Stones: a new B2B e-commerce platform integrated with company systems and powered by AI
A furniture retailer ordering twenty units of the same table wants to know three things before confirming: the price, the actual availability and the lead time. If finding out means sending an email or waiting for a phon
A furniture retailer ordering twenty units of the same table wants to know three things before confirming: the price, the actual availability and the lead time. If finding out means sending an email or waiting for a phone call, the order gets pushed back.
For Stones we built a new B2B e-commerce portal that replaces the old platform based on Spree Commerce and pulls its information from the same systems the company uses every day. The project brings together the sales site, the integration with the PIM and with SAP, and Widgeto, our AI assistant, which answers customer questions directly on the portal's pages.
Stones and the B2B channel in furniture
Stones is an Italian furniture company that sells large volumes of products to its network of retailers. Its story began in 1982, and logistics has been part of its growth from the start. The first warehouse with an automated system dates back to 1995. In 2006 came the expansion and the new headquarters in Montelabbate, in the province of Pesaro and Urbino, and today the storage area covers about 20,000 mΒ², as the company's About us page explains.
In 2021 Stones invested in technological innovation, process digitalization and staff training, with an IT infrastructure that connects its management systems across the entire value chain. This is laid out in the timeline published by one of its retailers. Alongside the retail channel there is also Duzzle, the online shop for consumers that sells Stones products and that we manage on Shopify.
Market figures help explain why the portal was worth investing in. According to the B2b Digital Commerce Observatory of the Politecnico di Milano, B2B product orders handled digitally in Italy were worth β¬278 billion in 2024, up 5% from 2023. Yet they still account for only 22% of total transactions, a lower share than in Germany, France and the United Kingdom. Companies selling to retailers still have plenty of ground to cover.
On the consumer side, the furniture and home living sector grew by 6% in 2025 to reach β¬4.7 billion, according to data on B2C product e-commerce. This concerns consumers, so it should be read as context. Still, it describes people who are used to choosing a sofa from a screen, the same people who then walk into the stores supplied by Stones.
Why we moved away from Spree Commerce
Until the new project, Stones' B2B channel ran on Spree Commerce, an open source platform written in Ruby on Rails and first released in 2007. Spree is a solid tool and we know it well. The version the old portal was built on, however, had become outdated.
In the meantime the project has reached version 5, which its own developers present as the biggest open source release in their history. Those who have analyzed it in detail describe it as a complete rewrite of the architecture rather than an upgrade. If you are starting from older versions, the move looks a lot like a brand new project.
At that point the practical question was where to put the effort. Together with Stones we decided to invest in a solution built from scratch and fully customized. The goal was maximum flexibility and a portal that would adapt to the systems already in use and to the way the company works, instead of forcing Stones to adapt to the platform.
A portal built around the way retailers buy
A B2B e-commerce site follows different rules from a consumer shop. There are usually commercial terms tied to each individual customer, volume orders and frequent reorders of the same items. Bending a generic platform to these rules often leads to a chain of plugins and customizations that, over time, nobody wants to touch anymore.
Anyone who has managed an upgrade on a heavily modified system knows the problem: before changing versions you have to work out which extensions will break and why, and the cost often exceeds that of the change you actually wanted to make.
For Stones we designed the platform starting from the operational and commercial needs of the sales team and the warehouse. The advantage shows when something changes. A new order flow or a different sales rule is implemented in the code, without working around the limits of a system designed for someone else. Maintenance also becomes more predictable, because the software contains what the company needs and the people working on it know where to make changes.
Catalog and warehouse aligned with PIM and SAP
A PIM (Product Information Management) system collects product information in a single place and distributes it across all channels. In furniture, product sheets are heavy: dimensions, finishes, materials, packaging, images. Keeping them updated by hand across several channels is the quickest way to end up with a website that says one thing and a price list that says another.
The Stones portal is integrated with the PIM and synchronized with SAP for managing and replenishing warehouse stock. The result is a coordinated flow of information between the company's systems. The people who manage the catalog work in the PIM, those who track inventory use SAP, and the e-commerce site shows retailers data that comes from there, with no manual copying from one screen to another.
On this point, the decision to start from scratch made a real difference. With an off-the-shelf product, integration tends to follow whatever connectors are available. By writing the portal from the ground up, we were able to model the data exchange on the procedures Stones already uses, instead of redesigning them around the limits of the tool.
A connection like this also lays the groundwork for more advanced automation. If at some point you decide to add AI agents alongside your management systems, it pays to think about permissions and controls early on, as we explain in our article on the security risks of AI agents connected to ERP and CRM systems.
Widgeto answers customers on the site
The final piece is Widgeto, integrated directly into the Stones website. It reads the published pages and documents and answers users based only on the company's own information. When a retailer looks for a measurement, a product detail or a clarification on terms of sale, they get an answer right away, without opening a ticket.
Its strength lies in where the answers come from. A generic chatbot tends to make things up or talk around the question; one that draws on official content gives answers that can be verified and takes pressure off customer support. There is plenty of room for improvement: a Gartner survey of 5,728 customers, conducted in August 2024, found that only 14% of requests are fully resolved through self-service, as reported in this collection of statistics on AI for customer support.
Gartner also notes that customers are more willing to use AI when they know they can easily reach a human agent. That is why an assistant like Widgeto works best alongside customer service: it handles recurring questions and leaves the cases that need judgment to people. If you want to know how it came about and how it connects to the other tools we develop, you will find the full picture in ConSergio, Widgeto and Kleio: three products, one approach to software.
Where to start if you run a B2B e-commerce site
Before choosing a platform or an AI assistant, there are a few checks we always carry out with our clients, and you can do them yourself too.
- If your shop runs on an old version of a platform, ask for a concrete estimate of the upgrade and compare it with the cost of a new solution.
- List where your catalog, inventory and price list data live today, and decide which system is the source of truth for each.
- Define which information needs to move between e-commerce, PIM and ERP, and how often, before writing a single line of code.
- For a couple of weeks, collect the questions that reach your support team and note how many already have a written answer in a document or on a page of your site.
That set of questions tells you how much work an assistant like Widgeto can take on, and what content is missing for it to work well. To choose where to begin, you may find our 8-question checklist for deciding which process to automate with AI useful.
If you would rather discuss your own systems directly, you can start with our free AI assessment. We look together at your current workflows, the platform you use today and the connected management systems, then show you where an integration or an assistant would deliver a measurable result.
Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.