[INFLUENCE MAKE] - AI Assistant Product Feedback

Hello :wave:

We just released a new major Beta version of our AI.

We are excited to introduce the latest update to our AI Assistant and would love to get your feedback on it! :tada:

Our goal with this update is to make creating powerful scenarios and automating your workflows even easier. Whether youā€™re generating new scenarios, crafting content, or exploring example scenarios to get started, our AI Assistant is here to help.

Hereā€™s what you can do with the new AI Assistant:

  • Build Scenario: Generate new scenarios for various use cases with ease.
  • Find Templates: Quickly find and explore example templates.
  • Ask Anything: Get answers to all your queries about features, bundles, and more.
  • Create JSON: Create JSON with specific parameters effortlessly.
  • Help with Error: Get assistance with troubleshooting errors.
  • Create a Mapping: No more tedious mapping between modules.

We would love to hear your thoughts on the following:

  1. Usability: How easy is it to navigate and use the new AI Assistant?
  2. Features: Are there any features you find particularly useful or any that you think are missing?
  3. Performance: How would you rate the performance and responsiveness of the assistant?
  4. Overall Experience: How does this update improve your workflow and productivity?

Your feedback is invaluable to us as we strive to enhance your experience with Make. Please share your thoughts and suggestions in the comments below or reach out to us directly. Share us what you think!

Thank you for being a part of our community and helping us make Make better for everyone!

Best regards,

SEB
Head of applied AI | Make

8 Likes

Hi @SebastianMertens,

In-tool JSON validation? That is cool!

From my learning days, I fondly remember copy-pasting large JSON objects to a JSON validator to check where an error occured. This is a huge help for sure!

Two points of feedback:

  • (small point) Make AI is able to validate the JSON and map it to a Module. However, the JSON that is inserted is not ā€˜prettyā€™, as in the answer Make AI gives, making it less readable when you later open the module.
  • The scenario is auto-aligned automatically after inserting the JSON. Now I have a scenario that is wrangled because the auto-align made a messā€¦ :smiling_face_with_tear:

Have not found time to play around with everything else.

Cheers,
Henk

1 Like

Thanks a lot!

Here are use cases one can inspire from to use with our AI:

  • Use ChatGPT to translate each of my incoming emails from Gmail to German and then create a draft email with the translation.
  • Add [[module]] at the end of the scenario.
  • Convert Word files from Google Drive to PDF using pdf co and upload them back to Google Drive.
  • How can I make a scenario?
  • Map the first module to the second.
  • For every email I receive, add a row to a Google Sheet with the date, sender, and subject. Then summarize the email using ChatGPT and send me the summary as a Slack message.
  • I donā€™t know what new blueprint to create. Do you have any ideas?
  • Map every relevant output from modules with IDs [[1]] and [[2]] to the module with ID [[3]].
  • Remove the last module.
  • Create a JSON with a list of attributes ([[name]], [[age]]) and fill the values from [[module]].
  • What does the aggregator do?
  • Create a JSON with a list of 5 largest European countries, their population numbers, and capital cities.
  • What templates are there with [[app]] and [[app]]?
  • I want to get inspired how to build a scenario with [[app]] from a template.
  • How do I fix this error?
  • For every new Jira ticket, select all users in my Jira team and assign the ticket to one of them.
  • Replace the [[app1]] module with an appropriate [[app2]] module.
  • Create an email invoice. Read the output of [[module]] and use it for the email content.
  • How can I parse an HTTP response?
  • Why did I get this error?
  • Create a JSON from the first module response.
  • Map the output from [[module1]] to [[module2]].
  • Create an email response with output from [[module]]. Thank the sender for the input and say that we will reply shortly.
  • Show me an example blueprint.
  • I want to create a blueprint that sends an email. Show me some examples of blueprints like these.
  • What is a bundle?
3 Likes

I love the direction this is going! Great steps forward.

