Transcript:
0:02 In this video, we're going to show you
0:04 how we can generate AI summaries for
0:08 your business
0:10 rules. So, what we have is a Power
0:12 Automate uh flow here which is set up to
0:16 uh trigger when a record is
0:18 selected. Uh it is set up on the north
0:22 52 formula table. So you'll be able to
0:25 select that uh in a list of formulas or
0:29 on an individual formula
0:30 itself. Second thing we have in the flow
0:33 is we set up a variable here which is
0:36 called the prompt. Uh it is type string
0:40 and this is the information we pass to
0:43 the uh LLM model to generate the formula
0:49 description. So here we've got uh a
0:53 prompt saying you're a technical writer
0:55 specializing in business rules. Write a
0:57 plain text summary that describes the
0:58 following rules. Then what we're doing
1:01 here is adding in the formula
1:03 description field. Uh and we get we get
1:06 that directly from uh the record that
1:10 was selected. We just click on formula
1:12 description there to add it to our our
1:14 variable from there.
1:17 So the next thing we do is we pass the
1:19 uh details uh through to a a language
1:22 model and uh for this example we're
1:25 actually using a service called open
1:28 router. Um this service allows you to
1:30 buy credits and then uh select uh from a
1:34 a large number of AI models. Um what
1:38 we've found uh in our testing is that
1:40 the Claude 3.7 sonnet it does a very
1:43 good job of interpreting 52 rules. I
1:46 mean this obviously may change over time
1:48 uh but at the moment we can use this
1:50 model uh with open router and if it
1:53 needs to change later on we can easily
1:55 change the
1:56 model. We need to fill out the
1:58 parameters here. So the URI uh here um
2:02 we will provide a link in the knowledgeb
2:04 article too of of the documentation for
2:06 this. The method is post uh then the
2:09 content type is application JSON. For
2:12 authorization, um it is bearer followed
2:15 by your API key. And we're actually
2:17 using uh an environmental variable here
2:20 to store the API key. And you can get
2:22 that by uh using your parameters uh
2:25 functionality in here to insert that.
2:29 Then we send through the the body uh
2:31 with the model name and the prompt from
2:34 our variable in the previous step. And
2:36 to access that variable, we just put our
2:38 cursor in here. And again just come
2:41 through into our search. We can see our
2:43 variables and you just click on
2:48 prompt. So the service then uh executes
2:52 that and it returns a value. And then we
2:54 want to update a row in data versse uh
2:56 to do this. So what we're doing again
2:58 we're selecting the the formula table.
3:01 Uh the row ID of is the the formula that
3:04 we get from when a record is selected.
3:07 So we can we get that from um in in
3:11 here. Uh so it's this one
3:15 formula and then we're looking at uh the
3:19 formula summary. Perhaps you probably
3:20 need to show advanced parameters to show
3:23 them all. But then in the formula
3:24 summary we need to insert an expression.
3:28 uh and this is pulling out the response
3:32 uh from the HTTP call and we're pulling
3:36 out the first choice and the text value
3:39 of the JSON that is returned from the
3:41 API service and then populating that
3:43 into the uh formula description field.
3:47 So let's have a look and see how this
3:48 actually executes. We'll come across to
3:51 our formula list here and I'm just going
3:54 to choose one of these uh claim
3:57 adjudication formulas here. Some
3:59 advanced rules that we've got here. You
4:00 can see none none of these have got any
4:01 formula summaries in the at the moment.
4:04 We can select multiple ones here. I'm
4:06 just going to show you with one. So from
4:09 our flow, we're going to generate AI
4:11 summary for business rules. We're going
4:14 to run this flow. Um and once that flow
4:17 is done uh it's going to uh execute in
4:21 here. So what we'll do is just run this
4:24 and we'll refresh
4:27 this and we can see here that our uh
4:30 formula summary has been populated. Now
4:32 if we open up the formula itself
4:38 uh then we can see the um formula
4:42 summary here. We can see that's actually
4:43 fairly comprehensive uh summary of all
4:46 the different decision uh tables and the
4:49 decisions that are on those tables. So
4:52 what you can do is an experiment with
4:54 the prompt uh itself. So if you wanted a
4:57 shorter summary um what we could do in
5:00 here in the initial IP we variable we
5:03 could write a plain text summary. You
5:06 could say write a plain text one
5:11 paragraph summary and that would
5:14 obviously create a shorter version uh
5:17 for your uh
5:19 summary. Hopefully that helps. Once
5:21 you've got that uh set up on your
5:23 system, then you'll be able to utilize
5:25 that information to generate various
5:28 reports, exports to Excel, Word, etc.