We solve

Too much copy-paste, slow quoting

The sales team loses 30min-1h / offer. They look for each item in the ERP, ask the agent if they have noted it well, sometimes they even go back to the customer. The customer waits, yet the opportunity cools down.

Product and package retrieval errors

There is confusion between almost identical variants and to choose one must either investigate the invoices associated with the parts or ask the agent or customer. If the wrong part is delivered, the customer will return the goods.

Missed upsell opportunities

Quoting often does not include upsell and bundles recommendations that could be automated for growth. They must be automatic, centrally controlled by management and applicable via chatbots to ecommerce clients.

Manual Quoting vs. AI Sales by OPTI

The Classic Process Our Process
Slow search: The agent searches for exact codes. If they search by title and make a typo, the ERP returns nothing. Semantic search: The AI understands "concrete screws" and finds the correct items, even without a code.
Manual stock check: The agent opens 3 windows to check stock in different warehouses. Total visibility: Stock from all inventory locations and the client-specific price are displayed instantly.
Missed Upsell: The agent forgets to propose accessories (washers, consumables) due to rushing. Automatic Upsell: The platform automatically suggests complementary products defined by the manager.
Time: 30-60 minutes for a complex quote. Time: < 5 minutes for the same quote.
Ask for a Demo

We deliver

Aplicații web și portaluri

Writing (OCR) and vocal transcription

Extracts products from huge PDFs, recognizes handwriting / photos for hand notes, allows audio commands.
Data și business intelligence

Semantic and hybrid search

Finds the product "by meaning", not just by code. Google Cloud Tech understands and complete the customer's intent.
Securitate și calitate IT

Complementary products

We define client business rules and machine learning rules for adaptation and upsell (ex: screw+washer).
ERP/CRM, software 3rd party

Continuously synchronized data

Multiple prices, product details, sales and customer history, catalog milestones — rapid refresh (e.g. every 15 minutes).
Aplicatii iOS si Android

Dashboard for offer creation

Each sales agent has access to their own dashboard to quickly create offers while adhering to common limits.
Soluții personalizate

On-site sales chatbot

Trained on your catalog, responds anytime (24/7), collects leads and schedules meetings.

Proven results: industrial client

-68%
time per offer. From 33 to 7 minutes for a medium complexity offer.
<1%
errors in catalog search. From 5-10% of searches leading to errors.
+14%
average order value. AI recommendations lead to upsell and increased sales.
<2 months
delivery in 7 weeks, approx. 200 hours of work, for full operation.
Interesat?

Interested?

Schedule a meeting

Get a Free Audit

Who uses the platform:

B2B Distributors

For companies with catalogs of +5,000 items (SKUs). Transforms the sales ERP into a fast search engine for field agents with upsell opportunities.

Retail & eCommerce

For online stores that want to provide 24/7 support. The public chatbot turns visitors into buyers using data from the existing eCommerce solution.

Technical and Support Teams

For IT departments seeking business process automation (technical support, document processing) without changing infrastructure.


References

Phases of Setup

Data & Analysis: objectives, product catalog, attributes

ERP Integration: API/ODBC/CSV + price list mappings, units, stocks

Gemini AI Pilot (2–3 weeks): OCR + Vocal + Semantic search on real data

Testing period (30 days): real agents, chatbots, sales reports

Stable implementation: Continuous optimization, A/B tests

Process and methodology

See the architecture and value flow


Most companies have their data locked in a rigid ERP or in Excel files. AI Sales by OPTI Quoting Platform connects as an intelligent middleware layer. We extract in real-time, we process for indexing and consistency and we deliver a fast dashboard for agents, chatbots for clients, and quotes back into the ERP.
Technologies: Google Cloud, Google Gemini, Document AI, Speech to Text, Vision, BigQuery, Vertex AI Search, Vertex AI Recommendations, Dialogflow CX, Cloud Logging, ERP, Senior, Charisma, Entersoft, Magister
Methodologies: OCR, Speech-to-text, Semantical search, Lexical search, Upsell rules, Model Training, A/B tests, Audit & Compliance, Integration via API, SQL, CSV, RPA

ERP platforms: Romanian and international

Romanian:

The platform is a smart quoting extension for Senior, Charisma, Entersoft, Winmentor, Nexus, Magister, Pluriva. We import the catalogue, offers and stocks in (almost) real-time, without modifying them.

International:

Microsoft D365, SAP Business One, etc.

We integrate any ERP via API, ODBC, RPA or scheduled imports (CSV, CDC, flat files etc).

We do not replace your ERP solution, we make it smarter!


