Hey there 👋

There’s a scene inside most large retail organizations right now that would have seemed implausible five years ago. A small team of no more than four or five people gets a brief, a deadline, and access to a few API keys. Three weeks later, they ship something that previously would have occupied a product roadmap for 18 months and required a dedicated engineering team to build.

Nike’s conversational shopping assistant is exactly that story.

Faced with a conversion problem they had diagnosed clearly but solved poorly, a cross-functional squad inside Nike’s digital product org decided to stop waiting for a “big AI transformation” and just build. They used existing infrastructure, a well-designed prompt architecture, and an orchestration layer to wire it all together. 

Let’s get into how it was done. 🚀

How We Built Nike's Personalized AI Shopping Assistant in Just 3 Weeks. With thousands of product SKUs and data from over 170 million loyalty program members, Nike envisioned a hyper-personalized, conversational shopping experience. One that felt like a natural, one-on-one conversation and made customers feel understood and supported by surfacing relevant products in real time and at scale, in an intelligent, dynamic, and non-intrusive way. [Source: Rangle]

AI Mode is Google’s next ads engine, and it already knows how to monetize it. As AI Mode expands, ad formats, reporting, and control are starting to take shape, with Google playing a long game that competitors can’t match. As conversational search gains traction, the bigger question isn’t who has more users, but who can monetize them. [Source: Search Engine Land]

Jeff Bezos reportedly wants $100 billion to buy and transform old manufacturing firms with AI. Jeff Bezos is reportedly seeking $100 billion for a new fund that will be used to buy up companies in major industrial sectors and, ultimately, modernize and automate them with AI, according to sources cited by The Wall Street Journal. [Source: TechCrunch] 

Generative AI improves a wireless vision system that sees through obstructions. With this new technique, a robot could more accurately detect hidden objects or understand an indoor scene using reflected Wi-Fi signals. This new technique builds a partial reconstruction of a hidden object from reflected wireless signals and fills in its missing shape using a specially trained generative AI model. [Source: MIT News] 

In a case study published by Rangle, Nike partnered with the agency to build a generative AI-powered shopping assistant designed to personalize the product discovery experience, and the entire build took three weeks. 

That speed is the real story here. Unpacking how it was possible tells us something important about where conversational commerce is going.

The problem Rangle and Nike set out to solve is one most Direct-to-Consumer (DTC) brands will recognize: mobile shoppers were landing on the footwear category page and leaving without converting.

The standard browse experience, with filters, sorting, and category pages, wasn’t giving them a fast, intuitive path to the right product, and the answer turned out to be a conversation.

Typical architecture for a generative AI shopping assistant of this kind draws on a short, well-established list of components, and understanding that list is useful regardless of which specific tools any given team chooses. 

An LLM API (such as GPT-4o or Claude) handles conversational understanding, context retention across a multi-turn session, and response generation. This is the engine that makes the assistant feel like a dialogue rather than a search box.

An orchestration layer (such as LangChain or LlamaIndex) manages conversation memory, routes user intent to the right downstream system, and enforces the guardrails that keep the assistant on-task. Without this layer, the assistant can’t reliably reference what the user said three turns ago.

A product search integration (such as Algolia or Elasticsearch) translates the conversational context the assistant has gathered, shoe type, use case, fit preference, and budget into structured queries against the product catalogue, returning results ranked by relevance to the conversation rather than keyword match alone.

An existing product catalogue and inventory API ensure the assistant surfaces actual in-stock options rather than ghost recommendations. The ability to plug into infrastructure that already exists is a significant part of why these builds are achievable.

All of these components are available off the shelf, but Rangel and Nike thoughtfully assembled them.

The most important artefact in an early sprint like this is the conversation architecture.

This means defining the assistant’s persona, the qualifying question sequence, the decision logic for when it has enough context to make a recommendation, and the guardrails that keep it focused. Getting this right up front determines everything that follows. Infrastructure integrations can be scaffolded in parallel.

Before any customer sees the assistant, the team needs to simulate the full range of user behaviours.

