Retail and supply chain leaders are being told that agentic artificial intelligence adoption is a straightforward cost-saving decision. The pitch is simple: automate a task, reduce headcount or hours, and watch the savings appear. However, the financial case for AI can be harder to establish than the technology itself, particularly for omnichannel retailers whose initiatives touch stores, e-commerce and fulfillment at the same time. Before a retailer can determine whether an AI initiative will save money, it needs to understand what the existing process costs, what the AI solution will require to implement and operate, and how those costs will change as usage scales across channels, categories and peak periods.

That discipline is increasingly important as retailers move from experimenting with AI to putting it into operational workflows, from personalization and customer service to inventory visibility, fulfillment and returns. A solution can work technically and still fail to deliver the expected financial return if the business case didn’t account for the full cost of implementation, ongoing operation, support, data preparation and scaling.

A Business Case Without a Complete Cost Picture

The more defensible concern is that retail organizations can enter an AI initiative without a sufficiently detailed understanding of what the solution will cost relative to the process it’s intended to improve.

That gap is important because the cost of an AI initiative extends beyond whatever appears on a vendor’s quote. Data preparation, technical architecture, integration with e-commerce, order management, inventory and customer data platforms, implementation, monitoring, training, workflow changes and ongoing management can all affect the economics of a solution. If those factors aren’t included in the original business case, a technically successful pilot in one channel may look very different once it becomes an operational system across the enterprise.

Where the Business Case Breaks Down

For retail organizations, the larger challenge may be even more fundamental: many AI initiatives are not being evaluated through a consistent business case or measurement framework. According to newly released supply chain AI research, 78 percent of supply chain leaders say their organizations are pursuing AI without a formal plan, and 85 percent said AI opportunities are selected without consistent criteria or a formal evaluation process. The research also found that developing or validating the business case was the most frequently identified barrier to AI progress, cited by 64 percent of respondents. Perhaps most telling, only 11 percent of organizations establish a baseline and measurable success criteria before launching an AI initiative, and just 3 percent go back afterward to compare results against the original business case.

Those findings point to a problem that comes before vendor selection: organizations often lack the measurement discipline needed to determine whether an AI investment is delivering what it promised. If a retailer doesn’t define what success looks like before it starts an initiative, it has no benchmark against which to catch a cost overrun after the contract is signed. A pilot can succeed technically, automating a task a human previously performed, and still fail economically once the full cost of implementing and operating the solution is measured against the value of the work it was meant to support. That doesn’t necessarily mean the AI solution was overpriced. It may mean the original comparison didn’t capture the full economics of either option.

Modeling True Cost Before Signing Anything

None of this means retailers should steer clear of AI. It means the buying process needs a step that’s currently missing from most organizations: a true cost model built before a contract is signed, not after.

That model should start with the fully loaded cost of the human process being replaced or augmented, such as a contact center agent, marketing analyst or returns processor, including wages, benefits, training and management time, expressed as a cost per hour of output. But the comparison shouldn’t stop there. Buyers should also document the volume, cycle time, error rate and other performance measures associated with the existing process, such as contact resolution, conversion or returns processing time, so they’re comparing the AI solution against a meaningful baseline.

It should then build a range, not a single number, for the AI alternative based on realistic usage volume and the full lifecycle of the initiative rather than a proof-of-concept scenario. Buyers should ask vendors directly how pricing scales with volume, whether by transaction, customer interaction or another unit; what implementation and integration costs are expected; what internal resources will be required; how ongoing support and monitoring are handled; and whether there’s a cap or a way to throttle usage before costs exceed the assumptions in the business case.

The key question is not simply, “What does the AI tool cost?” It is, “What will it cost us to put this capability into production across every channel and keep it producing measurable value?”

Measurement Cannot Be an Afterthought

The survey data points to where organizations should focus first. Respondents identified execution-related needs, not AI capability itself, as their greatest gaps: preparing data and technical architecture ranked highest at 63 percent, followed by planning and managing implementation at 60 percent, measuring outcomes and expanding what works at 52 percent, and evaluating technology and vendors at 49 percent. That combination is revealing. Organizations are asking for help with the practical work required to make AI function in live retail operations, while the research also shows that relatively few organizations have established the measurement practices needed to determine whether those investments are working.

The retailers that will benefit most from AI will not be the ones that adopt the fastest. They will be the ones that treat every AI investment the way they would treat any other capital decision, such as a new distribution center or a platform migration, with a baseline, a forecast range, and a defined point at which results get checked against what was promised. That discipline, more than any specific tool, is what turns AI from a hopeful bet into a measurable one.

This article was originally published by Total Retail, the sister brand of Women in Retail Leadership Circle. 

Rachelle Butler is the director of Redwood Advisory at Redwood Logistics, North America’s leading open fourth-party logistics (4PL) provider.