Bunnings Rolls Out Gemini Enterprise for CX Shopping Agent Buddy in New Zealand After Doubling Online Conversion in Australia
Bunnings, the leading home improvement retailer in Australia and New Zealand, has launched Buddy, an AI shopping agent built on Google Cloud's Gemini Enterprise for CX, in New Zealand. We explain its Australian results, design and caveats, and what they mean for retailers selling project-based products.
Key Takeaways
- Bunnings, the leading home improvement retailer in Australia and New Zealand, has launched Buddy, an AI shopping agent built on Google Cloud's Gemini Enterprise for CX, on its New Zealand website and app. It is the first New Zealand retailer to do so, following an Australian rollout that began in April 2026
- Buddy starts from a goal such as "I want to build a deck," works out the materials and tools needed, and can find products from a photo or a handwritten shopping list. Bunnings and Google say online conversion more than doubled with Buddy in Australia, but the comparison baseline and specific average order value figures are undisclosed
- For project-based categories where customers buy many items at once, the agent's job is less product search than telling customers what they need. Whether a retailer's own how-to content and staff expertise exist in a form AI can draw on becomes the differentiator
A first for New Zealand retail: Bunnings launches Buddy
Bunnings has become the first New Zealand retailer to launch a shopping agent using Google Cloud's Gemini Enterprise for Customer Experience platform.
www.reseller.co.nzHome improvement giant Bunnings has added its AI shopping assistant, Buddy, to its New Zealand website and app. As Reseller News reported on October 1, 2026, it is the first New Zealand retailer to launch a shopping agent built on Google Cloud's Gemini Enterprise for Customer Experience (CX).
Buddy breaks DIY projects into steps, finds the products needed and returns practical advice. Beyond typed questions, customers can find products by uploading a photo or snapping a handwritten shopping list, and Buddy anticipates what they may need next. Checkout continues through the Bunnings online store as before.
Much of Buddy's guidance draws on Bunnings' library of how-to content. Melissa Haines, general manager of Bunnings New Zealand, put it this way.
This is about using AI in a practical and responsible way to complement the advice our team members provide every day, while giving customers another way to interact with Bunnings when and where it suits them.
The New Zealand launch follows the Australian rollout that began six months earlier. The numbers and design thinking are clearer from the Australian case, so the rest of this article looks at what happened there.
Six months in Australia: from product search to project search
In its April 4, 2026 announcement, Bunnings began progressively rolling out Buddy on its Australian website. It was available to some visitors first, with broader access over the following weeks. Buddy replaced Ask Bunnings AI, the chatbot previously on the site.
The replacement stemmed from the limits of that first-generation chatbot. According to Computer Weekly, Vivek Pradhan, Bunnings' general manager of data and AI, told Google Cloud Next 2026 that Ask Bunnings AI could surface the retailer's 15-year archive of DIY content but made for a clunky experience. Customers had to open multiple browser tabs, recommendations could not be refined, and images could not be processed.
What Buddy set out to do is what Bunnings calls the shift "from product search to project search." Tell Buddy you want to build an outdoor deck, for example, and it recommends decking boards, the underlying structure, measuring tools and power tools, and links to Bunnings' how-to videos.
Images changed things too. Photograph a handwritten shopping list and the items can go straight into the cart. Send a photo of an unfamiliar part, such as a furniture cam lock (a fitting that holds panels together) that has fallen out of a bookshelf, and Buddy identifies it and tells you which aisle of your nearest store stocks it. It also checks local price and availability, connecting online advice to in-store purchases.
The speed of development stands out. According to Google Cloud's customer story, Bunnings worked directly with Google engineers and brought Buddy to market in about six weeks. Groundwork laid earlier with Agent Search (formerly Vertex AI Search) is credited with making that short timeline possible.
The published numbers and how to read them
Google Cloud's customer story lists more than double the online conversion rate and a significant lift in average order value as Buddy's results. Bunnings CIO Genevieve Elliott said much the same at Google Cloud Summit in Sydney in June. According to Reseller News, she reported more than a doubling of online conversion rates and larger baskets, particularly among users who had several turns in the conversation.
These figures should not be taken at face value. The baseline for "double" has not been stated. If the comparison is with visitors who do not use Buddy, shoppers with stronger purchase intent may simply be more likely to ask Buddy questions, so the figure does not separate Buddy's effect from differences in who uses it. The order value lift is described only as "significant," with no number disclosed. Buddy's user numbers, usage rate and New Zealand results have not been published either.
What is Gemini Enterprise for CX?
Gemini Enterprise for CX, the foundation under Buddy, was announced by Google Cloud on January 11, 2026, at NRF in New York. It aims to handle everything from product discovery to post-purchase support through a single interface, and Google says its prebuilt agents can be deployed in days.
At its center is the Shopping agent. The announcement described three features.
