 99.8% inventory accuracy and cutting cycle-counting time by 90%. PAL Robotics' StockBot, Simbe's Tally 4.0, and Ubica's autonomous scanning robots are expanding the vendor bench.
Service-based delivery is reshaping adoption economics. Brain Corp and Driveline's ShelfOptix offers fully managed robotic shelf scanning, while Simbe for Merchants provides SaaS analytics dashboards across hundreds of stores. These models shift cost structure from capital expenditure to operational, removing the ownership barrier that stalled earlier waves. Simbe's global footprint spans ~1,000 stores across 3 continents in 5 industries, achieving 60% reduction in out-of-stocks, 90% reduction in pricing errors, and 50% reduction in order fulfillment time. The ecosystem is diversifying beyond shelf recognition: ML models now predict inventory inaccuracies with 60-81% accuracy and 51-84% of items behave predictably enough for targeted intervention, enabling proactive rather than reactive approaches. Kenvue's ShelfScan deployment across Africa, Middle East, and Europe demonstrated 50% field productivity gain, 98% inventory accuracy, and 40% year-over-year sales growth in enabled stores—validating the business case across geographies. The broader in-store analytics market is $5.58-6.28B (2026) across three segments: ESL (Electronic Shelf Labels, $2.75B+ by 2034), computer vision (fastest growing at 24.28% CAGR through 2031), and RFID systems. Named deployments show ROI: Walmart's 2,300 ESL installations shortened label-change time from days to minutes; Morrisons completed 10.8M ESL installations; Focal Systems achieved 99% on-shelf availability in pilots; Albertsons targets $1.5B in productivity gains over three fiscal years.
Market analysis confirms ongoing expansion despite implementation challenges: Coresight Research quantifies in-store execution inefficiencies at $196.4B annually (6.4% of gross sales). The business case is clear: multi-retailer randomized controlled trials across 7 European retailers and 1M SKUs found 4-8% sales lift from correcting inventory inaccuracies. However, a critical gap persists: despite clear ROI evidence, 95% of retail AI pilots produce zero measurable business return, with 85% of failures traced to poor data quality rather than algorithmic limitations. Gartner projects 60% of AI projects without AI-ready data will be abandoned through 2026. The root cause is structural: inventory data sits across disconnected systems (ERP, POS, WMS, ecommerce), each maintaining different schemas and refresh cadences, with batch synchronization creating hours-old information that defeats real-time decision-making. Specialized retail vendor tools succeed 67% of the time versus internal builds at 33%—suggesting domain expertise and data orchestration matter more than raw compute.
However, critical barriers remain despite quantified ROI and proven technical maturity. Starbucks abandoned its NomadGo AI inventory system after only 9 months (September 2025–May 2026) due to accuracy failures (miscounts, mislabeling, missed items). Post-mortem analysis attributed 97% of root cause to governance and change management rather than technology immaturity—revealing that prerequisites like supply chain consistency, infrastructure maturity, and forecasting accuracy must precede CV deployment. Even where detection accuracy is high, 75% of identified issues go unresolved within 48 hours without dedicated response teams, indicating that data-to-action workflow integration remains the binding constraint. Practitioners identify workflow coordination and organizational readiness as the residual barrier: "retailers are automating broken processes instead of redesigning them first." Data fragmentation across CPG and large retail organizations creates a "silent failure" where bad inventory data flows into forecasting models and recommendations undetected, surfacing as mispricing, stockouts, and margin erosion days later—after action windows close. The technology has outpaced organizational capacity, and data orchestration now gates adoption more than technical capability or market readiness.
— 7-retailer RCT across 1M SKUs and 100 stores found 4-8% sales lift from inventory accuracy corrections, validating ROI foundation.
— ML predictive model achieved 60-81% accuracy flagging SKU-store records with inaccuracy (51-84% behave predictably), validating AI capability.
— FORM CRO affirms "CV has largely solved the recognition problem at shelf" but identifies downstream orchestration and data quality as binding constraints.
