Improving Quote Turnaround with AI-Powered Fixture Image Analysis

A U.S.-based provider of custom retail environments and fixtures needed to streamline its manual draft quote generation process involving image analysis, SQL data retrieval, pricing validation, and historical quote comparison.

The company’s draft quote generation process was largely manual, requiring intensive coordination across teams and systems due to the absence of a unified workflow. This led to operational inefficiencies, limited scalability, and reduced consistency in quote preparation.

Key challenges included:

Input Variability and Data Gaps

• Multiple customer-specific document layouts and product depictions made it difficult to standardize inputs
• Fixture images varied in structure and detail, requiring manual interpretation
• Increased dependency on individual expertise led to inconsistent quote preparation
• Inputs often lacked key specifications such as length, breadth, and quantity
• Missing data required manual assumptions or follow-ups, increasing turnaround time and risk of inaccuracies

Manual Data Retrieval and Tagging

• Product details had to be retrieved manually from the MyPD database
• The process involved navigating multiple systems, reducing efficiency and increasing the risk of errors

Tagging and Data Mapping Complexity

• Images and detected objects required custom tagging before quote generation
• Tags had to be mapped with MyPD terminologies to retrieve relevant product and quote records

Pricing and Validation Complexity

• MyPD did not provide standard pricing, requiring manual price validation
• Pricing for existing customers depended on averages from previous MyPD quotes and required additional validation

Dependence on Historical Data

• Criteria had to be defined for retrieving relevant MyPD information and generating new quotes from similar previous quotes

AI Adoption Constraints

• Image selection and model training could take longer and extend project timelines
• Prediction accuracy was expected to be low initially and improve gradually with model training

Evoke developed an AI-based application named “AI Ballpark” to generate a draft quote by analyzing existing quotes and data available in the Customers DB and existing applications.

The solution included:

Application Access and User Interface

• Secure login to the AI application using Active Directory authentication
• A user interface that allowed users to upload fixture images and enter the required quantity

Image Analysis and Object Extraction

• Image analysis using a Vision Language Model, such as Azure Custom Vision, to identify objects and fixtures from uploaded images
• Extraction of identified objects and fixtures from the uploaded image

Data Mapping and Retrieval

• Matching of extracted object names with MyPD object terminology using a combined view from the MyPD SQL database
• Retrieval of the top five matching records from MyPD based on predefined criteria jointly defined by PD Instore and Evoke
• Display of matched records and related summaries within the application

Quote Generation Workflow

• Ability for users to select relevant matched records from the application UI for quote generation
• A “Provide Ballpark” feature to generate a sample quote based on selected historical records and predefined quote-generation criteria

Quote Distribution and Tracking

• Option to download the generated quote in PDF format
• Option to send the generated quote as an email attachment to specified recipients
• Logging of quote-generation activities, including the email addresses to which quotes were sent
• A log-view screen to filter and review application activities

Scalability and Future Readiness

• A scalable application design to accommodate multiple AI agents for PD Instore in the future

Evoke’s AI-based application helped the company accelerate ballpark quote creation while enabling PDF downloads, email sharing, and activity tracking. This resulted in the following

Faster Quote Generation

• AI-based draft quote generation for uploaded fixture images
• Object and fixture identification from uploaded images

Enhanced Data Mapping and Retrieval

• Matching of detected objects with MyPD terminology
• Retrieval of relevant MyPD records based on predefined criteria

Streamlined Quote Creation and Output

• Selection of matching historical records for quote generation
• PDF download capability for generated quotes
• Email attachment capability for generated quotes

Monitoring and Tracking

• Logging of quote generation and email activity
• Activity filtering through a log-view screen

Improved Platform Security

• Platform readiness to accommodate multiple AI agents for PD Instore
• Secure application login through Active Directory authentication

With AI-powered image analysis, streamlined data mapping, and accelerated draft quote creation, Evoke demonstrated how intelligent automation can transform a manual, fragmented quoting process into a scalable, consistent, and insight-driven capability — enabling faster response times, improved accuracy, and greater confidence in decision-making