Modern technology stacks, combining automation platforms, flexible databases, and generative AI, are revolutionizing customer acquisition processes in complex B2B industries. They enable the creation of highly personalized yet fully automated customer service systems that do not replace human experts but enhance their strategic competencies. The following case study provides a detailed analysis of the transformation of a construction company that, thanks to such an implementation, moved from operational chaos to intelligent process orchestration.
Initial State Diagnosis – The Hidden Costs of an Unstructured Process
The initial situation in the analyzed construction company was typical for many enterprises in the B2B services sector. The inquiry handling process, although crucial, was characterized by operational chaos. Inquiries from contact forms, emails, and phone calls were handled in isolation, making it impossible to create a coherent sales funnel. There was no central repository, and information was scattered across email inboxes and employee notes, which prevented real-time answers to basic business questions.
Each new inquiry triggered a sequence of repetitive, manual tasks. A specialist had to personally read the message, assess its potential, and manually extract key data, such as the investment's location or its area. This was not only time-consuming but also prone to errors. The fundamental problem was the lack of data structure at the source – unstructured text was the cause of a whole cascade of inefficiencies. Therefore, the key to solving the problem was not writing emails faster, but automating the conversion of unstructured data into structured metadata at the very beginning of the process.
Designing the Solution – Agile and Intelligent Platform Architecture
Instead of building a monolithic system from scratch, the company chose an agile approach, which involved connecting specialized tools that communicate via APIs. Such an agile and intelligent platform architecture ensures faster implementation, lower cost, and unparalleled flexibility. The heart of the system became a carefully selected technology stack, in which each element played a precisely defined role.
How does the automated inquiry handling process work?
Inquiry Initiation
The client sends an inquiry through the website form or directly to the email inbox.
AI Capture and Classification
The system automatically captures the data. An AI model (e.g., Gemini) classifies the inquiry and extracts key information (location, area).
Data Saving and Enrichment
All data is sent to a central database in Airtable. The system automatically calculates the distance from the company's headquarters (Google Maps API).
Intelligent Question Generation
AI creates personalized, clarifying questions to gather the complete information needed to prepare an offer.
Specialist Verification (Human-in-the-loop)
The expert reviews, edits, and approves the message content in Airtable, maintaining full control over communication.
Message Preparation and Sending
The system creates a ready-to-send email draft in Gmail. The status in Airtable is updated automatically.
Client Response Analysis
After receiving a response, AI analyzes its content and automatically updates the data in Airtable.
Personalized Offer Generation
After the sales data is completed, the system automatically generates a professional offer in Google Slides and PDF.
Finalization and Analytics
A ready email draft with the offer awaits sending. Data from the entire process is used for sales funnel analysis and optimization.
Make.com became the process orchestrator, controlling the data flow. Airtable took on the role of a flexible database and user interface. The API of an LLM model (e.g., Gemini 1.5 Pro) was used as the operational brain for classifying inquiries, extracting data, and generating content. The system was complemented by the Google Workspace ecosystem for finalizing communication and a dedicated Google Apps Script application, allowing non-standard email messages to be included in the process with a single click.
| Component | Role in the Process | Key Function / Rationale |
|---|---|---|
| Contact Form / Gmail / Other (e.g., phone) | Input Gateway | Capturing inquiries from the standard website form and non-standard emails. Ability to manually enter data (e.g., from a phone call). |
| Make.com | Process Orchestrator | Visual business logic building, no-code API integration, webhook handling. |
| LLM Model (e.g., Gemini 1.5 Flash/Pro) | Analytical-Generative Engine | Classification, data extraction, and content generation. The choice of model depends on the client's preferences. |
| Airtable | Central Database / Interface | Storing structured data, managing statuses, "human-in-the-loop" interface. |
| Google Apps Script for Gmail | Manual Process Initiator | Allows inclusion of non-standard email inquiries that were not captured automatically into the workflow. |
| Google Maps API | Auxiliary Service | Automatic calculation of distance from the company's headquarters based on the address. |
| Google Workspace (Gmail, Slides, Drive) | Finalization and Delivery Platform | Generating email drafts, creating offers based on templates, archiving files. |
The Inquiry's Journey Through the System – A Detailed Analysis of the Automated Workflow
The implemented system transformed chaos into a precisely orchestrated journey. In the first phase (Capture and Qualification), a new inquiry from the form or a manually indicated email goes to Make.com. Then, the LLM model classifies the inquiry in terms of business relevance and extracts key metadata (area, location). Simultaneously, the Google Maps API calculates the distance. The complete information is saved as a new record in Airtable.
In the second phase (Interactive Communication), the system, based on the investment type, generates a set of intelligent clarifying questions with the help of AI. These questions appear in Airtable, where a specialist verifies and approves them. This action automatically creates a message draft in Gmail. The time to prepare a response is reduced to a minimum.
In the third phase (Response and Finalization), the client's response is again analyzed by AI to extract additional data, which completes the record in Airtable. After filling in the commercial data, the specialist changes the status to "Generate Offer". The system automatically creates a personalized presentation in Google Slides based on a template, exports it to PDF, and prepares a ready-to-send email draft with the offer and attachment.
