AI In The Accessories Industry Statistics

PWC reports AI can reduce fraud losses by 50% and more [1]

Juniper Research estimates retailers/brands deploying chatbots could save $8 billion annually by 2022 (and continuing growth) [2]

Gartner estimates that chatbots can reduce customer service costs by 30% [3]

IBM reports that AI can reduce operating costs by up to 30% [4]

McKinsey reports that AI can reduce inventory holding costs by 20% to 50% [5]

McKinsey reports that AI can improve demand forecasting accuracy by 10% to 20% [6]

McKinsey reports retailers can increase supply chain productivity by 20% to 50% using AI [7]

Deloitte reports that AI can cut customer service costs by 30% to 50% [8]

PwC reports that automation and AI can reduce average costs per transaction by 30% to 60% in certain workflows [9]

Capgemini reports that AI-driven recommendations can increase average order value by 10% to 30% [10]

Salesforce reports that personalization can reduce churn by 15% [11]

McKinsey reports that AI pricing can increase profits by 2% to 5% [12]

NielsenIQ reports that smart shelf/AI retail analytics can reduce out-of-stocks by 10% or more (case benchmark) [13]

RFID/AI inventory analytics: GS1 reports that RFID-enabled processes can reduce inventory inaccuracies by up to 50% [14]

IBM reports that AI can detect fraud with up to 95% accuracy in some deployments [15]

McKinsey reports that generative AI can reduce time spent on marketing by 30% to 60% [16]

McKinsey reports genAI can increase marketing productivity by 10% to 30% [17]

Gartner estimates AI can improve customer satisfaction by 10% to 20% [3]

BCG reports that AI and analytics can boost margins by 2% to 3% [18]

Deloitte reports AI adoption can increase labor productivity by 30% to 40% in some operations [19]

Oracle reports that AI can reduce call center handle time by up to 20% [20]

Zendesk reports that chatbots can reduce support costs by 50% for some organizations [21]

Juniper Research estimates chatbots could save over $11 billion annually by 2024 [22]

IBM reports that AI can reduce planning and scheduling errors by 25% to 30% [23]

McKinsey reports that AI-powered personalization can increase revenue by 5% to 15% [24]

Harvard Business Review cites that in retail, recommendation engines can yield a 10% to 30% increase in sales [25]

MIT Sloan Management Review reports companies using AI and analytics improve efficiency by 10% to 20% (reported range) [26]

Accenture reports that retailers could gain $2.8 trillion in value from AI by 2035 (including efficiency) [27]

BCG estimates that genAI can deliver $200B+ value to retail (productivity and marketing) [28]

IHL Group estimates that AI-driven retail inventory optimization can reduce stockouts by 20% to 30% [29]

PwC’s Global Consumer Insights Survey (2024) found 33% of consumers are open to using AI assistants for shopping [30]

Salesforce reports 88% of customers expect an experience tailored to their preferences [31]

Salesforce reports 72% of customers expect companies to understand their unique needs [31]

Accenture found 40% of consumers will choose the retailer that offers a personalized shopping experience [32]

McKinsey reports personalization can reduce acquisition costs and increase revenue (personalized marketing can deliver 5% to 15% increases in revenue) [33]

McKinsey reports AI-driven personalization can increase marketing ROI by 20% or more [24]

Deloitte reports 61% of consumers expect companies to use AI to improve customer service [34]

IBM’s 2024 study found 67% of consumers expect to interact with AI regularly in the future [35]

McKinsey states that personalization can lift sales by 10% or more [36]

KPMG found 39% of consumers are more likely to shop at retailers that recommend products based on their behavior [37]

Capgemini’s World Retail Report 2023 found 58% of shoppers are more likely to shop when retailers offer AI-driven recommendations [38]

A McKinsey study estimates that personalization improves customer experience, with a potential increase in sales by 15% and marketing ROI by 10% to 15% [24]

Salesforce’s “State of the Connected Customer” reports 88% of respondents say the experience a company provides is as important as its products/services [31]

Shopify reports merchants that use AI features see higher conversion (example: Shopify’s AI-powered product recommendations can improve conversion by up to 20% for some merchants) [39]

Amazon reports that improvements in personalization led to significant increases in customer engagement (company has stated that personalization has historically contributed meaningfully to higher conversion) [40]

In a 2023 report, 73% of consumers say they would trust an AI chatbot that provides accurate answers [41]

