Dumpling AI — AI Chat-Agent with GPT-4o for Automatically Saving Local Business Data to Airtable

What is the biggest headache for small businesses today? It is that you are doing ok selling things and have customers, then you spend most of the time on nonsense like transferring customers' contact information from Telegram into an Excel Spreadsheet. This is crazy! The owner of a Moscow coffee shop spends 3 hours a day just moving orders from Telegram into Excel. 3 hours! He could be launching new menu items or taking some rest instead of wasting his time doing so!

Dumpling AI was created to help this type of situation! It is a chatbot powered by AI using GPT-4o that gets the customer info from the chat and puts it into Airtable. Plus, you can set this system up without any programming skills in less than 1 evening. Simply attach the pre-built workflows and you're done! It will read the text of your order and then automatically read what field to put you in, who ordered from you, and when to deliver it.

To sum it up, you receive hours of routine savings, clean data input, and more peaceful nights.

"One intelligent AI chatbot can take the place of a few employees, reduce the number of errors, and enable you to work."

Let's look at the process of how it works, how you will save money with it, and how to get started with it!

What is Dumpling AI and Its Key Features

Dumpling AI is an AI chatbot that works directly with your client in the Telegram App, WhatsApp, or WebChat. An agent's primary function is to convert an unstructured message into a structured format. Below is a list of agent features:

• Automate Data Collection: An agent automatically identifies contacts (Name, Phone, and Email) as well as the details about their order (Product, Quantity, Address, and Requests). With the agent, you no longer have to manually copy and paste into Excel.
• Integration with GPT-4o: The OpenAI model uses Natural Speech and understands the meaning of unstructured messages. For example, if you wrote to say, "I want to have 2 kg of apple delivered to Lenina St. 15, contact me after 18:00." Dumpling AI detects what product, quantity, address, and time you requested, while saving this data into different fields in Airtable.
• Supporting Local Small Businesses: If you run a coffee shop, car wash, courier service, or home repair business, then you can use Dumpling AI to support your business needs without having to hire an IT department. Rather than spending hundreds of dollars on overpriced CRMs, you can purchase a subscription to Airtable, then use it in conjunction with pre-built workflows from ASCN.AI.

Real-world Example: A car wash located in the Moscow region processed 10 requests per day. The secretary spent approximately 15–20 minutes processing the requests. It was common for the car wash to misplace request numbers and/or addresses. Since implementing Dumpling AI, clients simply send requests through Telegram. The agent now recognizes the date, time, phone number, and automatically creates an Airtable entry with the appropriate fields. This gives the secretary more time to work with clients and has helped increase the number of requests received by 40% without having to hire additional staff.

What GPT-4o Does in Dumpling AI

GPT-4o is a multimodal language model that can process both text and images/audio. In Dumpling AI, GPT-4o performs two primary functions:

• Natural Language Processing (NLP): it "understands" messages and is able to extract pertinent information out of the message.
• Improved accuracy: it provides structured data to the agents, saving time and reducing errors.

An example of how GPT-4o improves accuracy:

Message: "Hi! Need 3 cakes for Friday, delivery to Nevsky 10, paid by card."

Parsed Airtable Entry: Product (Cakes), Quantity (3), Date (Friday), Delivery Address (Nevsky 10), Payment Method (Card).

• Correcting Typos: For example, the word "phonne" is a typographical error for the word "phone." GPT-4o will correct any typographical errors it encounters when parsing the data.
• Normalizing Numbers: When GPT-4o parses a number, it will convert it to its standard format (e.g., "8 (916) 123-45-67" → "+79161234567"). Additionally, GPT-4o will convert conversational phrases to dates/times (e.g., "This Friday" = "Friday at 10 AM").

Automated parsing through the OpenAI API works as follows — GPT-4o extracts the client's name, phone number, delivery address, and delivery date and time, returning them as JSON:

Fields: client_name, client_phone_number, client_delivery_address, client_delivery_time
If any field is missing — those fields will be left null.

A message processed by GPT-4o takes between one and three seconds to process through the OpenAI API, and once the parsing process is complete, the parsed data will be immediately available to the user through Airtable.

There is a primary distinction between the traditional parsing process and the parsing process that takes place through the OpenAI API: GPT-4o gives consideration to the context of the dialogue. For instance, if the client decides that he/she no longer wants to order cakes, the GPT-4o agent will understand that the order has been canceled, and therefore, the data will not clutter the database with additional orders.

Automatic Collection of Local Business Data

What Type of Data is Available for Local Businesses?

Contact Information: name, last name, phone, email, and Messenger ID(s). GPT-4o will standardize numbers, e.g. convert "8 (916) 123-45-67" to "+79161234567."

Reviews and Feedback: All reviews and feedback will be analyzed by the GPT-4o system according to their sentiment (positive, negative, neutral), and any negative reviews or complaints will be highlighted. The results of this analysis will be placed in a separate table within Airtable.

