Original subtitles

All right.

So as you probably know over the past year or so Microsoft has been making some pretty incredible strides

integrating artificial intelligence into tools like power by now.

Some of these tools like Azure machine learning models or sentiment analysis or text analytics are currently

only available to enterprise or premium capacity users.

But the good news is that there are some great A.I. tools that are accessible to free or pro users as

well.

So what I want to do here is cover some of the A.I. driven visuals that were recently introduced to

provide desktop.

Now important note what you're looking at here is the December 20 19 release.

Make sure you've got a current version of power b ice that you see these new visuals if you're not sure

head to the upper left corner.

Jump in to help about and you should see the version number so I'm on to 7 6 which is the December 20

19 release.

Now the first A.I. visual I want to talk about is the Q and A VISUAL which looks like a little callout

box or text box here.

We'll see a little light bulb in the lower right corner.

There are three visuals here that I'm showing with that light bulb that indicates that it's an A.I.

visual.

So we're going to use our venture works report here that we've built throughout the course.

Let's go ahead and clear this filter on our fields.

Just want to work with a data model that we're already comfortable and familiar with.

So first things first we're going to add a new tab or page let's call it Q and A nothing fancy.

And here we go.

Got our blank canvas.

There are two ways we can insert a Q and A visual and click the button just like any other chart or

if you go to the Home Menu you'll see an ask a question option which does the exact same thing and delete

that duplicate.

You can drag this out.

This is the only thing we're gonna be showing here.

And the first thing you'll probably notice is that we have a question bar here and some suggestions

to get started.

So Microsoft is looking at our data model and offering some suggestions things I might want to look

at.

So total cost by SKU category total revenue by SKU category top genders by year to date can show some

more kind of get a full list.

Now listen I'm going to be honest with you guys I very rarely will use these suggestions that kind of

feel like a shot in the dark.

Sometimes we'll get some interesting suggestions.

But in this case you know we're not that interested in the SKU category.

Some of these are just a little weird like Day of Week over time that really doesn't make much sense.

But if you click on one you can see kind of how this visual works.

Right.

It populates the query turns that query into a visual.

And because this is an A.I. visual you're not going to get the same chart type every time powered by

is going to look at your query it's going to interpret what you're asking and what you're looking for

and it's going to render the most appropriate visual to answer that question.

So here we're looking at SKU category with total profit as our values.

So let's go ahead and clear that out and ask a question of our own.

Now what's important here is that we want to be able to ask a question the same way that we would speak

to a friend right.

This visual wouldn't be very powerful or useful if we had to type the same way that we create a tax

measure for instance.

Right things like a W. underscore sales bracket.

That's not natural language.

That's not a natural query.

So what we can do here is type the same way that we would speak.

All right.

So maybe want something like revenue by category presenter.

We got a result.

Granted it's maybe not the one that we want but power by recognized keywords in the query which it underlined

in blue and it took a stab at rendering what it thinks we're looking for.

So skew category and I can see what the tool tip that it is showing adjusted revenue not quite what

we want here.

The good news is we can click on that keyword and we can use a different suggestion.

So in this case we want total revenue and instead of skew category we want the product category name.

There we go and now we get something a little bit more reasonable.

Right.

We get bikes accessories clothing and based on the tool tip.

We're looking at that total revenue measure.

So that looks great.

Now powered by pretty smart when it comes this visual.

So let's say that I was typing very quickly and I did something like this

revenue by category name.

Well it gave me the correct results.

Same to you that we just showed.

And now what it's done is underlined that first term revenue in red.

And it's saying hey you know we think you're probably trying to type revenue here.

So we're going to show you the results for revenue by category name just like Google does.

You know the auto correct tool powered by AIS driven by obviously a behind the scenes and as part of

that A.I. and understands synonyms and corrections just like the one we're looking at here.

So again I can click in and I can say you know what I want total revenue by category name and boom we're

off and running.

Now this is really just scratching the surface of what this visual is capable of.

You know this is a simple query one measure one column or dimension name get a little bit more sophisticated

in this and we could do something like total revenue for red bikes.

And because this query is designed to aggregate values to a single result power RBI has displayed the

answer or the result in the form of card one value.

Four point eighty seven million.

If you're not sure if you don't trust that that's right we can go ahead and test that.

We can add a card ourselves.

Let's expand our filters and in this card we want to show that total revenue measure.

And we want to filter it down for the bike category.

There we go.

And for the color product color red four point eight seven million.

Boom.

There we go.

The Q and A VISUAL.

Did its job it return the proper result which is actually pretty impressive when you consider the complexity

of this query.

We're looking for a measure aggregated values based on an attribute read which has a value in a dimension

from one table the product table and also category equals bikes which is a dimension from a second table.

The product category table.

So the fact that Part B I was able to interpret that instantly and render the proper result is actually

pretty incredible.

Now we can take this a step further.

Beauty of of AI and machine learning is that you can train these models and train these tools to get

smarter over time.

So let me show you what that looks like.

Going to click this gear icon here this is going to open up this menu.

Granted this is a brand new visual.

Things are changing and evolving so you might not see this exact view but basically I've got three options

here I can review questions that have been asked in this visual I can teach the Q and A VISUAL to get

smarter and understand new synonyms and I can manage any custom terms or synonyms that I've defined.

If you click this link here you'll go to the official Microsoft documentation if you want to dig a little

bit deeper.

But let me show you how this works.

Now if I click review questions I'm going to see a list of questions that I've been asked in this visual.

Now you won't see any questions here.

If you've only interacted with the visual in power b I desktop in order for this to work you have to

publish this to a workspace in the power be a service environment which we do cover in a separate course

publishing it Power by Service.

So as a test what I've done is upload or or publish this report to a workspace in the service environment.

I asked a bunch of questions inside of this visual and they populated right here.

So the important thing to note on this tab or this view is that any query that has a red underline or

a keyword that power b I didn't understand has this pencil icon where it says fixed needed.

And when I click that icon takes me to the teach Q and A tab it populates that query.

And when I hit submit what it's going to do is show me a preview of what the Q and A VISUAL would return

here on the right side.

So in this case a card with the value twenty four point nine one million.

And it also lets me define the term and read the term that power be I wasn't able to interpret.

So in this case you know I'm looking for product type which isn't actually a column name or field in

my model but I know that this user because it was me is actually looking for the category right bikes

clothing accessories.

So what I can do is tell power b I hate whenever you see someone type product type what they really

mean is category name.

OK so I've just defined a new synonym and watch this preview update.

Boom.

There you go.

So now instead of the card it's showing that bar chart with the breakdown by category name.

So when I save that synonym I've now made this visual smarter and I can manage any terms that I've created

right here in that last tab.

So pretty cool stuff if I close out of that dialog box now and I head into our model tab you'll see

this kind of populate one other place which is in the properties tab.

So if we look at the category name inside of our category lookup table there's this box called synonyms.

Check it out category name category name no space category and product type which is the one that we

just defined in the Q and A VISUAL.

So pretty cool stuff pretty user friendly.

One last query that I want to show you.

Let's say we do something like orders by order quantity.

In this case power b I interpreted this as a line chart but really we're not showing anything in a time

series.

We don't really want a line chart here.

I think a bar column chart would be a little bit more appropriate.

So what we can do is convert this to a visual using this button here just turns it right into a line

chart and all we need to do is click and we can modify it change the chart type.

We can format this we can filter it any way that we like.

Just like any other visual.

So there you have it pretty powerful stuff.

That's the A.I. driven q and a visual.