NEW AI Tech Guides for Companies

For clients who want AI implementations in their company, we publish tech guides. Read the guides for statistics, problems, and technological solutions. Read the published AI for Sales Guide #1

LIVE

Recommendations, upsell, and rules

Problems: Agents don't propose compatible accessories and don't respect management rules.

KPI: +20% increased AI-assisted sales.
Read

Integrating exact data with AI (RAG)

Problems: AI doesn't search inside ERP data. When we upload fragmented data, "hallucinations" appear, including for price and stock.

KPI: 94% hallucination evitation in RAG.
Preview Guide #2

Input processing: text, voice, photo

Problems: Orders on WhatsApp, photos of handwritten notes, voice messages, RFQs hundreds of pages long.

KPI: <5 sec transcribing 50 products from image.
Preview Guide #3

Hybrid search & assistants (chatbots)

Problems: The client wants "wood screw", the ERP has "self-tapping screw". Yet for "SKU-8521", they must not get "SKU-8522". 40 applications and chatbots.

KPI: -90% no-result searches.
Preview Guide #4

Continuous optimization and learning

Problems: Static software that doesn't adapt to users and the market.

KPI: 2.5x scaling speed.
Preview Guide #5

Security, audit, and standards

Problems: GDPR/ISO/NIS2 compliance, protecting company data.

KPI: 0 detected deviations from company policy.
Preview Guide #6

Packages

Internal Sales

  • Quote generation in 5 minutes
  • Romanian ERP integration
  • 1 upsell per category (max. 7)
  • 1 internal search chatbot
  • Agent dashboard
  • Warranty and technical support

Ecommerce Starter

  • All features from Internal Sales
  • Ecommerce integration: Woo/Magento/Custom
  • Multiple upsell rules
  • Public website search chatbot
  • Sales reports
  • SSO or login integration

Omni-Channel Sales

  • All features from Ecommerce Starter
  • CRM integration: HubSpot etc.
  • Chatbots with custom flows
  • Continuous optimization
  • Search reports (A/B testing)
  • International ERP: D365, SAP
Ask for Pricing Details

Steps Overview

Step 1: ERP Analysis & Integration

We start by analyzing your ERP data (products, prices, stock levels). We establish a secure connection via API, ODBC, or scheduled CSV imports to synchronize your product catalog with our platform.

Step 2: Pilot Phase with Gemini AI

We configure and train the Google Gemini AI model on your product catalog. We implement the semantic search and OCR features and run a 2-3 week pilot on real-world examples (e.g., agent notes) to validate its accuracy.

Step 3: Dashboard & Upsell Rules Configuration

We develop the quoting dashboard for your sales team. Together with you, we define and implement the business rules for complementary product recommendations (upsell/cross-sell) to increase order value.

Step 4: 30-Day Testing and Agent Training

We train your sales agents to use the new platform. We measure key KPIs: reduction in quoting time, decrease in product lookup errors, and increase in average order value.

Step 5: Go-Live & Continuous Optimization

After a successful trial, we launch the platform for the entire team. We provide ongoing support and use machine learning to continuously optimize the performance of the AI model and its recommendations.

Frequent questions

How long does it take to see results?

Typically, a PoC is delivered in 2–4 weeks; full projects in ~7 weeks according to standard implementations.

If my ERP doesn't have an API?

We can start data ingestion via CSV/Excel or ODBC to connect to SQL databases. We use automation technologies (e.g., UiPath) when these options do not work. Data can definitely be extracted!

Can the data remain in the EU?

Data can be hosted in EU regions (e.g., Belgium/Netherlands/Germany/Poland) or in EU multi-region. For AI services, we select EU endpoints/regions where available; if a service does not yet offer "EU-only", we inform in advance and propose alternatives.

How do we avoid item errors and product code retrieval issues?

We use semantic (vector) search to understand the customer's intent, not just the exact code match. Thus, even if the customer does not remember the correct code, Gemini can suggest the appropriate items based on the description or context. In addition, we perform a parallel exact search for cases when the code is correct.

Can the public chatbot give wrong prices?

A full implementation based on RAG (Retrieval-Augmented Generation) technology reduces hallucinations by 50% and can bring error rates down to 4-5%. Additionally, we have special safeguards to ensure prices and stock levels are retrieved from standard relational databases.

Do you have specific industries where the solution is applied?

Yes.

1. Construction Materials & DIY: There are thousands of similar items and different units of measure (sqm vs box). Plus, tender requests consisting of hundreds of pages are received.

2. Auto Parts & Service: Part codes are complex, compatibilities are strict, and there is a very high volume of products.

3. The Book Industry: There are tens or hundreds of thousands of titles requiring advanced search and intelligent recommendations, including dialogue with the highly rich content.

Schedule a Meeting

Last updated: 16.02.2026