Automate customer information gathering with a WhatsApp AI chatbot

The problem everyone knows (and no one solves)

How much time do you spend each week chasing your clients for missing documents?

Two hours? Five hours? Some of the brokers I work with told me ten hours. Ten hours a week sending emails like "Hi, I still haven't received your latest payslip," calling clients who don't pick up, manually checking whether a file is complete.

The problem isn't that your clients are unwilling. It's that the process is painful for everyone. You send an email with a list of 12 documents to provide. The client opens it three days later, on their phone, between two appointments. They don't have time to do it all at once. They send 3 of the 12. You follow up. They send 4 more, but they forgot to sign a form. You follow up again. The file takes two weeks to complete. Sometimes three.

Meanwhile, you're not doing your real job. You're doing admin.

There's an alternative. A WhatsApp chatbot powered by artificial intelligence that guides your clients step by step, collects the information and documents, validates in real time, and automatically follows up with those who haven't finished. All on WhatsApp, with a 98% open rate -- compared to 21% for email.

This isn't science fiction. It's a tool that brokers, accounting firms and real estate agencies are already using every day.

What a collection chatbot actually does

Forget the basic chatbots that reply "our hours are 9am-6pm." Here, we're talking about an AI assistant specialized in a single goal: getting all the information and documents you need from your client, as quickly as possible, with no friction.

Concretely, here's how it works.

You're onboarding a new client. Instead of sending them an email with a PDF to fill in or a link to a 3-page web form, you send them a WhatsApp message: "Hi Marie, welcome to [your company]. I'll guide you through putting your file together. Shall we start?"

From there, the chatbot takes over. It asks the questions one at a time, in a logical order. It adapts the following questions based on previous answers. If Marie says she's a salaried employee, the chatbot asks for her payslips. If she's self-employed, it asks for her financial statements. No pointless questions, no generic form.

The chatbot receives photos and PDFs directly within the conversation. Marie takes a photo of her ID from her couch, the chatbot receives it, checks that the image is legible, and moves on to the next question. If the document is blurry or unreadable, it says so immediately and asks her to resend the photo.

If Marie stops halfway through the conversation -- because she's at work, because her child is crying, because that's life -- the chatbot picks up exactly where she left off when she comes back. No need to start over.

And if Marie doesn't come back? The chatbot sends an automatic follow-up 24 hours later. Then another 48 hours after that. Politely, naturally, without sounding like a robot.

On the professional side, all the collected data is centralized in a dashboard or pushed straight into your CRM. You see at a glance which files are complete, which are in progress, and which are stuck. No more digging through 47 emails to find a client's bank details.

To understand the technical architecture behind this kind of solution, check out our guide on building a WhatsApp AI assistant with Amazon Bedrock.

5 sectors where it changes everything (and banks aren't one of them)

Let's be clear from the start: this kind of solution is aimed at SMEs and independent professionals, not at big banks that have multi-million IT budgets and IT services firms to run their projects. Here are five sectors where a WhatsApp collection chatbot has an immediate, measurable impact.

1. Real estate and mortgage brokerage

The problem: A mortgage broker spends, on average, 2 hours per file collecting the client's information before even starting to work on it. Income, employment status, down payment, expenses, supporting documents... The list is long, and clients systematically forget half the documents.

What the chatbot collects:

• Employment status (salaried, self-employed, civil servant)
• Net monthly income and bonuses
• Available down payment
• Current expenses (rent, existing loans)
• Last 3 payslips (photo or PDF)
• Latest tax assessment
• Bank statements from the last 3 months
• ID document

The measurable gain: The chatbot pre-qualifies the file automatically. Even before the first appointment, the broker knows whether the client can be financed and at roughly what rate. Qualification time cut by 4, number of files handled multiplied by 3.

2. Accounting firms

The problem: Onboarding a new client is an administrative nightmare. Company registration number, tax regime, projected revenue, bank details, engagement letters to sign... And every quarter, here we go again: you have to collect invoices, statements, supporting documents.

What the chatbot collects:

• Company registration number (SIRET) and legal form
• Tax regime (micro, simplified, standard)
• Projected revenue
• Business bank details
• Director's ID document
• On an ongoing basis: purchase invoices, sales invoices, monthly bank statements

The measurable gain: Onboarding goes from 5 days on average (emails, follow-ups, missing documents) to 48 hours. The quarterly collection, which used to tie up an assistant for a week, now happens continuously and automatically. Clients send their invoices as they go, directly into the WhatsApp conversation.