— Gartner projection 60% of AI projects without AI-ready data abandoned through 2026; Concentrix identifies data fragmentation across disconnected systems as root cause.
— $196.4B annual in-store execution losses, 60% enterprise deployments scaled, BJ's 40% picking efficiency and Albertsons $1.5B productivity target.
— Kenvue deployed ShelfScan across Africa/ME/Europe achieving 50% field productivity gain, 98% inventory accuracy, 40% YoY sales growth in enabled stores.
— 95% retail AI pilots produce zero ROI, 73.8% fail; 85% of failures trace to data quality; specialized vendor tools succeed 67% vs internal builds 33%.
— $5.58-6.28B market with three segments (ESL, CV, RFID); named deployments: Walmart 2,300 ESLs, Morrisons 10.8M ESLs, Focal Systems 99% OOS detection.
2018: Bossa Nova deployed robots to 50 Walmart stores for shelf inventory auditing; CMU and Microsoft Research published analyses of both technical advances and fundamental limitations in retail computer vision; survey data showed only 4% of retailers with extensive AI adoption but 33% experimenting.
2019: Bossa Nova scaled to 350 Walmart stores (7x growth); Simbe Robotics piloted Tally robots at Giant Eagle (35,000 products) and expanded to a dozen other retailers; Trax entered Japan and Asia-Pacific markets; Sephora deployed video-based customer analytics achieving >95% accuracy; academic research activity surged (113+ papers on shelf recognition by 2019); legal/privacy concerns about customer tracking emerged as an adoption barrier.
2020: Bossa Nova announced expansion to 650 Walmart stores (2x growth) in February, but Walmart cancelled the programme by November after 500-store trial, citing simpler human alternatives and shopper concerns; Trax launched Retail Watch in Americas with major grocery retailer deployments and Deloitte partnerships; COVID-19 drove inventory distortions ($1.8T globally) and spotlighted need for real-time inventory solutions; academic research advanced algorithmic solutions for inventory anomaly detection.
2021: Bossa Nova shut down European operations (January) after parent company funding halted, signaling fundamental viability challenges for robotics-only deployment models; Trax secured $640M Series E (April) from SoftBank Vision Fund 2, BlackRock, OMERS and Sony IGV, validating SaaS-based camera-and-ML market; industry sentiment showed retailers accelerating automation post-COVID despite visible robotics failures, masking a shift from robots to software-based shelf monitoring.
2022-H1: Walmart's Sam's Club announced chain-wide rollout of Inventory Scan robots (~600 stores, January) using Brain Corp's BrainOS platform, signaling large-scale automation commitment; Scandit launched ShelfView computer vision platform (January) integrating with BrainOS; Trax acquired Planorama (February) to consolidate image recognition and expand EMEA footprint, indicating ecosystem maturation around camera-based shelf monitoring despite Bossa Nova's exit.
2022-H2: Badger Technologies scaled autonomous robots to 25-store building supply chains managing 33K SKUs (September); Regami Solutions deployed AI shelf monitoring to national retail chains reducing out-of-stock incidents (December); Trax expanded internationally through Retail-Scan partnership in India (September) and Storecheck partnership in Mexico (December), both targeting FMCG markets; market-wide UK survey (November) revealed 52% of retailers still report 20% of inventory unavailable for sale, signaling persistent adoption barriers despite successful large-scale deployments and proven technical performance.
2023-H1: Scandit case studies documented large-scale retail deployments (400+ stores, 50% time savings in price verification, 20% improvement in on-shelf availability); Trax released MuleSoft Certified Connector integrating shelf analytics with Salesforce Consumer Goods Cloud, signaling ecosystem maturity; independent ECR Retail Loss study across 7 European retailers found 60% of SKUs affected by inventory inaccuracy but showed 4-8% sales lift from correction, quantifying ROI; industry survey (RIS/CGT) confirmed inventory planning as top analytics priority (42% of retailers) yet 52% cite budget constraints and 48% lack skilled staff, indicating continued mid-market and broad-retail adoption barriers despite strong business case.
