Română

AI for "blue collar" jobs: 5 (out of 10) projects to increase sales in the real economy: manufacturing, distribution, pharma

AI for "blue collar" jobs: 5 (out of 10) projects to increase sales in the real economy: manufacturing, distribution, pharma
24.02.2026
Updated on 11.03.2026

The IT market is saturated with AI for the office activities of "white-collar" workers (fr. Col blanc): generating emails, programming code, images and reports. What we bring to market with the series of technical guides "AI 2026 in B2B" is artificial intelligence for the backbone of the economy. This means:

  • Distribution: the movement of physical goods that reach the corner shop
  • Manufacturing: bringing new physical products to market
  • Any business-to-business negotiation: meeting contractual conditions for stock, price and quality.

AI Engineering

Traditionally considered "blue-collar" (fr. Col bleu), essential B2B activities depend on engineering, not magic. This is why we are publishing five projects drawn from the 10 in OPTI Guide 1 - AI for Sales.

Each project includes an architecture diagram in which we highlight the business rules that must be respected. This approach: artificial intelligence controlled by rule-based systems is the hybrid architecture we propose to the market.

"This guide proposes a pragmatic approach: hybrid architecture. You do not need to replace your current systems. A layer of intelligence will be built on top of existing data, generating new sales, under the strict control of business rules: minimum margin, stock levels, technical compatibilities, permissions."
Dr. Marian Călborean, Fulbright scholar, founder of OPTI Software
Marian Călborean

View on-prem diagram | Download the full guide (172 pages)


Copilot for Sales Agents

Recommended for teams with junior sales agents. Attention is required when implementing in legacy ERP systems to really be of help to the agent.

Side panel in CRM or ERP, or as an overlay over legacy Windows systems (e.g. AS/400, old Epicor instances)

What does AI do?

It recommends contextual upsell or cross-sell to the agent, e.g. during a call. With a custom model, it also recommends the best action (Next Best Action Model), e.g. "send them an email"

A chatbot (assistant) can also be added.

Ideal industries

Distribution of technical equipment (HVAC, Electrical) and Automotive parts

These are sectors where the catalogue contains thousands of references, and the training period for a new agent easily exceeds 6 months.

Logic flow

Logic flow Copilot for Sales Agents
Logic flow Copilot for Sales Agents

See in the guide: ROI indicators for shortening the learning curve of junior agents


Smart Substitutes Engine

Killer app for everyone who loses customers when out of stock. Probabilistic AI is not sufficient here; we recommend clean data, a product specification sheet and a classic constraints engine.

Product substitution when out of stock, for faster rotation or for delivery from a single location.

What does AI do?

Saving the sale when stock is 0 (e.g. automotive, IT, construction, pharma) due to supply or rapid distribution issues.

Transport optimisation when multiple locations exist (WMS).

Ideal industries

Construction materials distribution, Restaurant Supply, Medical and Pharma

Construction sites and end-users cannot wait, and a stockout in the presence of valid substitutes immediately results in losing the customer to the competition.

Logic flow

Logic flow Smart Substitutes Engine
Logic flow Smart Substitutes Engine

See the deliverable and why probabilistic AI is not sufficient without rules (guardrails)


Dynamic Bundle Generator

Effective in B2B. The key is that the discount should not be visible per product on the invoice, but global, so as not to accustom the customer to discounted prices.

Creating dynamic bundles to increase rotation without undermining positioning

What does AI do?

Suggests bundles of a bestseller plus a slow-moving product based on conversion probability (pCVR).

A discount is applied to the bundle (classic or with an AI model).

Ideal industries

FMCG (fast moving consumer goods), IT and electronics, Beauty / Cosmetics.

There are always stocks with short expiry dates or high-margin accessories to clear.

Logic flow

Logic flow Dynamic Bundle Generator
Logic flow Dynamic Bundle Generator

See in the guide the indicators that distinguish a bundle from a simple price reduction


"Frequently Bought Together" Module with Compatibility Awareness

Should be industry standard. Pay attention to volumes: if someone has already bought 1000 screws, the correct recommendation is an equal number of washers or screwdrivers, not more screws.

The classic "Customers who bought this also bought..." from sites like Amazon, but calibrated for B2B (industry and distribution).