Example: A bakery in Yekaterinburg received 150 reviews in two months, and as a result of analyzing those reviews, they identified a problem with stale bread being delivered on Fridays and adjusted their schedule; as a result, their negative review percentage dropped from 12% to 3%.

Sales Tracking includes details such as date ordered, items ordered, total order amount, how the order was paid for and how it was delivered. For example, if someone were to say "I'm ordering two bouquets of roses for 1500 rubles each, I'm using a courier to deliver them to me, and I will pay cash on delivery", that message would then become a new entry into the system that includes the product name, price, and payment type. The flower shop owner would be able to quickly see how sales have changed over time and also see what items are most popular, without having to do any additional work.

Advantages to Automatic Data Capture

1) Speed of processing data. Manually entering an application takes between 1.5–2 minutes per entry. By comparison, Dumpling AI can handle this in only 3–5 seconds per entry. Assuming that a user would receive 50 applications per day, that saves approximately 90 minutes per day.

By utilizing the time that was saved for additional sales consultations, a nail studio located in Novosibirsk was able to increase their up-sell percentage by 18%.

2) Reduced Errors. When manually processing data there is approximately a 5–8% chance of experiencing typographical errors and inaccuracies. By using Dumpling AI, this number has dropped to between 1–2%. Additionally, with GPT-4o providing checks on both phone and date formats, it helps to identify mistakes that would have otherwise gone unnoticed.

3) Improved Analytic Capabilities. Because data has been structured into databases, it's now possible to generate reports directly within Airtable, without having to spend time exporting data. A few examples:

• 40% of orders originated from one specific neighborhood — branching out to that location reduced the delivery time for customers from 45 minutes to 20 minutes.
• 60% of customer inquiries were generated between 18:00 and 21:00 — introducing evening shifts resulted in a 25% increase in total load for those hours.
• Personalized offers extended to 30 customers resulted in increased retention from 22% to 34%.

4) Cost. Business process automation has the potential to lower operating costs of SMBs by 30–50% (McKinsey Digital Report 2022). A company with 500,000 roubles of revenue a month (average 15,000–25,000 roubles saved) will have saved enough to cover the subscription for both ASCN.AI and OpenAI by integrating Dumpling AI with Airtable.

Airtable Integration

Airtable is a cloud-based database that has an open API structure. The database allows users to save up to 1,200 records and provides several ways to view records, including the calendar, kanban and table.

To set up an Airtable integration, follow these steps:

1. Create a base and then create a table within this base with the following fields: Name, Phone, Email, Address, Order Date, and Source.
2. Obtain a Personal Access Token for API use from the Airtable Settings (Account → API) — keep this token private.
3. Find your Base ID and Table ID using either the base URL or the Airtable API documentation.
4. In ASCN.AI, set up the HTTP Request Node as follows:
   URL: https://api.airtable.com/v0/{Base_ID}/{Table_Name}
   Method: POST
   Headers: Authorization: Bearer {{$secrets.airtable_token}}, Content-Type: application/json
   Body (JSON): fields from AI Agent — Name, Phone and others.
5. Activate the workflow and test it by adding a few sample entries to Airtable.

⚠️ If there are any issues: 401 means the token is incorrect; 404 means there is an incorrect Base or Table ID; 422 refers to a field name mismatch.

Example: A courier service in Kazan reduced the amount of time it takes to process 20–30 applications from 40 minutes to just 5 minutes — improving the speed and quality of delivery service.

Technical Specifications

To access the integration, you will need:

• A computer or smartphone with internet access and a browser (minimum 2 Mbit/s)
• An ASCN.AI account (starting at 2,000 ₽ per month)
• An Airtable account (either free or paid)
• An OpenAI account with enough funds to process GPT-4o requests — ranging from $0.002 to $0.005 per request; for 100 daily requests approximately $15–20 per month

A detailed tutorial on registration, token acquisition, node configuration, and workflow testing can be found on the ASCN.AI technical documentation page.

Examples of Automated Use Cases Using Dumpling AI for Business

Bakery in St. Petersburg: Previously, the manager was entering orders into an Excel file for 2 hours a day while frequently inputting the incorrect date or requesting multiple products on the same order, generating an approximate 8% error rate. With Dumpling AI, the time spent processing orders is reduced to 15 minutes, and the error rate has been reduced to less than 1%. Orders have increased by approximately 1/3, allowing the manager additional time to reach out to existing clients and generate repeat business.

Car Service in Nizhny Novgorod: Appointments were managed through Google Sheets, leading to multiple instances of conflicting appointments. With Dumpling AI, the service now recognises the service requested, the date and time of the appointment, displays available appointments and creates calendar entries, therefore eliminating duplicate appointments. Conflicting appointments are now at zero and the technicians have been able to increase their capacity by 20%, while the overall number of cars serviced has increased by approximately 40%.

Yoga Studio in Moscow: The administrator was manually monitoring and tracking the remaining sessions for each client; this took about 3–5 minutes per client and caused a backlog during peak hours. With Dumpling AI, the system now automatically deducts sessions as they expire, and notifies clients of their need to extend their sessions, speeding up processing and growing revenue by around 22%.