The shopper who knows exactly what they want, the one who doesn’t know where to start, the one who asks something completely off-topic. Every edge case gets a defined response built into the system prompt. The goal is to eliminate surprises before any real traffic hits the experience.

A phased deployment that exposes a portion of real traffic to the assistant while keeping the remainder on the standard experience allows the team to observe behaviour, identify drop-off points, and iterate on the prompt architecture before any wider rollout.

This is where conversation design and engineering work together: the data tells you where the assistant loses people. The system prompt is where you fix it.

Nike designed a conversation, connected it to the infrastructure they already had, and deployed it. The intelligence in this system lives in the prompt architecture, and that’s a content and strategy problem.

That matters enormously for marketers because it means the bottleneck for building something like this is the clarity of thinking about how your customers make decisions.

There’s also a broader lesson in who leads the conversation design work. The most valuable contributor in a sprint like this isn’t typically an engineer. Instead, it’s the person who understands how customers move from uncertainty to decision, and can structure a conversation that shortens that journey.

That’s a content strategy and UX skill applied to an AI context, and it’s a capability most marketing teams already have, or can develop quickly.

You think 4x faster than you type. So why are you typing prompts?

Wispr Flow turns your voice into ready-to-paste text inside any AI tool. Speak naturally - include "um"s, tangents, half-finished thoughts - and Flow cleans everything up. You get polished, detailed prompts without touching a keyboard.

Developers use Flow to give coding agents the context they actually need. Researchers use it to describe experiments in full detail. Everyone uses it to stop bottlenecking their AI workflows.

89% of messages sent with zero edits. Millions of users worldwide. Available on Mac, Windows, iPhone, and now Android (free and unlimited on Android during launch).

Try Wispr Flow free →

Use this prompt with your preferred LLM to design a conversational shopping assistant architecture for your own product category.

The scenario:

You are the Head of Digital Marketing for a mid-sized direct-to-consumer retail brand. Your mobile conversion rate is underperforming desktop by more than 30%; exit survey data suggests that shoppers are leaving because they can’t quickly identify the right product from your catalogue. You’ve been tasked with designing a conversational shopping assistant that can be built and deployed within four weeks, using existing API infrastructure with no custom model training or months-long engineering project.

The Prompt:

Act as a senior AI product strategist with deep experience in conversational commerce. I need your help designing the full architecture for a conversational shopping assistant for my brand. I’ll share context on our product category, customers, and current tech stack below.

Based on that, help me design the conversation flow, the system prompt structure, the recommended tooling, and the A/B test plan for a four-week build sprint. Be specific, I want something I can hand to my team as a working brief, not a set of general principles.

Current Situation

QUESTIONS

These are the most popular AI tools we use at Rise Up Media. If you're not using them already, they're worth a look.

Full disclosure: some links above are affiliate links. If you sign up, we’ll earn a small commission at no extra cost to you.

This is a real article from New York Magazine. 2026 is weird, man… 😆

What Nike’s sprint makes clear, perhaps more than anything, is that the organizations moving fastest on AI are the ones willing to define the problem precisely and resist the temptation to solve it with a standard transformation programme.

The conversational shopping assistant isn’t a novel concept. Retailers have been trying to build it in various forms for years. What’s new is that the infrastructure required to make it work reliable LLMs, flexible orchestration, and real-time search integration is now accessible to virtually any team. The constraint today is the clarity of thinking that goes into designing the conversation itself.

Understanding how customers make decisions, knowing the questions that reduce uncertainty, and being able to write words that build trust without friction are the work that make a conversational AI experience feel like a brilliant in-store associate rather than an elaborate FAQ.

If there’s one thing to take from this issue, don’t wait for a perfect AI strategy document to emerge from a committee. Find the conversion problem you can name precisely, find the team that can own it, and give them three weeks.

Until next time, keep exploring the horizon. 🌅

Alex Lielacher

P.S. If you want your brand to show up in Google AI Mode, ChatGPT, and Perplexity, reach out to my agency, Rise Up Media. That's what we do.