What does AI do?

Identifies strong correlations (e.g. "Customers who buy pipe also buy fittings") based on historical orders. The model is trained on the transaction history of the past several years.

Ideal industries

DIY / Home Improvement, Plumbing and Heating Installations, Furniture Industry

Sectors where products are interdependent (e.g. pipe + fittings), and a wrong recommendation blocks the customer's project.

Logic flow

Logic flow Frequently Bought Together Module
Logic flow "Frequently Bought Together" Module

See the deliverable, risks and ROI indicators in the guide


Intelligent Replenishment

Can dramatically increase sales. AI is the only method that detects real seasonality without a fixed recurring interval (e.g. August and November in agriculture).

A system that learns the consumption cycle of each customer and proposes replenishment

What does AI do?

Predicts the date a customer's stock will run out and sends a proactive reminder (e.g. email).

If a farm buys seeds seasonally, the system alerts it (for fixed-cycle repetition, non-AI methods exist).

Ideal industries

Agriculture, Stationery and Office Supplies, Pet food / Veterinary

They have either climatic seasonality or recurring consumption with seasonal influences (e.g. tax filing season).

Logic flow

Logic flow Intelligent Replenishment
Logic flow Intelligent Replenishment

See the deliverable, risks and ROI indicators in the guide


Four ideal industries for AI - read the guide:

AI in Wholesale: Technical and IT Distribution

AI in Healthcare: Medical Equipment and Pharmaceutical Supplies

AI in Construction: Building Materials

AI in Agricultural Inputs: Seeds, Fertilisers, Treatments

Predictive AI vs Generative AI

The market today is flooded with generative AI solutions (such as ChatGPT), excellent for writing emails, generating images or summarising texts.

Our guide focuses on predictive AI. It does not hallucinate texts. It analyses millions of historical transactions to anticipate the next step: which stock will run out, which customer will churn, which product is compatible.

On top of predictive AI, generative applications such as assistants and chatbots can be built, and in the future AI agents. Always with the strict application of the company's business rules.

Read more in the guide

In B2B, engineering matters. We must know, in order to foresee, in order to act (fr. "Savoir, pour prévoir, afin de pouvoir." - Auguste Comte)

Discover all 10 deliverables by reading the OPTI guide




Quick Questions

What is the hybrid AI architecture proposed by OPTI?

An approach where predictive AI is controlled by business rules (minimum margin, stock levels, technical compatibilities, permissions). It does not replace existing systems but adds an intelligence layer on top of current data.

What is the difference between predictive and generative AI in B2B?

Generative AI (ChatGPT etc.) generates texts and images. Predictive AI analyses millions of historical transactions to anticipate which stock will run out, which customer will churn, or which product is compatible, which business guardrails then verify.

Why does Smart Substitutes require clean data and a constraints engine?

Probabilistic AI alone can propose technically incompatible or unprofitable substitutes. The constraints engine applies mandatory business rules (compatibility, margin, available stock) before presenting the suggestion to the customer.

Which industries are ideal for the AI projects in the guide?

Technical and IT distribution, Medical and pharmaceutical, Construction materials, and Agricultural inputs, sectors where recommendation errors have immediate financial impact.

What is the TLDR (conclusion)?

What we bring to market with the series of technical guides "AI 2026 in B2B" is artificial intelligence for the backbone of the economy: distribution, manufacturing, and business-to-business negotiations.

What technologies and methodologies are involved?

Technologies: Hybrid AI architecture, Predictive AI, Generative AI, CRM, ERP (Winmentor), WMS, Next Best Action, Constraints engine, Collaborative filtering, Vertex AI
Methodologies: Contextual copilot for agents (real-time upsell/cross-sell), intelligent product substitution on stock-out with constraints engine, dynamic bundle generation (bestseller + slow-moving product), historical order correlation with compatibility check, stock depletion prediction and proactive reminder, predictive AI on historical transactions with business guardrails

Marian Călborean

Article written by

Marian Călborean

Manager, Software Architect, PhD. in Logic, Fulbright Visiting Scholar (CUNY GC, 2023)

See on LinkedIn →
Interesat?

Interested?

Schedule a meeting

Get a Free Audit

News and Guides

More News