Pew Research Center (2023) found 79% of US adults say it is likely that AI will be used in everyday life [42]

Twilio’s 2023 customer engagement report found 73% of consumers prefer to interact with businesses using their preferred channel [43]

Bain & Company reports companies that excel at customer experience outperform their peers by up to 80% [44]

Capgemini found 76% of customers want more personalized interactions [45]

PwC found 52% of consumers are willing to share data in exchange for a better experience [46]

McKinsey reports that visual search can improve product discovery and reduce search friction; they cite potential improvement in conversion (industry estimates commonly show 15%+) [47]

Retail sales personalization research: 44% of customers are likely to become repeat purchasers after a personalized offer [48]

Adobe found that 30% of consumers would abandon a website if content is not personalized [49]

McKinsey reports that people spend 8% more time on websites when personalization is used [36]

Meta reports that personalized ads can increase purchase intent; they cite improvements for advertisers (e.g., conversion lift benchmarks) [50]

Google/Think with Google reports that 77% of companies using AI for personalization see improvements in customer satisfaction [51]

Forrester reported that AI personalization can increase revenue by 15% and reduce marketing costs by 10% [52]

Accenture’s 2023 report found 79% of shoppers are more likely to buy from brands that offer personalized recommendations [53]

Microsoft states that their Responsible AI Standard includes 4 pillars [54]

NIST AI Risk Management Framework (AI RMF 1.0) is organized into 5 functions: Govern, Map, Measure, Manage [55]

EU AI Act (2024) sets a framework where fines can be up to €35 million or 7% of global annual turnover, whichever is higher (for certain prohibited practices) [56]

EU AI Act (2024) requires high-risk systems to undergo conformity assessments; the summary states that fines can be up to €15 million or 3% for some obligations [56]

GDPR allows administrative fines up to €20 million or 4% of global annual turnover, whichever is higher, for infringements [57]

CCPA statutory damages can be $100 to $750 per consumer per incident [58]

FTC penalties: FTC may seek civil penalties up to $50,120 per violation for some rules under the Inflation Adjustment Act (example figure) [59]

EU GDPR: data breach notification requirement is “within 72 hours” of becoming aware [60]

NIST AI RMF includes Measurement as a function with categories and subcategories; details are documented in the framework [55]

US White House Executive Order 14110 requires agencies to consider AI safety and security; it sets deadlines such as publishing assessments by specified dates (documented) [61]

UK ICO states organizations must keep records of processing activities (Article 30) and that failure can lead to enforcement; specific thresholds and rules are in guidance [62]

ISO/IEC 42001:2023 defines requirements for an AI management system; it was published in 2023 (data point) [63]

ISO/IEC 23894:2023 provides AI risk management guidance (published 2023) [64]

IEEE 7000 series: IEEE 7001 defines transparency of automated systems; published guidance exists (standard) [65]

OECD AI Principles include 5 principles; it was adopted in 2019 [66]

NIST documents that explainability is part of Trustworthiness; their AI RMF includes Measure/Manage for performance, robustness, etc. [55]

IBM reports bias detection tools can reduce bias; for fairness metrics they report improvements (benchmark) within their case studies [15]

Microsoft states its Responsible AI Standard requires documentation, monitoring, and human review; the policy provides required controls [67]

US National Institute of Standards and Technology notes that AI systems can cause risks including bias and discrimination; NIST highlights those risk types in the framework [55]

European Data Protection Board guidance: DPIAs must be completed before processing when high risk is likely (legal requirement) [68]

NIST AI RMF provides “Govern” function categories including policies, processes, etc. (structured categories) [55]

Global AI in retail market size was valued at $7.9 billion in 2023 and is projected to reach $39.6 billion by 2032, with a CAGR of 20.0% from 2024 to 2032 [69]

In 2024, 83% of retailers say AI is important to their growth strategy [70]

In a 2023 survey, 52% of retail organizations reported they use AI in at least one business function [71]

McKinsey reports AI could add $2.6 trillion to $4.4 trillion annually across retail by 2030 [16]

Gartner predicts that by 2026, 80% of customer service organizations will use AI-enabled capabilities [72]

Oracle reports 84% of organizations expect to leverage AI for customer service within the next 2 years [73]

Insider Intelligence (eMarketer) projects ecommerce sales to grow to $7.6 trillion by 2025, providing a large addressable spend for AI-driven personalization in commerce [74]

McKinsey estimates that AI could automate 45% of work activities in marketing and sales [75]

