Build a WhatsApp AI assistant for voice, images, and PDFs

Find Your AI Training Job
Chat with DigiNo Toucan and get matched to remote roles in one minute.

Busy executives and solo consultants want a single interface for managing their day — this AI automation turns WhatsApp into a fully functional personal assistant that reads voice notes, scans documents, and takes action across email and calendar on their behalf.

What This Automation Does

Receives WhatsApp messages in any format — text, voice, image, or PDF — and routes each to the correct processing path before a unified AI agent handles the request
Transcribes voice messages via OpenAI Whisper and replies in audio using OpenAI TTS, so the entire conversation feels natural and hands-free for the end user
Connects the AI agent to Gmail, Google Calendar, Google Drive, and Airtable so it can send emails, schedule events, search files, and look up contacts from a single WhatsApp message
Returns a spoken audio reply for voice inputs and a text reply for everything else, keeping the response format consistent with how the user reached out

Tools Used

n8n
Claude
OpenAI
Gmail
Google Calendar
Google Drive
Airtable
WhatsApp

Work faster

Stop Typing and Start Speaking With Typeless

Typeless turns what you say into clean, formatted text in any app on Mac, Windows, Linux, iOS and Android, which it rates at 4x faster than typing. It strips filler words, keeps only your final phrasing when you change your mind mid-sentence, and works in 100+ languages. Emails, AI prompts, client messages and docs, done by voice.

Try Typeless →

Where to Get Hired for This Skill

On Contra, top freelancers across this stack have earned 311 combined verified reviews from real client projects.

Source: Contra freelancer search · refreshed 30 May 2026

Start Earning as a Freelancer on Contra

Contra is a commission-free professional network for independents. Browse live AI automation work and keep what you earn.

Join Contra Free →

How To Build It

Wire WhatsApp to the message router

Configure the WhatsApp integration to receive incoming messages and branch the flow based on message type — text, audio, image, or PDF — so each input reaches the correct processing logic downstream.

Set up audio transcription and image analysis

Connect OpenAI Whisper to download and transcribe incoming voice messages, and configure GPT-4o mini to analyse image attachments, normalising both outputs into a single text field before they reach the AI agent.

Extract and validate PDF content

Add a PDF handling path that validates the file, downloads it, and extracts its full text so document-based requests — such as summarising a contract or pulling figures from a report — are converted into readable input for the agent.

Connect Claude to the full tool suite

Configure Claude Sonnet as the central AI agent and attach its full set of action tools: Gmail for sending and searching email, Google Calendar for reading and creating events, Google Drive for file search, Airtable for contact lookups, and SerpAPI for live web queries.

Route replies by original input type

Add a conditional output step that checks whether the original message was a voice note — if so, convert the agent's text response to audio via OpenAI TTS and send it as a WhatsApp audio message, otherwise send a standard text reply.

Pitfalls

WhatsApp Business API credentials expire or get flagged for unusual message volumes, causing the trigger to go silent mid-deployment — always build in error notifications and confirm the client's verified Business account before going live.
Claude's tool-calling can drift when a single WhatsApp message is ambiguous across multiple action types, such as a voice note that mentions both an email and a calendar event — define clear agent instructions and test edge-case prompts before handoff.
Google OAuth tokens for Gmail and Calendar require periodic re-authorisation, and a lapsed token will silently fail the action without alerting the user in WhatsApp — set up a monitoring check and document the re-auth process for the client retainer.

FAQ

Can I build this without coding?

Most of the build is visual configuration — connecting credentials, setting routing logic, and writing the agent's system prompt. The only section that may involve a short code snippet is PDF text extraction, but template-level examples make this straightforward for a low-code builder.

How long does it take?

A focused first build typically takes four to eight hours, including credential setup across WhatsApp, Google, Airtable, and OpenAI. A second deployment for a new client can be completed in under two hours once you have a reusable base configuration.

What can I charge?

Pricing is your call based on your market and the client scope — common structures include a one-time setup fee for the configured build plus a monthly retainer covering credential maintenance, OAuth renewals, and prompt updates as the client's needs evolve.

Which tool is required vs optional?

WhatsApp, Claude, and OpenAI are required for the core voice-to-text-to-voice loop. Gmail, Google Calendar, Google Drive, and Airtable are modular — you can launch with just one or two and add the rest based on what the client actually needs in their workflow.

This is original DigiNo analysis. The underlying automation pattern is a community workflow template – view the original on n8n.