The Latest Google Fail— Dialogflow CX and Why I Hate It | Solely My Opinion
Working as a software engineer in a sector that requires me to know about chat and voice bot building, meant that I’ve come across the original Dialogflow (now renamed ‘Dialogflow ES’ where ‘ES’ stands for ‘Essentials’) and the latest beta release of Dialogflow CX (where ‘CX’ stands for ‘Customer Experience’).
And what you’re about to read will probably never be published in my company’s publication because booooy is this new Dialogflow a nightmare of a product and you’re going to see why.
The Basics of Dialogflow
If you’ve used Dialogflow before, feel free to skip this section.
If you haven’t, Dialogflow can be used as a chat and voice bot with the right integrations. How it works is that when a user says something, the bot will find an intent to match it with and respond back. The way intents are matched are based on the training phrases that you include.
For example, you have a “Welcome” Intent. In this intent, you have the following training phrases: “Hi”, “Hello”, “What’s up!”, “Good morning”. And so on. You can also add some sentences that your bot can respond with, such as: “Hi, what can I do for you today?” or “Good morning, is there anything I can help you with today?”
When a user greets your bot with “Hi!”, the “Welcome” Intent is triggered and either one of the responses will be returned. Even if the user says something that sounds remotely like a ‘hello’ greeting, Dialogflow is intelligent enough to pick up on it and trigger the right intent (well, most of the time).
And that’s it. It’s that simple.
If you want your bot to be able to do more, like understand whether the user wants to schedule something, or buy something, or speak to someone, you can always create more intents to cover those cases.
Well… it was simple until Dialogflow CX came along.
Why I Had to Explore Dialogflow CX
The reason why I even explored CX was because of 2 main functionalities CX promises:
- CX promises a more advanced Natural Language Understanding (NLU) compared to the ES. You could even adjust a classification threshold to improve your intent matching results. And it is always a good thing to have more control over your bot; however I’ve yet to try the NLU out on our systems because the integration isn’t available yet, so let’s just cross our fingers and hope it comes soon
- It has a drag and drop feature which honestly… looks intuitive on the surface… But when you realise the tutorial is so god damn long, there should’ve been gigantic warning signs, loud-ass warning bells that should’ve gone off at Google telling them that it really isn’t that intuitive
Although sure, this could suggest that you could build a multi-page agent much easily (e.g. 1 page handles ordering, 1 page handles after sale service and another page provides information), similar to building a mega-agent in Dialogflow ES.
But ironically, they made building a single page/agent so much more complicated.
Dialogflow CX: How it Ruined Bot Building for Me
To help you better understand my frustrations, let me just show you the steps on how to build your bot’s pages and intents without any screenshots.
- Go into a tab that isn’t the main tab you automatically land on: i.e. Click the “Manage” tab
- Then click into the “Intent” tab
- Click “Create” to create your intent
- Name your intent with some kind of weird dot naming to your intents’ name
(as of now I don’t understand why it’s the case, but their tutorial suggests “store.location” instead of something more intuitive for the average user like “Store Location Intent” but ok) - Add your training phrases
(minimum 10; which ok, I have no qualms about because it helps to improve the bot’s ability to link to the correct intent) - Click “Save”
- Then, go back to the original tab you came from: “Build”
(ok, I’ll be honest, past step 7, I just copied the instructions from Dialogflow CX’s tutorial because oh my god it is so long to even get things started) - Select the “Default Start Flow” in the “Flows” section.
(Are you lost yet?) - Click the “Start” node in the graph. This is the start page for the “Default Start Flow”.
(Do you need the screenshots?) - Find the intent route with the “Default Welcome Intent” as an intent requirement and click it. This opens a panel to edit the intent route information
(Would it help if you have the screenshots of the entire page but the catch is that you don’t have a glaringly red box to show you exactly what to click on??) - Find the fulfillment section
- Add your own responses
- Click “Save”
- Test out your bot and you’ll see that FINALLY it works
Now, let me compare and contrast it to how Dialogflow ES works:
- From the main page you are on, click on the “Intent” tab
- Click on the “Create” button to create your intent
- Add in your training phrases
- Add in what your bot should respond with
- Click “Save”
- Test out your bot and you’ll see that the intent will be matched and your bot works, just like that.
Tell me, which one could you see yourself building a bot with if you only had the instructions without the screenshots?
Now, imagine paying 2.5x more than ES, with that kind of bot builder in this economy?!
While some of you might go ‘But CeLiA, that’s not fair on Dialogflow CX. They are providing you with a drag and drop. Your example isn’t meant for CX. CX is expanding on the features so that you can create bots with more interesting features and paths and options and… blah blah blah’.
And ok, I get it. CX is meant for bots that are far more complex, usually on the enterprise level. But considering you can do the same with ES to build mega-agents, CX isn’t simplifying building a mega-agent by that huge of a margin to justify the increase in price.
Sure, it looks prettier and it might be somewhat easier to see all of your agents in one glance, but just because it has a visual builder doesn’t mean it makes it any easier.
Now don’t get me wrong. I was excited for Dialogflow CX because it’s promising. They have plans that run parallel to what my team and I envision for our product.
But damn, I for sure know that the biggest takeaway from this train-wreck is: Google needs to reassess the products that are actually beta because this should probably go back to the labs.
And fine, at the end of the day we have to take into consideration that it is only a beta. And it might continue to improve before it is officially released. But come on Google, aren’t you hiring the best of the best? What’s going on in your UX team?
Note: There are other functionalities that’ve expanded and there’s also limitations but I understand it’s a beta so I won’t go into it because things will change (and I hope like hell it will before the official release).
Honestly, I think I’ve tried my best trying to understand the different concepts in Dialogflow CX. If you’re just as confused about the process, I completely understand because I don’t even think I completely understand it either.
I feel like every time I look at Dialogflow CX, I have to run through their tutorial 10 more times before I think I can create my own bot. And even then, I’ll probably still refer back to the documentations to help me out.
Have you used Dialogflow CX yet? What do you think about it? Let me know, I’d honestly love to hear your thoughts!
Also, if you enjoyed this post, please leave me some claps! It would make a world of difference to know if you enjoyed my content! :)