---
title: "AI Automators as tools for AI"
video_id: PXfmh3QXjN0
url: https://www.youtube.com/watch?v=PXfmh3QXjN0
workflow_page: https://workflows-of-ai.com/workflow/ai-automators-tools-ai
language: en
captions: generated
---

0:00 Okay, tools being written in automator
0:04 uh for the agent. So I will just
0:08 showcase the end result of setting up
0:10 everything. So can you tell me what
0:12 people say about Tiffen at Rocket
0:14 Center? So this will do a Google. So
0:17 there's a automated chain in the
0:18 background running a
0:20 Google places look up into a address
0:23 field taking that address field uh and
0:26 looking up all the Google reviews and
0:28 then there's a summary of this and we
0:30 get a real summary. So this is based on
0:32 the Google reviews
0:35 uh in Google maps of uh this stiffness
0:39 that is somewhere around rockall
0:41 center. Um so let's showcase how we did
0:45 set up this. So you start with
0:47 automators automator chain settings set
0:50 up a new automator chain type and
0:55 uh you can just see it here. So it
0:58 basically starts from the input search
1:00 place which is a text field that is a
1:03 required field. So that's where we start
1:05 adding stuff. This uh looks up this
1:09 place into a address field using the uh
1:13 Google place a search to address field.
1:15 So it gets the real address
1:19 uh and from that it's using the Google
1:22 uh places extractor. So that's a field
1:24 that can extract many things and what's
1:26 important here we're extracting the
1:28 reviews. We're extracting some images
1:30 and display name as well. I was just
1:32 wanted to check so that works but they
1:35 are added into reviews that is a
1:36 multiffield because I think we get 10
1:38 reviews and then in the end of the chain
1:41 there is a review summary of the
1:45 reviews. Uh and what we can do if we
1:49 know the URL here we we should link it
1:51 here since this is hidden by default and
1:54 you have to turn it on. So you can run
1:56 an automated uh chain here. So if we
1:59 take this
2:01 uh and this is how you would search
2:04 basically
2:06 uh in Google maps right so you could
2:09 probably find something with that
2:11 information without having to say New
2:13 York it understands that what the follow
2:15 center is in New York and it's the
2:18 Tiffany's in New York and you can see
2:20 here it got the right address
2:23 uh and here is the uh review summary
2:27 there's images and display names But
2:29 here you have the reviews that that this
2:31 is based on. So you can try it out there
2:35 uh to see that the actual automator
2:38 chain works. And then when you have done
2:40 that under AI automators there's
2:42 automators tool where you can create a
2:44 new tool uh where you give it a label
2:47 you give it a description you find the
2:50 workflow that you have set up. So if you
2:52 start searching here it's going to fill
2:55 it out right and then you decide
2:57 basically which of all these fields are
3:00 input which are output which should be
3:04 ignore which uh you can set a default
3:06 value. So currently we mostly support
3:09 text fields uh integer fields list
3:12 fields and so on. U but later we should
3:17 basically be able to support everything.
3:18 So we have the review summary being the
3:21 output and the input search please being
3:23 the input.
3:24 Then then we just go ahead under agent
3:27 settings and create a agent. So we have
3:30 created a Google review summary agent
3:33 uh some system prompt and it's using
3:36 this get Google review summaries. We can
3:38 test the tool with tool explorer. So we
3:41 once again write Tiffany's
3:45 uh local feller center run this function
3:50 take some time and it will output a
3:52 value here
3:54 uh where it write out the value
3:58 right and basically what we're saying
4:01 here use this tool if it's a place and
4:04 then uh uh rewrite the review
4:09 uh we don't need any detail you know
4:12 force any values or something here and
4:15 then we can of course use
4:17 the explore here as well to try it. So
4:21 yeah in Berlin another it's a famous
4:24 place here in Berlin. So you can see
4:26 it's using the actual tool first and
4:30 then it's outputting the information
4:35 right
4:36 uh and when we have the agent of course
4:38 we can connect that to the assistant I
4:42 should probably I wrote very little here
4:47 uh please answer word for word as the
4:51 agent is answering because it was adding
4:55 stuff I saw there enable agent action
4:57 the Google review summary
5:00 agent and then we can use this then uh
5:04 what does people say
5:07 about in
5:11 Berlin and we have a end to end process
5:14 where we can set up a fairly
5:17 complex workflow that's being used as a
5:20 tool by agent by our assistant
5:24 uh without writing a single line of
5:27 code.