Skip to content
Studio Beltramelli

Integrated digital systems and artificial intelligence for SMEs

We build high-performance B2B and B2C e-commerce platforms, connect ERP, logistics and CRM systems and bring artificial intelligence models into business processes, including on the company's own infrastructure.

Integrated systems

Company systems integrated around the middleware. Inside the company perimeter ERP, DMS, WMS, PIM and the middleware with a message queue to the channels. Outside the perimeter a SaaS CRM synchronised in batches, EDI with trading partners, a B2B portal, a B2C store with its payment gateway and a TMS for the carriers. No outside system talks to the ERP directly.DMSERPWMSMIDDLEWAREPIMQUEUECRMEDIB2BB2CPAYMENTTMSSYSTEMScompany perimeter

We build B2B portals synchronised with the ERP and high-performance B2C e-commerce platforms for large catalogues with variants, compatibility data, customer-specific price lists and live stock, where search quality and loading times directly affect sales and day-to-day operations. We design the data flows between ERP, warehouse, logistics and CRM so that prices, stock and orders stay aligned across every channel. Where needed, we develop the custom websites and web applications that complete the system. After release we look after the platform with maintenance, updates and further development.

SCOPE
B2B, B2C, catalogues of hundreds of thousands of items
INTEGRATION
ERP systems, warehouse, logistics, CRM
OPERATIONS
maintenance, updates, further development

AI integration

Artificial intelligence integrated with company data. Inside the company perimeter documents from the DMS go through OCR and normalisation into a vector database; RAG retrieves context from it and the ERP supplies master data, both to the on-prem LLM; the answer of the model goes through guardrails and human review (HITL) to the output; a monitor observes the model. Outside the perimeter a cloud LLM API as an option.DMSINGESTVECTOR DBRAGERPLLMLLM APIMONITORGUARDRAILHITLOUTPUTAIcompany perimeter

We apply artificial intelligence to the company's operational, commercial and marketing workflows: extraction, classification and normalisation of data from documents and orders, generation and enrichment of content and product data, semantic search, assistants and agents that work on company data. We connect models to internal archives and, where useful, specialise them on the terminology of the industry. Every integration is designed with security and human oversight first, in the cloud or with language models installed on the company's own infrastructure (on-prem LLMs) when data must stay in-house.

SCOPE
operations, documents, catalogue, marketing
TECHNIQUES
semantic search (RAG), model specialisation (fine-tuning), agents
DATA
owned and controlled by the company, cloud or on-prem

Infrastructure and security

The company's hybrid network. Inside the company perimeter a firewall on the border separates the outside, the DMZ with the reverse proxy and the LAN, which holds an identity provider (SSO, MFA) that authenticates VPN access, central logs, production and development applications on separate VLANs, a database reached by production only, an AI server and backup. Outside the perimeter remote access by VPN, a CDN with WAF and cache and the IaaS hosting cloud.IDPDMZLOGFWVPNDBAPP PRODCDNBACKUPAPP DEVCLOUDAI SRVINFRASTRUCTUREcompany perimeter

We design and run the infrastructure the platforms depend on, in the cloud or on the company's own servers. Separate development and production environments, devops practices for automated and repeatable releases, verified backups, monitoring and security updates. We size resources on actual volumes and scale them when needed, with service continuity and the protection of data and access as priorities.

SCOPE
cloud, on-prem, hybrid
WORK
environments, releases, backups, monitoring
SECURITY
data, access, service continuity

Method

Diagnosis first. Every project starts with an analysis of the systems in use and of the data as it stands. The work plan, with phases and priorities, comes before the code.

Sign-off before every change. Nothing reaches production without a review agreed with the company. Decisions are kept on record in the project documentation.

The company owns code, data and documentation. Everything we build for a client, with its documentation and infrastructure access, is handed over to the company, which is free to have it managed by whoever it chooses.

Data stays in-house when required. When data must not leave the business, systems and artificial intelligence models are installed on the company's own infrastructure.

Compliance support. We design with the EU AI Act, GDPR, NIS2 and accessibility requirements in mind and produce the technical documentation useful for audits and security questionnaires. We work alongside the company and its advisers as a technical partner, for the part that depends on the systems.

Artificial intelligence in our own work. We use artificial intelligence tools in developing and running systems, this website included. Technical choices, verification and responsibility for what we release remain ours.

Frequently asked questions

What kind of companies do you work with?

SMEs that sell through B2B or B2C digital channels and manage catalogues, orders and data across several systems. Size matters less than complexity. The typical case is a company with an ERP, a warehouse system and an e-commerce platform already in production, that needs to make them work together or grow.

Do you have the structure to handle a complex project?

Yes. A lean structure means quick decisions, a single point of contact for the whole project and low overheads that are reflected in our proposals. For the specialist parts of larger projects we involve suppliers and developers we have worked with for years, coordinated by us and under our responsibility.

Do you also work with software houses, agencies and ERP partners?

Yes. When a project calls for a platform on a large-scale catalogue, a complex integration or dedicated infrastructure, we work with the software house, agency or ERP partner that already serves the company. Each party's scope is written down before the start and the ERP, with the services the partner already provides, remains the partner's.

Do you work with our ERP?

Yes. We integrate Italian and international ERP systems, industry-specific solutions and custom-built systems. Integration goes through the interfaces provided by the vendor and, where needed, together with the vendor's technicians. On large volumes synchronisation is incremental and the ERP exchanges only what has changed.

Our integration uses an interface the ERP vendor is retiring. What happens?

The integration layer is rewritten, not the ERP. We identify which flows depend on the retiring interface and move them to the interfaces the vendor supports, once, with a switch planned so that orders, price lists and stock are not interrupted. Upgrading the ERP itself stays with its vendor.

Our ERP already has a B2B module and standard platforms exist. Why a specialised studio?

Not always needed. If the company has a single ERP, price lists by customer group and a catalogue of modest size, the ERP module or a standard platform is enough and we say so. The studio is for problems of scale and integration: catalogues of hundreds of thousands of items with variants and compatibility data, net prices and stock per individual customer, several systems and channels to keep aligned, search and loading performance that affects sales, code and data that must remain the company's. In those cases the preliminary analysis compares the two paths on actual price rules and volumes before any decision.

Do we have to replace the platform we use today?

Not necessarily. The preliminary analysis assesses whether the current platform can evolve or whether replacing it is the better option, with the costs and risks of both paths. When replacement is the right choice, the migration is planned so that sales are not interrupted.

Who owns the code and the data?

The company. The code we develop on commission, with its documentation and infrastructure access, is handed over to the company, which is free to have it managed by whoever it chooses.

Is our data used to train artificial intelligence models?

No. We use services that contractually exclude the use of data for training. When data must not leave the company, models are installed on the company's own infrastructure.

What happens after release?

We look after the system in production with monitoring, security updates, maintenance and further development. The service level is set out in a support agreement sized to the company's needs.