User Testimonials

"I spent half a day on orders just to process them. With Dumpling AI, all applications now go directly into Airtable and I just check them. I put that time I freed up into Instagram and gained about 1/3 more new clients."

"We integrated the agent with WhatsApp and Airtable together in one evening. The implementation is straightforward, all instructions are presented clearly. Clients receive confirmations without any delays due to conflicting schedules."

Efficiency and Business Metrics Impact

Speed of Response — A client would previously wait between 5 and 30 minutes for your response; with Dumpling, clients receive their response between 3 and 5 seconds. As noted in Harvard Business Review (2011), if you respond within five minutes, it is nine times more likely your client will close versus responding after 30 minutes.

Freedom to Scale Without Hiring — Previously, as the volume of requests increased, organizations would hire a second administrator for approximately 40,000–60,000 ₽ per month. Now, using Dumpling, one employee can manage up to 150 requests per day and save up to 50,000 ₽.

Decrease in Errors — The percentage of errors for manual entry ranged from 5 to 8 percent, whereas through the use of the validation function of GPT-4o, the percentage of errors for Dumpling AI has dropped to between 1 and 2 percent.

Improved Transparency of Operations — The owner can see every submitted application via Airtable in real-time. There is no longer any need to wait on employee reports to determine status. Application logs and workflow history are always accessible.

Manual Entry Versus Dumpling AI with Airtable: Comparison

Frequently Asked Questions (FAQs)

Is it Compatible with Other Platforms?
Yes! Dumpling integrates with Google Sheets, Notion, Monday.com, and other various CRM platforms using the HTTP API.

How Do I Integrate With WhatsApp Business?
To integrate, you must obtain official access to the WhatsApp Business API and receive approval from Meta.

What do the error codes 401, 404 and 422 mean?
401 means an invalid or incorrectly formatted access token; 404 means an incorrect base ID or table ID; and 422 means the fields in Airtable do not match up properly. You can fix these errors by verifying your API parameters and access permissions.

Do I have to be a programmer to connect with this service?
No. All configuration for your account(s) will be performed through ASCN.AI's user interface and will only require that you provide the necessary access tokens and identifiers.

Conclusion and Next Steps

Why You Should Use This Workflow Now:

• Fast responses to client requests create competitive advantages.
• Increased efficiency by reducing the amount of time spent on administrative tasks.
• Having one source of structured data makes it easier to find, analyze and manage your data.
• Scalability without needing to hire additional developers or engineers to grow your business.

According to a forecast by Deloitte (2024), the bulk of the time spent on routine accounting operations will be automated by 2027. Integrating this system into your workflow is going to be a large-scale project and it's important that you select a trustworthy partner who has experience integrating these types of systems and has a solid support structure in place.

Comparison Table of Major AI Tools Used to Work with Airtable

Expert Insights on Future Use of AI in Workflow Automation and Data Analysis

"In three years 'creating reports' by hand will become as foreign to us as searching for static paper directories is to using Google. AI agents are becoming the foundation for all workflows. They will not only save users time, but will also create a serious advantage for those individuals and companies implementing AI in the foreseeable future. Companies implementing the use of AI now will eventually evolve their operational and logistical models, allowing them to function at an extraordinary level of efficiency."

User Testimonials and Use Case Scenarios Around the Use of AI Agents with Airtable

Example Use Case: Cryptocurrency Arbitrage on Airtable

The team at arbitragescanner.io uses Airtable to keep track of price differences across exchanges. Historically, this was done by human analysts every 15 minutes; now the work is completed by a bot. With the use of the bot, automated tracking of prices takes place. When the price spread rises above 2%, the bot sends an alert to the Telegram channel. The speed of reaction increased to 10 times faster than before, resulting in a 35% increase in arbitrage profits.

Example: Flash Crash — October 11, 2024

On the night of the flash crash, the crypto market dropped by 20% in 30 minutes. Customers of ASCN.AI received alerts through their AI agents immediately after the initial crash. Those who were able to react quickly were able to exit the position before losing any money (and some made money), while those who depended on manually monitoring the price lost out on time and subsequent monetary loss.

"In the past I would spend 20 minutes each morning creating a Sales Summary Report. Today, I simply receive a Sales Summary Report via Telegram at 9:00 AM daily. I now have more time to spend developing and executing my Strategic Plan instead of wasting this time creating and preparing a Sales Summary Report." 

Using an AI agent via ASCN.AI is extremely user-friendly for anyone with a basic or limited understanding of technology. Users simply need to take advantage of an existing tracking template for arbitrage, connect it to their Airtable and automate the monitoring of price differences between exchanges. When the price spread exceeds 2%, the bot sends an alert via Telegram. Set up takes less than 15 minutes and initial clients have been earning from $500 to $3,000 per month, depending on the trading volume and frequency of the alerts.