VARIANT-AI

Furniture collections combine dimensions, materials, colours, fabrics, construction options and accessories into large numbers of possible configurations. This complexity makes production and inventory planning difficult. Customers expect individual choice, while manufacturers need sufficiently large production batches: small orders increase costs, whereas larger batches can generate excess stock and tie up working capital.

Brands, retailers, manufacturers and material suppliers often operate separate information systems. Orders, prices, material availability, production capacity and delivery dates are coordinated through repeated manual exchanges using incomplete or outdated information. Consequences include slow quotation and delivery-date confirmation, misaligned procurement, unsuitable minimum order quantities, underused capacity and delays caused by unavailable components.

Retailers and brands hold valuable information about customer preferences, sales trends, order histories and collection performance. These signals are often disconnected from manufacturing schedules and material purchasing. Businesses need to translate market demand into feasible operational plans and respond faster when demand or supply conditions change.

Cross-company coordination must accommodate different commercial objectives without requiring participants to disclose all sensitive information. It must also understand product configurations beyond individual stock-keeping units, prevent infeasible combinations and work with existing industrial systems. Companies need clear decision authority, traceable actions and human approval for consequential changes rather than unrestricted automation.

VARIANT-AI will combine a Demand & Collection Intelligence Agent for forecasting and assortment planning, a Sales & Negotiation Agent for configurations and quotations, a Manufacturing Planning Agent for batches and schedules, and a Procurement & Supplier Agent for sourcing and material availability. Product models will link variants to materials, processes, compatibility and manufacturing constraints.

Generative AI will interpret information and coordinate alternatives; mathematical multi-criteria optimisation will quantify costs, margins, stocks, capacities and lead times. A digital supply-chain model will simulate scenarios. The Governance & Human Approval Agent will distinguish recommendations, coordinated alternatives, permitted autonomous execution and mandatory human approval, with defined permissions, escalation and audit trails. Design- or quality-affecting substitutions and major discounts will require approval.

VLS d.o.o., operating the MIOTTO brand, owns the product and market use case and provides the sales-side pilot, linking customers, collections and production partners. It will define requirements and use cases, develop collection-management scenarios, validate sales and procurement processes, test agents, assess economic effects and prepare a commercial exploitation model.

We are looking for an AI/platform technology partner to develop architecture, communication protocols, negotiation, orchestration, governance and AI–optimisation integration; an industrial systems integrator to connect ERP/MES/PLM and other data sources; a Turkish furniture manufacturer to provide production data, modelling input and the production pilot; and a materials or components supplier to demonstrate availability, minimum-order and delivery coordination.

Work will begin with industrial process mapping, business requirements, pilot scenarios and baseline measures. Analysis of sales, orders, customer preferences and product variants will support variant-level demand forecasting, identification of promising configurations, assortment optimisation and new-collection planning.

Development will cover negotiation under incomplete information and conflicting objectives; compatible-order aggregation, batch optimisation, capacity- and resource-constrained scheduling; and replanning after disruptions. Procurement models will account for prices, minimum order quantities, transport costs, lead times and alternative materials. Data models, interfaces and controlled information sharing will connect existing systems while protecting commercial confidentiality, without replacing the IT infrastructure.

Planned linked pilots will connect VLS d.o.o., the manufacturer and the supplier. Sofa orders varying in dimensions, colours and fabrics will test feasibility checks, price and delivery negotiations, material procurement and responses to shortages or delays. Simulations will compare configurations, quantities, production sequences and delivery dates, assessing costs, margins, stocks and approval rules before execution. Deployment in real business environments will also assess capacity utilisation, working-capital effects and responsiveness to demand changes.

Comparison with existing workflows will test preliminary targets of 15–25% less excess stock, 30–50% shorter quotation-feasibility checking time, 10–20% better delivery reliability and 20–30% less manual coordination; these are not measured results. Commercialisation will cover results protection, market-entry planning, licensing, SaaS subscriptions, implementation services and industry-specific solutions, plus transfer to mattresses, hospitality and commercial interiors, other configurable products and networks with multiple suppliers and subcontractors.

• VLS d.o.o. 🇸🇮 SI