The quick rise of artificial intelligence-powered tools has reshaped retailers’ process of selecting technology partners for anything from marketing to supply chain to merchandising.

Surveys suggest customers want to use AI for shopping and to see AI tools from retailers. In a CI&T survey conducted in 2025, 58% of 1,040 U.S. consumers said retailers should use AI to improve the shopping experience, and almost 75% said they were already using AI tools at least occasionally in their path to purchase. In a separate survey from Gartner last March, 56% of millennials said they would be willing to let AI handle or assist with some of their shopping tasks.

Bringing AI into a retailer’s business could mean using models they already have access to through partnerships with OpenAI, Google or Microsoft; implementing AI features from existing software vendors; or working with a newer AI startup. Often, retailers will do a mix of all of these, depending on the desired use case.

Given how quickly AI has entered the collective consciousness, there has been more pressure on retailers to deploy AI features more quickly and experiment to see what works. This is all while still doing due diligence, as AI can be complex or costly.

The challenge is that the best solution can quickly shift, considering how rapidly technology is evolving, said Jason Goldberg, chief commerce strategy officer at Publicis Groupe. For example, he said that, historically, retailers may have spent 18 months searching for and rolling out an e-commerce platform.

“It’s exposing new problems that the old selection criteria and process are not sufficient for,” Goldberg said. “CIOs are spending a lot of time talking to each other about how they’re going to survive the new pace of change, and they’re trying different things, but the jury is still out on what the right approach is.”

Melissa Minkow, global director of retail strategy and insights at IT firm CI&T, said some retailers are questioning whether they have the internal resources to integrate AI into their businesses successfully. For example, she said she knows of several retailers who have questioned using AI for digital merchandise planning because planners fear it will make it harder to draw on their own human knowledge and expertise from working in the field.

“This is really the first time where consumers are more ready for the technology than the retailers,” Minkow said. “Retailers know that their business will benefit from implementing it in most cases, but they have a lot of trepidation around it.”

Nick Anderson, vp of data science and AI at CarMax, said that when selecting which vendors to go with — or whether to develop something in-house — the company starts with the business outcomes it’s trying to influence, like conversion rates or churn.

“Then you look at the process you use when you’re engaging with the customer, or where you’re engaging in financing,” Anderson said. “In the evaluation of that process, then you can see [if there is] a role for AI to play in making that process better, faster or cheaper.”

For example, Anderson said CarMax has worked with a vendor (he declined to name the firm) that allows customers to speak with an AI agent by phone. In that case, executives at the car retailer decided they wouldn’t have a reason to build that in-house and would be able to create a good experience through a vendor.

“Because of the massive advancement, the proliferation of tools out there and the rapid advancement and evolution of those tools, you always need a finger on the pulse of which tools are the best tools to have in the toolkit,” Anderson said.

Cost, however, could play a role in whether a company turns to an in-house solution or not. According to a Gartner survey conducted last year, 55% of marketing technology leaders said they face ongoing pressure from leadership to reduce their spending.

“We’re seeing more retailers look to their current marketing technology vendors to see what generative AI capabilities or tools they offer that they can benefit from, versus bringing in a net-new martech or generative AI tools,” said Greg Carlucci, a senior director analyst for Gartner who works with marketing executives.

Carlucci added that Gartner generally recommends retailers implement several AI marketing features the firm finds the most successful: conversational chat for customers, enhanced search and upselling, and product image attribute tagging. Often, he said, existing technology vendors can handle this.

CarMax generally uses Microsoft Foundry to build AI tools, but Anderson said there may be advantages in turning to specialized AI tools, like for resume screening or contract reviews. For those, he said, it depends on how much repeatable human effort is going on within the organization to warrant additional investments.

“Vendors have potentially more comprehensive solutions with more bells and whistles that come with more costs, and internal solutions can be more nimble, more cost-efficient, more time-efficient to create,” Anderson said. “It’s picking the right tool for the right problem.”

Like CarMax, Lowe’s builds its own AI features through partnerships with tech companies like OpenAI, Google and Nvidia.

“Our focus is on providing the best AI-driven solution, rather than chasing the latest trend, so that we can solve customer and business problems,” Chandhu Nair, svp of stores, data, AI and innovation at Lowe’s, said in an email.

Anderson added that, when turning to third-party tools, having the ability to fine-tune the software’s capabilities is important, given that the company wants to have control over the customer experience.

Minkow said retail executives will often rely on referrals for a tech platform from colleagues at other non-competing retailers or from engineers at the organization who may have experience with it. She said case studies are also important for executives to see. “Beyond just proof of concepts, you need to actually see that they’ve been able to implement it somewhere.”

Nair said Lowe’s focuses on three things in selecting tech partners: simplicity of use, speed to scale, and the ability to solve a meaningful customer or business problem.

“Capabilities matter, but what ultimately counts is whether a solution delivers measurable impact,” Nair said. “If an AI-driven solution doesn’t move the needle on adoption, conversion, efficiency or customer experience, it doesn’t get built or deployed.”