Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Retailers are using AI to improve demand and inventory decisions, optimize operations, automate warehouse work, and make shopping more relevant and convenient. Adoption is widespread, but many organizations are still piloting rather than scaling these systems. Open-source AI is attracting interest for its potential to give retailers more control over their data and reduce dependence on a single vendor—but it brings evaluation, security, integration, and governance work of its own.
How widely are retailers adopting AI?
Two 2025 surveys point to broad interest, but they describe different groups and do not establish that every adopter has deployed AI across its operations. The National Retail Federation (NRF) Center for Digital Risk & Innovation surveyed 56 U.S. retail AI leaders in summer 2025. NVIDIA’s 2025 survey described nine in ten retail and consumer packaged goods (CPG) organizations as adopting or piloting AI. Because the latter figure combines adoption with piloting, it should not be read as a measure of scaled production use.
The distinction matters: a promising trial does not by itself show that a system is accurate enough, integrated with daily operations, or delivering sustained business value. Retailers are applying AI across both back-office work and customer-facing services, but readiness varies by use case, data quality, and organizational capability.
How is AI changing retail supply chains?
Supply-chain AI can support decisions and automate tasks across planning, inventory, fulfillment, and warehouse operations. Its value depends on whether recommendations are reliable and can be acted on within existing processes—not simply on whether a retailer has a model or pilot.
#1 Best Overall
Demand forecasting and inventory decisions
AI can help retailers interpret sales and operational data to inform demand and inventory planning. Better-informed decisions may help balance availability against excess stock, but results depend on data quality and integration with planning workflows. A forecast that arrives too late, omits important context, or is not trusted by planners will have limited operational value.
Process optimization and warehouse automation
AI can help identify opportunities to optimize supply-chain processes, while automation can assist with warehouse tasks. Gartner reported that top-performing supply-chain organizations use AI to optimize processes at more than twice the rate of low-performing peers. That is an association between performance groups and AI use, not proof that AI alone caused the performance difference.
Physical AI
Physical AI applies AI to systems that interact with the physical environment, including automation in operational settings. In retail supply chains, it can be considered alongside warehouse automation and process optimization. Deployment decisions should account for how the system fits real workflows and what happens when its output is uncertain or an exception occurs.
The execution gap
The opportunity sits alongside a strategy gap: Gartner’s 2025 survey found that only 23% of surveyed supply-chain organizations had a formal AI strategy. A retailer can therefore have promising experiments without a coordinated approach to priorities, data, accountability, and scaling. The practical lesson is to connect a defined operational problem to measurable results and clear ownership before expanding a pilot.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteHow is AI improving retail customer experience?
Customer-facing AI aims to make it easier for shoppers to find relevant products, compare options, get service, and move between shopping channels. The experience only improves when recommendations and answers are useful, current, and consistent with what the retailer can actually sell and fulfill.
Personalization and product discovery
Retailers can use AI to tailor product discovery and customer interactions. In NRF’s 2025 findings, customer personalization was among the applications with the strongest reported returns: 48% of respondents identified it as such. This is a survey finding, not a guarantee that personalization will produce the same return for every retailer.
Unified commerce
Unified commerce connects the customer experience across channels, so shoppers can encounter more consistent information and service as they move between them. Salesforce reported in 2025 that 88% of retailers surveyed said unified commerce would significantly affect their goals. AI can contribute to such experiences, but it cannot compensate for disconnected data or inconsistent product, price, and availability information.
Shopping assistants and agentic service
Shopping assistants can help customers discover products or get answers; agentic service refers to AI systems intended to take actions, not just generate responses. Salesforce reported in 2025 that 75% of retailers surveyed expected AI agents to be essential by 2026. That is a forecast of retailer expectations, not evidence that agents had already become essential or that they can safely handle every service task.
For customer-facing systems, retailers need to define which questions or actions the AI may handle, what information it can use, and when a person should take over. Errors about a product, price, availability, or service policy can undermine the experience the tool is meant to improve.
Rank #4
What retailers should measure
Choose measures that match the use case rather than treating AI usage as a result in itself. For example, a product-discovery pilot could be assessed against relevant customer and business outcomes, while an assistant should also be evaluated for answer quality and escalation behavior. The exact measures should reflect the retailer’s goals and available data.
Why are retailers considering open-source AI?