3. Property management and real estate agencies

The problem: Putting together a complete tenant file takes 3 days on average. The applicant sends their documents in dribs and drabs. Something is always missing. And when you're handling 20 applications in parallel for the same property, it's chaos.

What the chatbot collects:

• ID document
• Last 3 payslips
• Employment contract or employer's certificate
• Latest tax assessment
• Proof of current address
• Guarantor's information (same documents)
• Bank details (RIB)

The measurable gain: A complete tenant file collected in 10 minutes instead of 3 days. Incomplete files are flagged automatically. The real estate agent can compare applications based on 100% complete files, without spending hours chasing.

4. Training organizations

The problem: Enrolling a trainee involves information from three different parties: the trainee themselves, their company, and the funding body (OPCO, CPF). The back-and-forth is endless, and an incomplete file blocks the start of the training.

What the chatbot collects:

• Trainee's personal information (identity, contact details)
• Company information (SIRET, address, HR contact)
• OPCO funding approval number or CPF number
• Signed training agreement
• Employer's certificate
• Copy of the ID document

The measurable gain: Enrollment time reduced from 2 weeks to 3 days. The chatbot automatically follows up with each stakeholder (trainee, HR, OPCO) until all documents are received. Fewer training sessions postponed because of an incomplete file.

5. Recruitment and temp staffing

The problem: A candidate validated in the interview, but you still have to collect a mountain of documents before placing them on an assignment: ID document, driver's license, certifications, diplomas, bank details, social security certificate... And the candidate is already replying to 5 other agencies.

What the chatbot collects:

• Up-to-date CV
• ID document, front and back
• Driver's license (if required)
• Certifications and accreditations (forklift license, electrical certification, etc.)
• Diplomas
• Bank details (RIB)
• Social security certificate
• Availability and salary expectations

The measurable gain: The candidate completes their file in 15 minutes from their phone, the very evening of the interview. The agency can place them on an assignment the next day instead of waiting 5 days for the documents to arrive by email.

To go further on automating customer relations via WhatsApp, read our dedicated article.

How it works technically (without the jargon)

You don't need to be an engineer to understand it. Here are the main building blocks.

WhatsApp Business API -- This is WhatsApp's official interface for businesses. It lets you send and receive messages programmatically. Unlike the free WhatsApp Business app, the API lets you automate conversations at scale.

A language model (LLM) -- This is the chatbot's brain. We use models like Claude from Anthropic or GPT-4 from OpenAI, accessible via Amazon Bedrock. The LLM understands the client's natural language, generates appropriate responses, and decides which question to ask next based on previous answers. It's not a fixed decision tree: it's a real conversation.

Secure storage -- The documents sent by clients (photos, PDFs) are stored in a secure, encrypted space, hosted in the European Union. Each document is linked to the client's file and accessible from your dashboard.

Connectors to your tools -- The chatbot can push the collected data directly into your CRM (HubSpot, Salesforce, Pipedrive...), your business software, or simply into a Google Sheets or Notion table. No double entry.

A follow-up engine -- If the client hasn't finished their collection, the system sends automatic reminders at configured intervals. The messages are natural, personalized, and comply with WhatsApp's rules (no spam, mandatory opt-in).

On the GDPR compliance side, the data is hosted in the European Union, consent is explicit (mandatory opt-in before any exchange), the client can request the deletion of their data at any time, and the documents are encrypted both at rest and in transit.

For the technical details on the Amazon Bedrock and AWS Lambda architecture that powers this kind of solution, check out our complete technical guide.

Why WhatsApp and not a web form?

A fair question. You already have a form on your site, maybe even a client portal. Why add WhatsApp?

Because your clients don't fill in your forms. Or they fill them in badly, half-way, three weeks later.

Here's the concrete comparison:

The most important point: WhatsApp is already installed on your client's phone. They have nothing to download, no account to create, no password to remember. They reply from their couch, on the subway, during their lunch break. The friction is almost zero.

The other major advantage: document photos. Asking a client to scan a payslip, save it as a PDF, log in to their client portal, and upload it... that's asking them for 10 minutes of effort. On WhatsApp, they take a photo and send it. Two seconds.

