Nike’s Chief Technology Officer has welcomed the beta release of NikeAI – a feature that the firm has rolled out to all Nike App iOS users in the US.
Nike CTO Dr Muge Erdirik Dogan said the generative AI application lets users “ask [NikeAI] for running shoes for a race, gear for your team, or products in your favorite color, size, even ones not in our standard filters."
"Most teams treat AI like it’s just software with more data. That mindset will lead to failure – quiet, expensive, and often irreversible." - Nike's Bikram Barman.
The former Amazon executive who joined the sports giant in November 2023, wrote on LinkedIn: “Consumer behavior is changing from searching by keywords to sharing problems to be solved such as ‘I want to run a 5k’.
"Fine-tuned by domain experts"
She added: “NikeAI understands our consumer's shopping intent, built by using best in class foundational models and fine-tuned by domain experts who understand Nike’s DNA. It’s simple, intuitive, and built for real moments in real lives. The early signs,” said Nike’s CTO, “are exciting.
Like all generative AI applications hitting the public, this was a multi-function effort, she emphasised, or “a great team collaboration across Global Tech, Nike Direct Digital Commerce, Marketing, and Legal.”
NikeAI: Team "worked tirelessly across the matrix"
Jason Loveland, VP of AI added that his team “worked tirelessly across the matrix to bring NikeAI to life. The dedication, collaboration, and innovation it took to launch Nike’s personalized shopping assistant responsibly is nothing short of impressive. The lessons learned and the body of knowledge you’ve built will shape the future of AI at Nike and beyond.”
Away from the consumer-facing side, Nike's TechOps and Data teams partnered to deploy two AI-powered tools that "quietly reimagine how we manage tech support", CTO Dogan said.
These "now handle thousands of repetitive tasks – from job failure alerts to common IT questions, freeing our teams to focus on what matters most," she said.
The two tools are:
AgentAutosys, which "automatically detects and resolves job failures in our core systems, quietly handling thousands of alerts and reducing operational noise."
Genius Results, which is a "self-service assistant trained on over 1M past incidents and 6,000 help articles, empowering employees to fix issues instantly, in their own language, without filing a ticket."
"What leaders need to recalibrate..."
Bikram Barman, a senior director at Nike India Technology Center, noted separately that “AI Development Isn’t Software Development. It’s a Different Game Entirely. Most teams treat AI like it’s just software with more data. That mindset will lead to failure – quiet, expensive, and often irreversible.
He wrote on LinkedIn: "Here’s what makes AI development fundamentally different, and what leaders need to recalibrate: "The Life Cycle is Iterative, Not Linear: Traditional software follows a predictable plan → build → test → release cycle.
"AI? It thrives in hypothesize → experiment → validate → deploy → monitor → retrain loops. You’re not building a feature—you’re shaping a living system that evolves with your data."
He added that the “new failure modes to watch for included:
- "Model Decay & Data Drift – Your model loses relevance over time.
- "Explainability Pressure – Stakeholders want to know “why,” not just “what.”
- "Runaway Training Costs – GPU bills can kill ROI if not governed.
- "Compliance Risks – Ethics, fairness, and privacy aren’t optional.
- "Misaligned Teams – You need scientists, MLOps, and product minds working as one.
Nike's Barman added: "Organisations need to track: Model accuracy and fairness; experiments/week; cost per 1000 inferences; drift rate & retraining frequency; business lift vs baseline..."
He described this as a shift in thinking from deterministic delivery → to dynamic systems thinking and from “code-first → to data, iteration, and outcome-first” adding that “if your roadmap includes AI, don’t retrofit your SDLC - [Software Development Life Cycle] rethink it entirely."
How are your AI application builds going? What lessons have you learned? How are you optimising for cost efficiency and accuracy? We'd love to share your thoughts, experiences, and case studies with technical detail. Get in touch with "Real world AI application" in the subject bar.
The link has been copied!
