Implementation of AI Agent in Customer Service for Anda Śruby

About our client

Anda Śruby is a company specializing in the sale of fasteners for B2B clients, handling a large number of quotation requests. As sales grew, there was a need to streamline inquiry handling and shorten response times. The company was looking for an agency that would implement an AI agent to manage quotation requests and integrate it with sales processes and CRM, to alleviate the workload of the sales team.

Key challenges faced by our client

Too long response time for quotes

The process of handling quotation requests was largely manual, which significantly extended response times. As a result, some clients opted to purchase from competitors without waiting for a quote to be prepared. The lack of a prompt response therefore directly impacted the loss of potential revenue.

Time-consuming manual data entry in CRM

Every inquiry required manual checking to see if the client already existed in HubSpot, and then filling in or updating their information in the CRM. This process was repetitive and consumed a significant portion of the sales team's time, limiting their ability to focus on actual customer service.

Long analysis of inquiries and product availability

Due to the industry's nature, Anda Śruby's customer inquiries often include long product lists in attachments. Before automation, sales representatives had to manually review items, extract key data, and check the availability of each product in inventory systems, which was very time-consuming.

Incomplete information on sales progress

Updating the CRM with information from phone calls was cumbersome for sales reps, which is why it was often neglected in practice. Call notes made it into the system very irregularly or not at all. As a result, managers lacked visibility into how sales calls were conducted.

To streamline sales at Anda Śruby, we implemented an AI system featuring

Automatic email monitoring and message filtering

We've implemented customer service automation integrated with our corporate email inbox. The system monitors all incoming messages, and an AI module classifies them for spam and determines if a message is a sales inquiry. Non-sales emails are marked accordingly and excluded from further processing.

AI Agent for client verification and lead scoring

Upon detecting a tender inquiry, the AI agent checks if the company already exists in HubSpot. If not, it automatically searches for company data and then creates a note in the CRM with a description of the client's business, an initial lead assessment, and recommendations for sales representatives. For existing clients, the system analyzes sales history and assesses the potential of the current inquiry.

Analysis of attachments using AI and OCR

Email attachments are automatically downloaded and analyzed using OCR technology (Google Vision API). The system rejects technically irrelevant files and notifies sales representatives about attachments requiring manual verification. This is a crucial element of AI utilization in Anda Śruby's customer service.

Generating product lists by an AI agent

Inquiry content and data from attachments are sent to an AI assistant. Following predefined instructions, it generates a list of desired products, including technical parameters and features specified by the client. This information is then fed into HubSpot and our custom-built offer panel in Airtable.

Automatic transcription and analysis of sales calls

Sales calls are now automatically recorded, transcribed, and aggregated thanks to integration with T-Mobile. This enables the AI agent to assign the transcription to the correct lead in HubSpot. Historical calls are also available to managers in a dedicated sales analysis panel.

Automated Offer Generation Panel

The final step is an Airtable panel where sales representatives verify AI-identified products and optionally make corrections. Within the system, they can resubmit unmatched products to the AI, incorporating additional discussions to clarify client requirements. Once the product list is approved, a proposal file is generated.