Open-source AI can appeal to retailers that want to use proprietary business data with more deployment choice, avoid dependence on one provider, or benefit from community innovation. NVIDIA’s 2026 article describes this appeal as a way to leverage proprietary data, avoid vendor lock-in, and benefit from open-source community innovation. Those are potential advantages, not automatic outcomes of choosing an open model.
Open-source AI is gaining broader enterprise interest, but the available evidence does not establish that it is already the dominant choice among retailers. McKinsey identified Meta Llama and Google Gemma among the most commonly used enterprise open-source AI tools as of January 2025. The same McKinsey research reported that 81% of developers highly valued open-source AI experience. That developer figure indicates interest in skills; it is not a count of retail deployments.
What open source can—and cannot—solve
- Data and deployment control: An open-source approach may give an organization more choices about how a model is deployed and adapted to its data. The retailer still has to protect sensitive information and establish appropriate access controls.
- Vendor dependence: Open-source components can reduce reliance on a single model provider, but they do not eliminate dependence on hosting, integration, support, or other vendors.
- Evaluation and governance: Retailers remain responsible for testing model behavior, defining acceptable uses, monitoring results, and deciding who is accountable when the system fails.
- Integration and skills: Greater flexibility can also require more internal expertise to deploy, maintain, and connect the system to retail data and workflows.
“Open source” is not a shortcut around responsible deployment. A retailer should assess the particular model and its terms, security needs, operating requirements, and performance on the intended task rather than choosing based on the label alone.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What are the main risks and adoption barriers?
Survey findings and industry analysis identify a set of recurring challenges: explainability, data quality and integration, skills shortages, governance, customer trust, and immature strategy. Their importance varies by use case, but each can prevent a technically successful pilot from becoming a dependable service.
- Data quality and integration: Incomplete or disconnected information can produce unreliable forecasts, recommendations, or customer answers.
- Explainability: If staff cannot understand or challenge a consequential recommendation, they may be unable to use it appropriately or identify when it is wrong.
- Governance: Unclear rules for access, review, ownership, and escalation make it harder to manage errors and changing model behavior.
- Skills and operating capacity: Teams need the capability to select, evaluate, integrate, and maintain AI systems—not just run an initial demonstration.
- Customer trust: A confident but incorrect response, or an experience that does not match actual price or availability, can erode trust.
- Strategy maturity: Disconnected pilots can consume effort without building toward a measurable operational or customer outcome.
How should a retailer choose an AI approach?
Compare options against the business problem and the retailer’s ability to operate the system. A model that performs well in a demonstration may still be a poor choice if required data is unavailable, integration is burdensome, or the retailer cannot explain and govern its use.
| Decision factor | Questions to ask |
|---|---|
| Business outcome | What specific supply-chain or customer problem should improve, and how will the retailer measure that improvement? |
| Data readiness | Are the necessary data accurate, current, accessible, and usable in the workflow? |
| Integration effort | What systems and teams must connect to the AI, and who will maintain those connections? |
| Explainability and governance | Can staff review outputs, identify uncertainty, and escalate decisions under clear rules? |
| Deployment control | Where will the system run, what data will it handle, and what level of operational control does the retailer need? |
| Vendor dependence | Which providers or components would the retailer rely on, and how difficult would it be to change them? |
| Time to value | How soon can a bounded pilot test the intended outcome using real operational conditions? |
There is no universally best choice between open-source and proprietary AI. The right comparison is between viable approaches for a defined task, including their data fit, integration burden, explainability, deployment control, vendor dependence, and measurable time to value.
Quick Recap
How can retailers roll out AI responsibly?
- Select one measurable use case. Start with a specific pain point in planning, inventory, warehouse operations, product discovery, or service. Define the desired outcome and the current baseline before choosing a system.
- Check data and workflow readiness. Confirm that the needed information is usable and that the people responsible for acting on AI outputs are part of the design.
- Set governance and escalation rules. Decide what the AI is permitted to do, how outputs will be reviewed, how uncertainty or errors will be handled, and who owns the system.
- Run a bounded pilot. Test the system in a limited setting against the agreed measures. For customer-facing use, assess the quality of responses and the path to human support as well as business outcomes.
- Scale only when results justify it. Expand when operational or customer metrics improve under real conditions and the retailer can support the integration, monitoring, and governance required for broader use.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
