We hosted a Masterclass webinar about how to use our Vaimo AI Assistant to increase in-store sales and boost customer loyalty. PJ Utsi, Co-Founder and Chief Creative Officer, and Firuzé French, Lead CX Consultant, conducted a live demo of the in-store AI Assistant we built for Elon, the leading Nordic home electronics and appliances retailer.
Watch the recording of the AI sales assistant masterclass »
Why did Elon implement an in‑store AI Assistant?
We’re past the initial hype of generative AI and into the practical phase. Buyer journeys are longer and more research‑heavy; customers arrive in stores highly informed and expect seamless online‑to‑offline continuity, personal advice, and fair pricing.
It’s the right moment to deploy AI that tangibly shortens discovery, supports staff, and connects channels, while still proving ROI through clear value areas (customer experience, sales enablement, business ops, and data/insights).
How does the assistant personalize recommendations?
Recommendations blend session inputs (category, price, family size, brand preferences, etc.) with proprietary signals: customer profile and order/browse history, return‑rate data, Zendesk support trends, local store inventory, and campaign/margin guidance.
This helps avoid dead ends like out‑of‑stock or negative‑margin items and steers toward products with lower likelihood of returns or support issues.
Price matching: how does the “campaign comparison” work?
The rep pastes a competitor product URL. The AI verifies the competing offer (price, model, features), maps it to Elon’s assortment, and generates talking points, either recommending a comparable/ better alternative at Elon or confirming a price‑match route. It arms the rep with fair, on‑the‑spot rationale customers can trust.
How does it help new or rotating staff sell confidently?
On any product page, the AI can switch into Sales Mode and present concise “sales highlights” and talking points (e.g., what SteamCare or SensiCare+ actually means in daily life). It doubles as a just‑in‑time training academy for categories a rep is less familiar with, speeding onboarding and standardizing how features are pitched.
What about post‑purchase issues and technical questions?
Support Mode surfaces common issues, deep links to the exact user manual, and step‑by‑step fixes (e.g., factory reset button combos, error code lookups like E20). If presets aren’t enough, the rep can ask the assistant in natural language or by voice for product‑specific troubleshooting.
How does the tool bridge online and offline in daily store work?
Reps can build local Lists (e.g., “washers on display,” “October campaign,” “favorite coffee makers”) to quickly access what’s actually present in that store. From lists or product pages, they can push items to a customer’s cart/wishlist or create a Quote that’s saved to the account and shareable via SMS/email/print, so the customer can complete later online or in another channel. The Unified Checkout supports paying online (e.g., BNPL, Swish) or in‑store via POS, while attributing the sale correctly to the store and rep.
How does sales rep attribution work?
As long as the rep is logged in on the device and operating in the right store context, all key actions (customer lookups, quotes, in‑store orders, and even subsequent online orders stemming from that journey) are tracked to that user/store. This removes the historic tension between channels by making digital an ally to physical retail.
What did Elon need in place to start?
Not perfection, just access. The assistant was trained on Elon’s existing stack and content: the ecommerce site (Adobe Commerce), PIM data, CRM/loyalty & personalization, order/return signals, POS exports, Zendesk tickets (where SKUs are referenced), and a trove of PDFs in SharePoint (sales briefs, campaign docs, manuals).
The AI ingests unstructured material well; the key is making sources reachable.
How does the system cope when stock, margins or campaigns change daily?
Recommendations update instantly. If inventory flips, a margin turns negative, a campaign switches, or the customer’s online behavior changes between visits, the assistant recalibrates.
That’s the point: keeping guidance accurate without relying on reps to memorize every weekly change.
Which AI tool should we start with?
Start from the use case and data you want to leverage (e.g., price comparison, troubleshooting, or staff training). In Elon’s pilot, we used OpenAI models as the primary engine, but the choice depends on your constraints (data residency, governance, latency/cost, multimodal needs, function-calling/RAG).
A practical pattern is: pilot rapidly with a top‑tier hosted model (OpenAI/Azure OpenAI), keep your proprietary data as the differentiator, and design a vendor‑agnostic layer so you can swap/augment with Google or open‑source models later without re‑platforming.
How are privacy and governance handled when reps “log in as the customer”?
The customer verifies their identity (e.g., via email) before the rep views the account‑tied elements like wishlists or loyalty offers. Access happens within the retailer’s authenticated environment, inherits existing roles/permissions, and is bound to the service moment. You can also log policy requirements (e.g., consent prompts, auto‑logout) and audit trails to meet governance standards.
Can we roll this out store‑by‑store or category‑by‑category?
Yes. Many teams start with one region and a few high‑value categories (e.g., white goods) to limit data wrangling and change‑management. From there, you widen lists, training prompts, and troubleshooting libraries.
What if our ERP/POS exports are messy or lack APIs?
That’s common. You can begin with cleaner sources (ecommerce, PIM, CRM, PDFs) and add legacy feeds later via scheduled exports and parsing. The assistant is intentionally resilient to imperfect, unstructured inputs; you’ll still get value while you modernize the harder systems in parallel.
Do we have to re‑platform our website to do this?
No. The assistant is a lightweight layer over your existing site/app experience with a staff‑mode UI. It’s designed to “meet your stack where it is,” not force a re‑platform.
What’s a pragmatic first use case if we’re unsure where to start?
Two low‑friction winners from the webinar: (1) Troubleshooting/Support Mode powered by your Zendesk/knowledge PDFs (fast to prove value), and (2) Price/Competitor Comparison for highly shopped categories. Both improve CX immediately and build confidence internally.
How do we measure ROI beyond immediate conversion?
Track: assisted revenue (with rep attribution), quote‑to‑order rate, time‑to‑first‑sale for new hires, average handling time on support queries in‑store, reduction in returns for AI‑recommended items, and customer satisfaction. Layer in qualitative store feedback to prioritize next steps on the roadmap.
Want a deeper dive?
We’ve got plenty of resources for you. Read about our AI-powered sales assistant, and check out how we implemented the assistant for electronics home giant, Elon, in our case study.
Be sure to watch a recording of our webinar below!
Watch the recording of the AI sales assistant masterclass »
