Two Different Products for Different People

Dialogflow is Google's conversational AI platform. It lets you build chatbots from scratch — define intents, write training phrases, create conversation flows, and deploy across channels.

Intent classification (like Supp) is a pre-built support automation tool. 315 intents are already trained. You set up routing rules and auto-responses. No chatbot building required.

Dialogflow is for builders. Classification is for users. The distinction matters.

What Dialogflow Requires

To build a support chatbot on Dialogflow, you need:

1. Define your intents. Write out every question type your customers ask. Create training phrases for each (Google recommends 10-20 per intent). For 50 intents, that's 500-1,000 training phrases to write.

2. Build conversation flows. Define what happens when an intent is matched. Response text, follow-up prompts, API calls (fulfillment), and context management.

3. Integrate with your channels. Connect Dialogflow to your website widget, WhatsApp, Facebook Messenger, or phone system. Each channel requires separate integration work.

4. Test and iterate. Dialogflow's NLU improves with training data. You'll spend weeks refining intent matching, fixing misclassifications, and adding training phrases for edge cases.

5. Maintain forever. New intents emerge. Language patterns change. Integrations break. This is an ongoing engineering project.

Timeline: 2-6 months to build, indefinite maintenance. Cost: Dialogflow CX starts at $0.007/request. But the real cost is engineering time.

What Classification Requires

1. Sign up and install widget. 5 minutes. 2. Write 5-10 response templates. 30 minutes. 3. Create routing rules. 15 minutes. 4. Test. 5 minutes.

Timeline: Under 1 hour. Cost: $0.20/classification. No engineering time.

When Dialogflow Wins

You need a conversational chatbot. If your use case requires multi-turn conversations where the bot asks follow-up questions, collects information, and responds dynamically, Dialogflow is built for that. Classification identifies intent from a single message — it doesn't have multi-turn conversations.

You need voice integration. Dialogflow CX integrates natively with Google Contact Center AI for phone-based conversational AI. If you're building an AI phone agent, Dialogflow is the starting point.

You want full control over NLU. Dialogflow lets you train custom NLU models with your exact training data. You control the intent taxonomy, the entity extraction, and the conversation logic. For teams with specific NLU needs, this flexibility matters.

You're building a product, not using a tool. If your business IS a chatbot (you're building a chatbot product for others), Dialogflow is infrastructure. Classification is a finished product for a different audience.

When Classification Wins

You want support automation, not a chatbot project. You don't need to build anything. You need your password reset questions answered automatically and your bug reports routed to GitHub. Classification does that in 30 minutes.

You don't have engineering resources. Dialogflow requires a developer to build and maintain. Classification requires someone who can fill out a form.

Speed matters. Dialogflow takes months. Classification takes minutes. If you're a startup that needs support automation this week, you can't wait for a Dialogflow build.

You want pre-trained accuracy. Dialogflow starts with no training data — you provide everything. Classification starts at 92% accuracy on 315 support intents because the model is pre-trained.

So Which One?

Dialogflow is an engineering tool for building custom conversational AI. It's powerful, flexible, and complex. If you have engineers and months to invest, you can build exactly what you want.

Classification is a product for people who want support automation now. It's simple, fast, and limited to what it's built for. If you want a pre-built solution that handles 60-70% of support messages, it's ready today.

Most small teams that start with Dialogflow end up 3 months in with a half-built chatbot that handles 10 intents. Most teams that start with classification have automation running in under an hour with 315 intents. The math on where to spend your time isn't complicated.