- Complex reasoning: narrowing products to those that meet several conditions at once, such as fabric durability, dimensions and budget
- Multimodal input: understanding requests made through images, video and voice. One example was reading a photo of a handwritten recipe and adding the ingredients to the cart
- Consented action: going beyond information and links to add items to the cart and even handle checkout, with the shopper's consent
Google also introduced Customer Experience Agent Studio for building support agents, so that context carries over between sales and service. The table below lists the companies named at launch along with Ulta Beauty, which disclosed its adoption in April.
| Company | Sector and region | Work with Gemini Enterprise for CX | Announced |
|---|---|---|---|
| Bunnings | Home improvement (Australia and NZ) | Shopping agent Buddy, rolled out progressively on the Australian website and now on the NZ website and app | April 2026 (Australia), October 2026 (NZ) |
| Lowe's | Home improvement (U.S.) | Enhancing its advisor Mylow to give guidance personalized to a customer's home, project and location | January 2026 |
| Woolworths Group | Supermarkets (Australia) | Evolving its digital assistant Olive to plan meals and spot relevant specials | January 2026 |
| Kroger | Supermarkets (U.S.) | Building baskets around meals and comparing products; using Customer Experience Agent Studio to analyze customer calls to stores | January 2026 |
| Ulta Beauty | Beauty (U.S.) | Ulta AI, the AI assistant on its own website and app | April 2026 |
What sets Buddy apart is that although the Shopping agent can handle checkout, Bunnings kept checkout in its own online store. The agent focuses on advice, product selection and cart building, while the final purchase happens on the familiar screens. Lowe's is using the same product to enhance its advisor Mylow, and Home Depot has teamed with Google Cloud to expand Magic Apron. In home improvement, the "advisor" model is the one taking off first.
Why agents work in home improvement
A DIY purchase rarely ends with one product. Building a shelf takes boards, screws, brackets and paint, and sometimes a power drill. Google Cloud's customer story names the problem Bunnings faced: online shopping tends to be more transactional than in store, and customers planning complex projects often do not know all the components or equipment they need.
That is where an agent has room to help. Keyword search assumes the customer already knows what to buy. Project-based shopping breaks that assumption. In store, you can tell a team member you want to build a deck. Online, there was no equivalent. Buddy is an attempt to bring that in-store conversation online.
The finding that baskets grow with more conversation turns follows from this structure. As customers go through a project step by step, they learn about materials and tools they had not thought of and add them to the cart. Unlike routine grocery restocks, in home improvement telling customers what they need is itself the kind of service that drives sales.
Having 15 years of DIY content already searchable since the Ask Bunnings AI days was a major asset. Elliott said Bunnings has been using AI for five years and had already deployed a conversational assistant for store teams on handheld devices, among other tools. Buddy is an extension of that work.
Caveats: safety, data and trust
Safety is the weightiest concern. Bad DIY advice can lead to injuries or accidents. Google Cloud's customer story says Bunnings codified safety guardrails with the Google team so the tool prevents customers from attempting dangerous jobs themselves. But the specific criteria, such as which jobs count as dangerous and when customers are referred to professionals or store staff, have not been made public.
How personal data is handled will also be tested. Computer Weekly reports that Bunnings plans to integrate loyalty program data into Buddy so that, with customer consent, it can suggest tools from brands a customer already uses. The consent-led design is emphasized, but exactly what data will be used, and how far, has not yet been explained in detail.
Consumer trust has limits as well. FashionNetwork, reporting on a study by French consultancy Converteo, found that 47% of shoppers trust an AI assistant before purchase, but only 30% do at the point of payment. Some 71% worry about their personal data being exploited. The study covers the French e-commerce market, but Bunnings' decision to keep checkout on its own screens fits with these consumer attitudes.
On the relationship with store staff, Bunnings and Google repeat only that Buddy "complements" their advice. They have not explained how staffing or the role of in-store service might change.
Next up: voice, product pages and trade customers
Bunnings plans to add voice conversations with Buddy during 2026. It also intends to embed Buddy on product pages and build solutions for trade and commercial customers. Computer Weekly reports a plan to have Buddy handle initial support queries as well. If that happens, Bunnings will be testing in practice the "unified sales and support" that Gemini Enterprise for CX promises.
What e-commerce operators can take from this
Buddy's guidance comes from the how-to content and staff expertise Bunnings has built up over many years. Retailers selling products that need explanation of combinations or usage, such as furniture, appliances or outdoor gear, would do well to check whether their own guides, videos and FAQs are in a form AI can read.
Letting customers start shopping from a handwritten note or a photo of a part is a way to avoid losing customers who do not know product names. Keeping checkout on your own screens, proving results as an advisor first and then widening what the agent is trusted to do, is also a realistic sequence for many retailers.
Summary
Buddy is a case of supporting project-based shopping with an agent by bringing the in-store conversation online. Bunnings launched it in about six weeks in Australia, reported more than double the online conversion rate, and six months later expanded it to New Zealand.
What to watch next is whether effect figures, including the comparison baseline, are disclosed, and how far voice and trade-customer features actually progress. If the specifics of the guardrails are published, they will also help answer how much can be entrusted to agents in a safety-sensitive area like DIY.
