Zremb

Key Requirements

We helped secure an innovation grant of PLN 3 million for ZREMB.

Thanks to a Lean approach, we successfully decreased the project budget for the MVP version from PLN 3 million projected by the client to PLN 500 thousand, optimizing the budget use.

Martha won the PropTech Festival 2022competition among the likes of Żabka Nano and Park Cash.

We greatly optimized the estimated implementation time for an MVP from 2 years to just 6 months.

About the client

ZREMB (eng. ZREMB Elevator Equipment Company) is an established company with over 70 years of successful presence in the Polish market. ZREMB excels in providing holistic technological solutions for elevators, with a distinct emphasis on sustainability and environmental preservation. Their expertise lies in delivering cutting-edge elevator systems engineered for heavy-duty use, ensuring prolonged operational lifespans without the necessity for frequent replacements.

What truly distinguishes ZREMB and contributes to their success are seamlessly integrated lift control systems. These systems boast universal compatibility with equipment manufacturers worldwide, setting a new standard for industry synergy. Their broad offer included extensive elevator solutions, encompassing small freight elevators, accessible platforms, passenger and freight elevators, escalators, moving walkways, and steel construction elevator shafts. From installation and modernization to repair and maintenance – they've got you covered.

About the project

Operating in a highly specialized field of industrial automation and seeking innovation that would further confirm their role as an industry leader, ZREMB has envisioned a state-of-the-art elevator management system.

Here comes Martha – an AI-powered robot with a mission to offer a personalized, human-like approach instead of the simple press-the-button interaction we know from elevators installed in our buildings. To turn our client’s vision into reality, Martha needed to utilize a well-rounded machine-human interface, incorporating automatic recognition and voice & touch interfaces.

Yet – that’s not all. It was necessary to introduce graceful degradation so that when one interface is not preferred by the user, the other one takes over. The users are, therefore, empowered to switch interfaces as they see fit. Combined with a must-have support for the Polish language and offline accessibility, Martha called for a custom solution.

The ZREMB project wouldn’t have been possible without ITCorner. It all started when Rafał Pisz, CEO of QuantUp, shared information about the Martha project within our community. With vast experience and a proven track record of successfully navigating innovative projects, Vazco was strongly recommended as a business & technology partner for ZREMB.

Strategic advantage

How did we navigate the client for success?

Comparative Model Testing

We evaluated four different speech-to-text and classification models, assessing:

Transcription Quality (Word Error Rate, Character Error Rate (CER), Sentence Error Rate (SER))
Contextual Comprehension (handling domain-specific terminology, brand references)
Robustness against background noise or varied accents

This multi-model approach allowed us to select the optimal pipeline for the client’s environment.

Rule-Based + AI Hybrid

While we leveraged advanced AI classification for nuanced conversation contexts, we combined it with a rule-based layer to capture explicit triggers:

Exact Phrases
Policy-Specific Markers

This hybrid approach reduced the chance of missing straightforward red flags or incorrectly categorizing calls where the conversation closely mirrored a known scenario.

End-to-End Call Analysis Flow

Audio Input

Exact Phrases

Transcription

Our team rigorously evaluated and measured each model's effectiveness to ensure it delivers optimal transcription accuracy and cost efficiency.

Policy Violation Detection

AI classifier + rule-based checks parse the transcript, identifying any deviation from the 18 internal guidelines (e.g., offering courses below age 6, misquoting pricing, failing to mention required disclaimers).

Scoring & Reporting

Each call receives an accuracy score, and flagged segments are highlighted. A structured report pinpoints any policy breaches, along with recommended next steps.

Key AI Techniques & Highlights

High-Accuracy Transcription Pipeline

We systematically compared four STT (Speech-to-Text) solutions, focusing on domain adaptation for education and sales.

Domain-Specific Language Model

Custom dictionaries and post-processing rules handle brand-specific terms, typical questions, and competitor mentions.

Contextual Classification

Our system uses ML classifiers to gauge context - was “free” used in a correct or violating scenario? Did the rep mention “PESEL” or “ID” requirements for the right target audience?

Self-Validation Against QA

We tested the system’s outputs against a human QA reference set. In one instance, the AI identified an error that human QA had incorrectly labeled as “passing” - resulting in an impressive “1:0 for AI”.

Handling False Positives

We discovered one false positive triggered by a transcription error. With minor refinements to domain-specific synonyms, we swiftly corrected the model - fully eliminating that error class in subsequent tests.

Results & Impact

Human-Level Accuracy

Side-by-side testing showed that on provided call sample, the AI not only matched QA performance but managed to outperform it through identifying policy violations in overlooked by QA (false-negative) and captured cases where QA wrongly marked fulfilled policy (false-positive).

Scalable Insights

The system aggregates errors into high-level reports - revealing patterns (e.g., repeated mention of wrong age limit or missed disclaimers) that inform targeted training programs.

Reduced QA Time & Cost

By automating the bulk of call analysis, manual review hours are dramatically reduced, allowing the QA team to focus on complex issues and targeted coaching. It showed potential for covering 100% of the calls in given time and cost.

Consistent Branding & Compliance

The AI solution enforces a uniform standard, ensuring all reps follow the same guidelines for pricing disclosure, disclaimers, and marketing consents.

x10

Test Coverage Increase

>80%

QA Cost Reduction

99%

AI Test Accuracy

Test Coverage Increase

500%

Test Coverage Increase

80% less manual input thanks to smart data sync across CRMs

Key Requirements

About the client

About the project

How did we navigate the client for success?

Key AI Techniques & Highlights

Results & Impact

Lead innovation in your industry,
just like ZREMB does

80% less manual input thanks to smart data sync across CRMs

Key Requirements

About the client

About the project

How did we navigate the client for success?

Key AI Techniques & Highlights

Results & Impact

Lead innovation in your industry, just like ZREMB does

Lead innovation in your industry,
just like ZREMB does