McKinsey estimates that AI could automate 60% to 70% of routine work in customer operations [76]

Salesforce reports 57% of companies use AI for some personalization or customer interaction [77]

IBM reports that businesses using AI can reduce costs by up to 30% [78]

Gartner predicts that by 2025, chatbots will manage 15% of customer service interactions globally [79]

Gartner predicts that by 2025, 25% of organizations will use AI to enhance the customer experience [3]

Gartner predicts that by 2026, customer service leaders will deploy AI assistants for most customer interactions [3]

The IBM Global AI Adoption Index 2023 found that 35% of surveyed organizations had adopted AI in at least one function [80]

IBM reports that 42% of respondents are implementing AI due to competitive pressure [80]

IBM reports that 19% of respondents had AI models in production [80]

Gartner says by 2024, AI will become the new standard interface for customer service; 70% of customer interactions will start with AI [81]

A 2024 survey by Kore.ai found 82% of enterprises are exploring or already using chatbots/virtual agents [82]

A 2024 Deloitte survey found that 57% of retail executives have deployed AI [83]

A McKinsey survey found 56% of respondents said they have adopted at least one AI use case [76]

McKinsey reports retailers can reduce markdowns by 20% through AI pricing and inventory decisions [84]

Optimizely reports A/B testing improves conversion; for some e-commerce use cases, conversion rate can improve by 10% to 25% [85]

Adobe reports that 49% of marketers say AI has improved content targeting [86]

NIST reports that facial recognition error rates can be as low as (false accept/false reject) in controlled conditions; examples show improvements to under 1% in some controlled settings [87]

NIST’s Face Recognition Vendor Test (FRVT) measures false match rates and false non-match rates; for one scenario, errors can be at ~0.1% or lower depending on threshold/setting [88]

OpenAI reports GPT-4 class models can achieve high performance on many benchmarks; e.g., MMLU score reported as 86.4% for GPT-4 [89]

OpenAI reports that GPT-4o achieved 87.2% on MMMU and 91.6% on MathVista (reported results) [90]

Google Gemini documentation reports tool use and reasoning improvements; for example, Gemini 1.5 Pro supports long context windows up to 1 million tokens [91]

Google DeepMind reports that AlphaFold2 predicts protein structures with high accuracy; typical use is reported with TM-score above 0.7 for many targets in CASP14 (context for AI capability) [92]

Microsoft reports that Azure AI Vision supports optical character recognition (OCR) with high accuracy; example: document intelligence uses models evaluated on standard benchmarks (reported F1/AUC ranges) [93]

NVIDIA reports that its AI-powered recommendation/vision systems can run at low latency; for example, TensorRT supports up to 2x speedups in optimized inference in many deployments [94]

NVIDIA states that TensorRT can improve inference performance by optimizing models; they report “up to” 40% for some cases [95]

Amazon Rekognition documentation indicates it can detect labels (objects) and faces; “FaceIndexer” supports searching millions of faces; example: up to 100 million faces per collection (service limits) [96]

AWS Rekognition face search provides “SearchFaces” returning the best matches; service supports collections up to the limits shown (example: 5 million faces for some configurations) [96]

Google Cloud Vision API supports 1,000 requests per 100 seconds per project by default (quota) [97]

Azure AI Vision service supports 20 transactions per minute per region by default for some tiers; example quota numbers provided in docs [98]

OpenAI’s GPT-3.5 Turbo was trained on 175 billion parameters (commonly reported) [99]

OpenAI’s GPT-4 technical report lists “~1.8 trillion parameters” (reported estimate) [100]

Meta’s Llama 2 paper reports model sizes up to 70B parameters [101]

Meta’s Llama 3 report states Llama 3 model sizes up to 405B parameters [102]

Anthropic’s Claude 3 report states it can handle long contexts up to 200K tokens for Claude 3 Opus (reported) [103]

Cohere’s Command R+ model supports 128K context window (reported) [104]

OpenAI’s Whisper model transcription: Word Error Rate (WER) benchmarks include low WER on LibriSpeech (example ranges under 5% for certain settings) [105]

OpenAI’s Whisper repo reports transcription accuracy benchmarks with WER values [106]

NIST’s Face Recognition Vendor Test includes metrics for “False Match Rate (FMR)” and “False Non-Match Rate (FNMR)” for different demographics; results vary [107]

NIST reports that some face recognition systems show large performance differences across demographic groups; the report provides specific error-rate disparities [107]