2023-H2: MIS Quarterly peer-reviewed field experiment with FMCG manufacturer showed AI-powered shelf monitoring significantly improves sales with low marginal costs; Trax expanded to serving 1.5M stores monthly with documented 8-16% revenue increases and 90% time savings; Honeywell survey of 1,000 retail directors showed 58% planned AI/ML/computer vision adoption within a year; Manhattan Associates data revealed inventory visibility declining to 70% accuracy (down from 74% in 2022); Carrefour deployed Adapta Robotics' ERIS robot for automated price-label scanning and shelf stock analysis in Romania; major retailers acknowledged critical adoption barriers including staff resistance to AI, difficult business case-building, and C-suite skepticism.
2024-Q1: Trax launched signal-based merchandising in US markets with named CPG partners (BlueTriton, Trü Frü, Dude Wipes), confirming continued commercial expansion; ECR Retail Loss research quantified inventory accuracy impact at 4-11% sales lift and €10M per 1% improvement; Grand View Research projected in-store analytics market growth to $16.51B by 2030 (CAGR 21.8%); IBM consumer survey showed 55-59% interest in AI retail applications but persistent satisfaction gaps with current deployments, signaling implementation challenges amid strong business case evidence.
2024-Q2: Focal Systems deployed AI shelf monitoring at Morrisons and other UK retailers achieving 2-3% sales lift from improved availability; ParallelDots ShelfWatch gained traction with 84% accuracy and documented compliance improvements; Manhattan Associates 2024 survey revealed only 68% of retailers with accurate inventory view (4% achieving 100%), indicating persistent adoption barriers; ABI Research market data confirmed growing installed base of inventory scanning robots with expanded vendor ecosystem including Simbe, Badger, and Brain Corp; Walmart terminated Bossa Nova robot expansion to 1,000 stores after five-year partnership, revealing structural deployment challenges despite technical viability.
2024-Q3: Market forecasts showed automated shelf monitoring expanding to $1.5B by 2025 with 15% CAGR through 2033; industry surveys indicated 79% of retail/CPG companies actively implementing or experimenting with AI computer vision; 60% of large retailers deployed AI video analytics for in-store security, signaling mainstream adoption of vision technology across retail operations; ongoing market consolidation and vendor ecosystem maturation continued despite persistent mid-market adoption barriers.
2024-Q4: Google Cloud and Infilect partnership delivered cloud-integrated shelf analytics across 20+ countries (December), signaling ecosystem maturity with major infrastructure vendors; Trax released Dynamic Merchandising combining computer vision with 1.4M crowd-sourced workforce, covering 99% of US retailers (November); EU regulatory drivers (Farm-to-Fork) drove 68% of European grocers to plan shelf monitoring by 2025; critical implementation perspective revealed: AI-powered alerts (e.g., spill detection) failed when not integrated into store workflows, indicating deployment barriers remain organizational rather than technical.
2025-Q1: Continued momentum in multi-vendor robotics adoption: Harmons grocer deployed Simbe Tally across 17 stores post-pilot with 20% out-of-stock reduction; Kroger piloted 70 stores combining Simbe and Badger robots validating multi-vendor approach; SaaS platforms expanded globally with HumbleBee AI integration reaching 600+ customers across 500+ retail locations with 90%+ detection accuracy and 80% audit-time savings. Adoption signals accelerated: Deloitte survey showed 60% of retail buyers report AI-enabled inventory management improvements in 2024; manufacturing sector shows 73% testing or using AI, though government initiatives target expanding AI in distribution from <3% to 30% within three years. Ecosystem consolidation continued with Google Cloud-Infilect partnership and Trax Dynamic Merchandising at scale. Persistent tensions remained: technology capability advanced, but organizational integration barriers (workflow adoption, staff buy-in) continued limiting broad adoption outside enterprise tier.