However, it seems to want to adjust things frequently and incorrectly, introducing random things (like a module that contains the name ā€œplaceholderā€ as an attempt to add placeholders) despite adding prompts to NOT do that.

On the other hand, it seems to give better suggestions, being able to adjust and create JSON using AI from inside Make is great. Also asking for suggestions on how to achieve certain things with functions is great. I just wish I could hit a switch that prevents the AI mode from adjusting anything in itā€™s current state.

1 Like

I asked this question and got this response:

Q: How do I get the 2nd item of an array using make

A: To get the 2nd item of an array in Make, follow these steps:

  1. Create a new scenario or open an existing one.
  2. Add the Get element module to your scenario.
  3. In the module settings, select the array from which you want to get the element.
  4. Set the index to 1 to get the 2nd item (since arrays are zero-indexed).
  5. Run the scenario to see the result.

This will extract the 2nd item from the specified array. If you need further assistance or have any other questions, feel free to ask!

### Mapping arrays #### Make Help Center

What is the ā€œGet elementā€ module? First time I have heard of this module. I was fully expecting an explanation of the get() function and got this instead.

2 Likes

I also tried to have it help Parse JSON but it created some sample JSON and didnā€™t add it to the Parse JSON Module leaving it blank.

It said this:

The JSON module has been added to your current blueprint and configured to parse the following JSON string:

{
  "countries": [
    {
      "name": "Russia",
      "population": 144.1,
      "capital": "Moscow"
    },
    {
      "name": "Germany",
      "population": 83.2,
      "capital": "Berlin"
    },
    {
      "name": "France",
      "population": 67.1,
      "capital": "Paris"
    },
    {
      "name": "United Kingdom",
      "population": 66.7,
      "capital": "London"
    },
    {
      "name": "Italy",
      "population": 60.4,
      "capital": "Rome"
    }
  ]
}

Would you like to map the parsed JSON data to another module or perform any other actions with it?

But the actual module it generated was this:

Scenario Blueprint
{
    "subflows": [
        {
            "flow": [
                {
                    "id": 1,
                    "module": "json:ParseJSON",
                    "version": 1,
                    "mapper": {},
                    "metadata": {
                        "designer": {
                            "x": 0,
                            "y": 0,
                            "messages": [
                                {
                                    "category": "last",
                                    "severity": "warning",
                                    "message": "A transformer should not be the last module in the route."
                                },
                                {
                                    "category": "setup",
                                    "severity": "error",
                                    "message": "Value must not be empty."
                                }
                            ]
                        }
                    }
                }
            ]
        }
    ],
    "metadata": {
        "version": 1
    }
}
2 Likes

Will work on this - great feedback!

Same here. Suggesting non existing modules in scenarios i evilish :confused:

I there a chance, some day i will ask it to write scenario that ā€¦ - and get template to import in make with all the modules :)?

Hi,

The AI is a great ideas. Here are some of the places where itā€™s lacking:

  • WHen it adds a module, the parameters of the content are often (always?) missing. So it will create a Parse JSON module and say itā€™s taking output elements from a preceding module, but when you open it, there is nothing there.
  • If I ask it twice in the same chat to modify my scenario, the second time it says there was an error and I have to delete the chat and create a new one
  • It hallucinates solutions, meaning it says it did something with the scenario that will do what I want, but when I run it itā€™s nowhere near a solution.

L

This is a great way to describe it ā€“ this is probably the worst thing an AI could do is suggest something that works and it is not nearly close to a solution.

I see AI sometimes as an overly confident friend that has trouble admitting they donā€™t know the answer to your questions and makes stuff up with whatever limited knowledge on the subject they haveā€¦ Thatā€™s why in my prompts I always add ā€œItā€™s okay if you do not have any suggestions for my question. Please let me know if you are unable to formulate a solution.ā€

3 Likes

This is amazing but why there is limitation? on number of times we can use ?