How Nike uses AI: direct-to-consumer data and generative product creation
From demand sensing for its own channels to AI-generated footwear concepts and an AI shopping beta, what Nike has built, what it has disclosed and what changed under new leadership.
KEY TAKEAWAYS Summary by the editors
- Nike acquired the demand-sensing company Celect in 2019 to make hyper-local demand predictions for inventory across its direct-to-consumer channels.
- For its A.I.R. project, revealed in April 2024, Nike used generative AI to create hundreds of concept images for 13 athletes, then refined them with designers, computational design and 3D printing.
- Nike's VP of NXT said generating creative starting points with AI took seconds rather than months, but Nike has not published cost, speed-to-market or sales figures for AI-designed products.
- In 2025 Nike launched a NikeAI beta for personalised product guidance, while its VP of AI left and the company cut a limited number of technology roles.
- Nike has not disclosed financial results attributable to AI, so its programme is best read as capability building whose payoff is not yet public.
Nike uses AI in two main areas: data-driven demand sensing and personalisation for its direct-to-consumer business, and generative AI in early-stage product creation. Its best-documented steps are the 2019 acquisition of the analytics firm Celect, the generative A.I.R. footwear concepts shown in 2024 and a NikeAI shopping beta in 2025. Nike has not published financial results attributable to AI.
What has Nike built?
The public record shows three strands:
- Demand sensing: in August 2019 Nike acquired Celect, a Boston-based company whose cloud platform uses data science to make hyper-local demand predictions and optimise inventory across channels.
- Generative product creation: in April 2024 Nike revealed A.I.R. (Athlete Imagined Revolution), prototype footwear for 13 athletes across track, football, basketball and tennis, developed with AI-generated imagery, computational design and 3D printing.
- AI-assisted shopping: in 2025 Nike launched a NikeAI beta that gives personalised product guidance, according to Digital Commerce 360, and in South Korea it worked with Naver's HyperCLOVA X model on a chatbot-style ad format.
Jason Loveland, Nike's VP of artificial intelligence until July 2025, listed among his achievements the integration of Celect and what he called the first fine-tuned generative AI model to design Nike products, shown at the Paris Olympics and Air Max Day, Retail Dive reported.
Why did Nike invest in AI?
The Celect deal came as Nike was building its direct business. Chief operating officer Eric Sprunk said at the time that the acquisition greatly accelerated Nike's digital advantage by adding a platform developed by data scientists. Selling directly creates first-party data on members and demand by location, and predicting that demand precisely is what allows stock to be placed where it sells.
In product creation the motive is speed and range of exploration. Roger Chen, Nike's VP of NXT, said AI-generated starting points that used to take months could now be created in seconds. For a company that designs a large number of styles each season, faster concept exploration shortens the most open-ended part of development.
Celect's chief executive John Andrews said at the time that the company would add its capabilities to the data and analytics foundation Nike had been building, and Celect's co-founders, MIT professors, were to consult for Nike. The acquisition was therefore a talent and intellectual property purchase as much as a software one.
How does it work (data, models, process)?
Nike has described the A.I.R. process step by step. It began with listening sessions with each athlete. Designers then used generative AI to produce hundreds of images based on those sessions, which served as inspiration rather than final designs. Designers sketched from the outputs, and computational design was used, for example, to reinforce traction patterns in high-wear areas. Prototypes were 3D printed at Nike's Concept Creation Center, and digital simulation tested support, containment and durability before physical prototypes were made.
For demand sensing, Celect's cloud platform, as described at acquisition, uses data science to make hyper-local demand predictions that help optimise inventory across channels. Nike has not published how the platform was integrated into its planning systems or which decisions it automates.
| Year | Milestone | Source |
|---|---|---|
| 2019 | Acquires Celect for predictive analytics and demand sensing (August) | FashionNetwork |
| 2024 | Reveals A.I.R. prototypes for 13 athletes created with generative AI and 3D printing (April) | Nike |
| 2024 | Elliott Hill becomes CEO (October) | Retail Dive |
| 2025 | Limited technology layoffs (May); VP of AI Jason Loveland leaves (July) | Retail Dive |
| 2025 | NikeAI beta launched for personalised product guidance | Digital Commerce 360 |
What results has Nike reported?
Nike has not disclosed figures linking AI to revenue, margin, inventory or development cost. The only quantitative statement on A.I.R. concerns time: concept starting points in seconds instead of months, a qualitative comparison by a Nike executive rather than a measured result. The A.I.R. shoes were prototypes, not products sold at scale. For the NikeAI beta, no adoption or conversion data has been published.
Nike returned to year-on-year revenue growth in its fiscal first quarter of 2026, Digital Commerce 360 noted, but nothing in Nike's statements attributes this to AI.
What are the limits and open questions?
The following is editorial analysis. Leadership priorities shifted. Retail Dive reported in July 2025 that chief executive Elliott Hill, who arrived in October 2024, was prioritising other initiatives over AI and virtual efforts, that Nike shut down its virtual products company RTFKT in early 2025 and that its VP of virtual studios had also left. That does not mean AI work stopped, but it shows that AI programmes depend on sponsorship at the top.
Generative concepts raise open questions about intellectual property when models are trained or fine-tuned on design archives, and about how much they change the slower stages of development, such as materials, testing and factory readiness. Demand sensing raises a different question: a model built for direct channels does not automatically cover demand that flows through wholesale partners, whose sell-through data Nike does not own.
What can other fashion companies learn?
- Use generative AI at the front of the design process, where exploration is cheap and humans keep editorial control.
- Link generative concepts to engineering tools such as computational design and simulation, or they remain mood boards.
- Treat acquired analytics capability as an integration project; the value lies in connecting it to planning decisions.
- Secure executive sponsorship and a clear business metric, because AI teams are exposed when strategy changes.
- Pilot customer-facing assistants as betas and measure them before scaling.
Frequently asked questions
How does Nike use AI in design?
In the A.I.R. project, Nike designers used generative AI to create hundreds of concept images based on athlete interviews, then developed them with sketching, computational design, 3D printing and simulation. The AI provided starting points; designers made the decisions.
What is Celect and why did Nike buy it?
Celect was a Boston-based retail predictive analytics and demand-sensing company. Nike acquired it in August 2019 to predict demand at a local level and optimise inventory across its direct-to-consumer channels.
What is NikeAI?
NikeAI is a beta feature launched in 2025 that gives customers personalised product guidance, for example on running shoes for a race, team gear or colour and size preferences, according to Digital Commerce 360.
Has AI improved Nike's results?
Nike has not published any figures attributing revenue, margin or inventory improvements to AI. Its public statements describe capabilities and time savings in concept creation, not financial outcomes.
One edition every weekday morning. Read in five minutes. Free for industry professionals.