2025-Q2: Market research confirmed sustained expansion: global in-store analytics market forecast to reach $22.34B by 2032 at 21.4% CAGR, with inventory management representing 25% of revenue share. Production-scale deployments accelerated among Fortune 500: Target expanded AI-driven Inventory Ledger to 40%+ of product assortment; Walmart deployed regional demand-prediction algorithms for climate-based inventory repositioning; Home Depot continued Sidekick app rollout. Negative signal emerged with publication of HBR case study documenting Walmart's termination of Bossa Nova robot partnership across 500+ stores, revealing structural deployment barriers (integration cost, labor economics, customer acceptance) despite technical viability. IHL Group analysis quantified high failure rates (80% of retail AI projects fail, 30% beyond pilot phase per Gartner), highlighting persistent implementation challenges. Technology maturity advanced: Trax released Next Best Action, AR-powered on-device recognition, and faster SKU recognition (under 2 weeks), signaling continued product evolution. Customer behavior analytics demonstrated deployment success via peer-reviewed research: 99 beacon deployment across 4,800 sqm real retail environment successfully tracked customer movements and purchase patterns using LSTM models. Practice remained at good-practice tier: enterprise deployments at scale, clear ROI evidence, but organizational integration barriers continued preventing broad mid-market and general retail adoption.
2025-Q3: Vendor ecosystem matured with new service-based models addressing capital cost barriers: Brain Corp and Driveline launched ShelfOptix as fully managed robot service targeting major retailers (Southeastern Grocers pilot); Simbe pivoted to SaaS analytics with Simbe for Merchants (realograms and multi-store planogram dashboards) deployed across hundreds of stores. Market acceleration continued: robotic shelf monitoring forecast $1.2B (2024) to $7.8B (2033) at 23.4% CAGR with North America at 38% market share. Planogram compliance market projected 33.2% CAGR (2023–2032) to USD 105B by 2032. Mid-market robotics diversified: Kroger piloted 70 stores with both Simbe and Badger robots, validating multi-vendor approaches. Camera and edge AI integration standardized with advanced specifications for autonomous shelf monitoring. Deployment barriers shifted from ownership economics to organizational integration: service models addressed capital concerns but labor economics, workflow integration, and staff adoption remained structural obstacles to broad retail adoption.
2025-Q4: Market demand accelerated: IHL Group research quantified 67% of major U.S. retailers facing shelf accuracy challenges (lost sales, brand erosion) with robotics ranked #1 solution and 72% of retailers ready to deploy; $1.73T global inventory distortion validated urgency. Walmart's decade-long autonomous shelf-scanning robot program demonstrated technical maturity at scale (RGB/depth cameras, CNNs feeding real-time data into inventory systems), though October termination of earlier Bossa Nova expansion to 1,000 stores after ~5-year trial revealed persistent robotics economics challenges. Service-model deployment momentum accelerated with ShelfOptix and Simbe for Merchants deployments deepening. Market forecasts firmed: robotics shelf monitoring at $7.8B by 2033 (23.4% CAGR), planogram compliance at USD 105B by 2032 (33.2% CAGR). Adoption remained stratified—Fortune 500 production deployments advancing, mid-market pilot phase accelerating, broad retail (70%+) still manual—signaling the practice's bottleneck had definitively shifted from technology capability to organizational integration and workflow adoption.
2026-Jan: New robotics vendors entered market (PAL Robotics StockBot, Cypher Robotics Captis) with strong early case studies (Leon's 300-store deployment achieving 99.8% accuracy and 90% cycle-time reduction). Simbe released Tally 4.0 with enhanced capabilities. Market analysis firmed: digital shelf analytics (online retail monitoring) projected at USD 2–5B by 2026 with 7–17% CAGR. Physical in-store practice remained at good-practice tier with enterprise deployments advancing and mid-market accelerating, but organizational integration and labor economics continued limiting broad adoption.