Finally, the follow-up. A follow-up email has about a 5% chance of being opened (clients filter, ignore, forget). A WhatsApp message is read within 3 minutes on average. Your file completion rate goes from 40-50% to over 85%.

How much it costs

Let's be transparent about pricing. As with any AI chatbot project, the cost depends on complexity. Here are the real ranges.

Simple collection (5 to 10 fields, no AI)

A rule-based chatbot that asks questions in a fixed order and stores the answers. No natural language understanding, no dynamic adaptation.

Build cost: 1,000 to 2,500 euros

Monthly cost: 50 to 100 euros/month (hosting + WhatsApp API)

Ideal for: basic collection with few fields and no documents to receive.

AI collection with validation and automatic follow-ups

The chatbot understands natural language, adapts its questions, receives documents, validates the information in real time, and automatically follows up with clients who haven't finished.

Build cost: 3,000 to 8,000 euros

Monthly cost: 100 to 200 euros/month (hosting + WhatsApp API + LLM calls)

Ideal for: brokers, real estate agencies, training organizations, accounting firms.

AI collection with CRM and business tool integration

Everything above, plus direct integration with your tools: CRM, business software, client portal, document management system. The collected data arrives directly where you need it.

Build cost: 5,000 to 15,000 euros

Monthly cost: 150 to 300 euros/month (hosting + WhatsApp API + LLM + connector maintenance)

Ideal for: companies with structured processes and a high volume of files to handle.

For a detailed breakdown of pricing by AI chatbot type, check out our complete article on costs.

A concrete example: a mortgage broker who cut his qualification time by 4

To make all this tangible, here's a real case.

Marc is a mortgage broker in Lyon. His firm handles around 30 files a month. Before, the process of qualifying a new client looked like this:

• First phone call to understand the project (15 minutes)
• Email with the list of documents to provide (ID document, last 3 payslips, tax assessment, bank statements, proof of down payment)
• Waiting... The client takes 5 days on average to send all the documents
• 2 to 3 follow-ups by email and phone
• Manual check of the file's completeness
• Second call for the missing information
• Total time per client: around 2 hours, spread over 7 to 10 days

With 30 files a month, Marc was spending 60 hours -- nearly 8 full days -- on information gathering alone. Not on advice, not on negotiating with banks. On pure admin.

Today, Marc uses a WhatsApp AI chatbot. Here's the new process:

• After the first call, Marc sends the client a WhatsApp link
• The chatbot takes over and guides the client step by step
• It asks the questions in a logical order: employment status, income, expenses, down payment
• It asks for the documents one by one, with clear instructions ("Take a photo of your latest payslip, front and back")
• It checks that each document is legible and complete
• If the client stops, it automatically follows up the next day
• Once the file is complete, Marc gets a notification with a structured summary and all the documents sorted

The results, three months after deployment:

• Qualification time per client: 30 minutes instead of 2 hours
• File completion time: 24 hours instead of 7 days
• Number of files handled per week: 10 instead of 7
• File completion rate: 92% instead of 65%
• Client satisfaction: clients find the process "simple" and "fast" -- some complete their file that very evening from their couch

The return on investment? The chatbot cost 6,000 euros to deploy and 180 euros a month to run. Marc estimates it saves him 40 hours a month. At 80 euros an hour (the value of a broker's time), that's 3,200 euros in monthly savings. The chatbot paid for itself in less than two months.

What it changes for your business

Beyond the numbers, the real change is qualitative. When you no longer spend your days chasing clients for documents, you do your real job. The broker negotiates with banks. The accountant advises their clients. The real estate agent shows properties. The recruiter places candidates.

The chatbot doesn't replace the human relationship. It removes the painful part -- the administrative collection -- so the human relationship can start sooner and on a better footing. When a client arrives at their first appointment with a complete file, the conversation is immediately productive.

If you want to explore how generative AI can transform other aspects of your business, our practical guide for SMEs is a good starting point.

Also worth reading

• What is a FinOps audit? A complete guide for SMEs →
• Why an AI agent connected to your CRM replaces your forms →
• How much does an AI chatbot cost for an SME in 2026? →
• Automating customer relations with WhatsApp and AI →
• WhatsApp AI assistant with Amazon Bedrock →
• Integrating generative AI into your SME: a practical guide →