Human-in-the-Loop – The New Role of the Specialist
The fundamental philosophy of the implementation was not to replace the specialist, but to augment them. The Human-in-the-Loop approach was implemented by deliberately designing checkpoints where human intervention is required. As a result, the specialist, relieved of repetitive tasks, could focus on building relationships, negotiations, and strategic thinking.
- Verification of AI suggestions: The specialist always has the final say and can edit any content generated by AI before sending it to the client.
- Enriching data with qualitative context: A human brings invaluable context, e.g., notes from phone calls, which enrich the project profile with nuances unavailable to AI.
- Final commercial decision: Key business decisions, such as setting the final price, remain the exclusive responsibility of the expert, who receives a complete set of data from the system to make an informed decision.
As a result, a specialist who previously spent 80% of their time on administrative work could reverse these proportions, which translated into increased job satisfaction and higher quality of customer service.
From Data to Knowledge – The Continuous Improvement Loop
The system not only automates work but also generates invaluable, structured data. This allowed for the creation of an analytical report that transformed raw data into strategic knowledge, enabling the company to transition to a culture of Data-Driven Decision Making. The analysis of metrics triggered a flywheel effect: the process generates data, analysis leads to better decisions, and these optimize the process, which in turn generates even better data.
| Metric (KPI) | Data Source | Strategic Question It Answers |
|---|---|---|
| Average first response time | Time difference between record creation and status change to "Awaiting Response" | How does response speed affect the conversion rate? Should we invest in shortening this time? |
| Conversion rate by inquiry type | Ratio of won deals to all inquiries, filtered by "Investment Type" | Which market segments are most profitable for us? Where should we direct our marketing efforts? |
| Correlation of price per m² with customer decision | Analysis of won/lost deals in the context of the proposed price | Is our pricing strategy optimal? At what price point do we lose the most deals? |
| Process bottlenecks | Average time an inquiry spends in each status (e.g., "Awaiting Offer") | Which stage of the process is the longest? Where should we look for further optimizations? |
| Rejection rate at the qualification stage | Ratio of inquiries marked as "is_relevant: false" to the total | Are our marketing campaigns attracting the right clients? Are the qualification criteria well-defined? |
Summary of Benefits – A Measurable Business Transformation
The implementation of the platform brought the company a range of measurable benefits. The inquiry handling time was reduced from hours to minutes, resulting in radical operational efficiency. The system ensured standardization and professionalism in communication, and the creation of a central database in Airtable eliminated the problem of information chaos. The customer experience (CX) also improved thanks to fast and personalized responses. Most importantly, the company gained access to strategic business intelligence and a flexible, scalable architecture ready for future growth.
Next Steps and Future Vision – From Automation to Prediction
The described implementation is just the beginning. Short-term plans include integration with messengers and automation of follow-ups. In the medium term, the goal is to integrate with a project management system and implement dynamic pricing. The ultimate vision is to move from automation to predictive analytics. The collected dataset will be used to build a machine learning model that can predict which inquiry has the highest chance of conversion (Predictive Lead Scoring), allowing for intelligent work prioritization.
Conclusion: Investing in Process Intelligence as a Competitive Advantage
This case study shows that the combination of agile platforms, flexible databases, and the power of AI can solve fundamental operational problems in service industries. The described implementation is a practical, achievable, and relatively inexpensive solution that any company can adapt. Investing in this type of platform is not a cost, but a strategic investment in process intelligence – the organization's ability to understand, optimize, and automate its key processes. In today's world, it is this intelligence that becomes the key factor in building a sustainable competitive advantage.
Frequently Asked Questions (FAQ)
Not necessarily. The cost depends on the scale, but an architecture based on no-code/low-code tools (like Make.com and Airtable) and pay-per-use AI APIs is significantly lower than building a custom system from scratch. Implementation requires expert knowledge, but it is much faster and more flexible than traditional IT projects.
Absolutely not. On the contrary, the system is designed to enhance the competencies of specialists. It automates repetitive, analytical tasks, allowing people to focus on what they do best: building customer relationships, negotiations, strategic thinking, and making key business decisions.
The choice depends on specific needs and budget. Models like Gemini 1.5 Pro offer very high-quality text analysis and generation for complex tasks. For simpler, high-volume operations (e.g., initial classification), faster and cheaper models like Gemini 1.5 Flash may be perfectly sufficient. The key is the flexibility of the system, which allows for changing the model in the future.
The implementation time depends on the complexity of the company's processes. However, thanks to the agile approach and the use of no-code platforms, the first working prototypes (MVP - Minimum Viable Product) can be launched within a few weeks, not months or years, as is the case with traditional development.
Yes. The described architecture and logic are universal for any B2B industry where the sales process is based on handling complex, unstructured inquiries. It will work perfectly for consulting firms, marketing agencies, law firms, IT companies, or design offices.
The biggest challenge is often organizational rather than technological. The key is to involve future users (specialists) in the design process and show them that the system is a tool to support, not replace, their work. Building trust in automation and AI is the foundation for the success of the entire project.