2026-Feb: Ecosystem consolidation accelerated: Gemspring Capital's acquisition and merger of FORM with Trax Retail's Image Recognition business created unified platform serving 750+ customers including 30 of top 50 CPG companies (February). Autonomous AI agents for inventory and shrink reduction expanded beyond shelf scanning to pricing and markdown automation (AgileSoftLabs, Trigo). IHL Group quantified continued adoption momentum at 23% AI CAGR with 15% of retail IT budgets directed to AI; independent analysis identified "pilot fatigue" and backwards automation (applying AI to un-optimized workflows) as critical implementation barriers. Practice remained at good-practice tier: technology capability and ROI well-validated, but organizational integration (workflow redesign, staff adoption) remained structural limiting factors for broad adoption beyond Fortune 500.
2026-Mar/Apr: Market scale and ROI validation deepened: Flame Analytics industry report sized global AI retail analytics market at $15.3B (36.6% CAGR) with 90% retailer investment acceleration; Kroger achieved 12% conversion lift from heatmap-based store redesign; Mondelez improved display compliance from 40% to 70% in 40 days using Trax image recognition and AI-driven corrective actions; Australian retail cases quantified recurring revenue impact ($5,000/month per location from layout optimization). Simbe Tally deployment scale confirmed at ~1,000 stores across 3 continents, 5 industries with documented 60% out-of-stock reduction, 90% pricing error reduction, 50% order fulfillment acceleration. IHL Group analyst research quantified adoption stratification: sales leaders 482% more likely early tech adopters, with Boggi Milano, Costco, AWG SmartMeals, and Amazon as named case studies. Independent practitioners documented persistent implementation barriers — measurement gaps (dwell time as descriptive rather than diagnostic), integration complexity, and privacy risks — alongside high-ROI use cases in queue management, staffing optimisation, and planogram testing. Technology proven at scale with ROI clear across deployment types, but workflow integration and staff adoption barriers remain the gating factor for broad adoption beyond Fortune 500.
2026-May: New deployment momentum confirmed: Ahold Delhaize's Albert (350 stores, Czech Republic) scaled Brain Corp shelf-scanning robots achieving 90%+ live-operation accuracy; SpartanNash expanded Tally pilot to a 15-store grocery rollout; SymphonyAI CINDE reported production scale at 1,000+ locations with 91% labor reduction and 2-5% sales lift; Simbe Index launched aggregating 45 billion shelf photos globally; Decathlon USA entered the segment. A regulatory milestone arrived with UL Solutions issuing the first UL 3300 safety certification for a public-space robot (Simbe Tally), with OSHA recognition removing a key regulatory barrier to large-scale retail automation adoption. AI inventory management market projections firmed at $5.85B (2026) growing to $17.42B (2031) at 24.4% CAGR, with Spar Austria validating a 6-store Tally rollout after a 5-month pilot confirming mid-market adoption momentum. Critical implementation analysis documented that detection accuracy is high but 75% of issues go unresolved within 48 hours without dedicated response teams, identifying workflow coordination — not technology — as the residual constraint.
2026-Jun: Starbucks abandoned NomadGo AI inventory (11K stores, 9-month rollout) after critical accuracy failures — mislabeling milk types, missed items, declining accuracy — with post-mortems attributing 97% of root cause to governance and change management failures rather than technology immaturity, providing the cycle's clearest signal that organizational readiness gates adoption. Contrasting signals reinforced the bifurcation: B&R Stores deployed Simbe Tally with 99% shopper acceptance; Kroger spotted expanding multi-vendor robot pilots nationally; Coresight Research survey (200 VPs) found 97% of retailers deployed or planning store intelligence, with 60% actively scaling. Practitioner confirmation arrived that "CV has largely solved the recognition problem at shelf" — with a 7-retailer RCT (1M SKUs, 100 stores) quantifying 4-8% sales lift from accuracy corrections and ML models achieving 60-81% accuracy predicting at-risk inventory records — while data orchestration and fragmentation across disconnected ERP/POS/WMS systems are identified as the binding constraints, with 95% of retail AI pilots producing zero measurable ROI and 85% of failures traced to data quality rather than algorithmic gaps.