Hello everyone, welcome back to

Analytical Guy. In this video, we will

talk about a very hot topic, can AI

replace Power BI developers? Power BI is

here for more than 10 years and now

everyone is thinking that AI can replace

Power BI. So, we will understand this in

this particular video. So, stay tuned

till the end. All right, so let's try to

understand if AI can replace Power BI

developers using three-tier framework.

So, first tier in this framework is AI

does this really well. So, it can be

technical execution such as using DAX

for various tasks or M code etc. So,

this is tier one. Then tier two would be

AI helps

but human leads. For example, AI can

accelerate the work but business

judgment and context are required to

make the right call. So, this is where

human leads, right? Coming to the third

tier is humans are irreplaceable. So, if

we talk about any particular skill such

as business communication, if you're

talking to stakeholders, understanding

the context, judgment, all these are the

things which are irreplaceable as of now

and then I'll explain you all the skill

sets required for this particular tier

because this tier is really important

for us to grow in our field. All right,

so let's start with tier one.

AI does this really well. So, we are

talking about technical execution. So,

in technical execution, I have seen

Cloud Code handles all these tasks

really well such as if you have to work

on DAX or M code and if you have seen my

previous videos, I have done end-to-end

Power BI report development using Cloud.

So, you can refer to the top screen and

you can have that link to see how it's

done. But, what we have seen earlier as

well, Cloud Code does really well and

various other AI platforms. So, what DAX

does this? For example, if you are on

Power BI report and you want to create a

new formula, right? And for example, you

want to have total sales. So, you need

to have a formula, right? So, formulas

are created using DAX language. DAX is

data analysis expression. This is a

language which is understood by Power BI

database, right? So, it writes measures,

it can create calculated columns, right?

So, these things can be done really well

using Cloud Code. That doesn't mean that

you don't need to learn DAX, but what

I'm saying is

uh you need to know the concepts, but

majority of the part can be done using

Cloud Code. Same goes with the M code.

So, M code is a language

understood by Power Query. So, if you're

cleaning your data or transforming your

data when you're importing your data set

in Power Query, so that's where M code

is required and these two things are

essentially done really well by Cloud,

right?

It handles time intelligence,

iterations, complex filter context. So,

it's all good.

Right? Second technical execution is

working on semantic modeling and this is

the most important part. For example, if

you have to create if you have multiple

tables, you will have to have

relationships, right? So, once you have

for example, geography related data, you

want to have hierarchies like

cities should come under state, states

should come under country, all those

things. So, hierarchies matter. Date

table matters. I mean, you should have a

date calendar table in your Power BI.

Star schema setup. So, if you're working

on multiple tables and you have

dimensions and fact, so this is where

the semantic modeling concepts come into

the picture and once you provide the

skill in Cloud Code, this is

predominantly handled quite nicely by

Cloud and various other AI tools.

Third is theme and visual formatting.

So, if you talk about JSON theme files,

visual level formatting, color palettes,

all those things, if for example, you

you have a particular set of brand guide

for your company, you can easily create

that JSON inside the skills and then the

Cloud will handle that in all the pages

it creates for Power BI.

So, the next technical execution over

here in tier one is custom visual

development. So, if we talk about PV PBI

viz projects, for example, if you're

working on custom reports, so you want

to create a custom visual which is not

as a native visual inside Power BI, you

can create that. So, you can have HTML

code and then add that in PBI viz and

that's what makes it even more

interesting because you can have more

you know,

whatever the visuals you were not able

to create earlier, now you can do that

using

HTML code that you can generate through

various AI and that's also D3 or React

is also feasible and you can build a

capability layer and formatting pane

options. So, this is also really nice.

And finally, we have one more skill in

technical execution, that's deployment

automation. So, if you are working on

for example, a project and multiple team

members are there working on the same

project and you want to connect through

API endpoints as well, so this is easily

handled by

AI. So, you can easily provide the

information for example, get related

information regarding REST API scripts,

CI/CD pipelines. So, this is all taken

care by

AI. So, this is

these are some of the skills where AI is

really strong and we need to cope up and

we have to match so that we understand

what AI is, you know, bringing in so

that we know what's happening and we can

improvise on that. All right, so now

let's talk about tier two of framework

where AI helps but human leads.

So, first point in this particular tier

is performance optimization and this is

the most important element because you

are creating reports. If you're

refreshing it and the business user is

not able to utilize it on time, then

there is no uh need to create any report

per se. So, it is really important that

we are focusing on performance

optimization. So, what AI can do is it

can suggest you differences between

import versus direct query versus

composite. So, what exactly you need to

they can suggest. It can write

aggregation tables. It can recommend

partitioning strategies. But,

what AI cannot tell you

these things. We know the actual data

volume, right? So, we know this. We know

the refresh SLAs. We know the gateway

infrastructure. We know the

organizational tolerance for latency.

So, we have so much of context and we

cannot provide everything inside AI,

right? So, these are some of the things

we still need to know. That's the reason

performance optimization in my opinion

is still a gray zone and we need to come

into the picture.

Second point is RLS

design. This is important. So, AI can do

these things. AI can write your DAX

security filter expression. So, you

know, we use user principal name for RLS

row level security, right? For all the

path-based hierarchy filters. But, what

AI can not do as of now. I mean, we need

to, you know, come into the picture for

this. Designing the organization level

access hierarchy from scratch. So, let's

say let's assume we have already created

this and we provide this context to AI,

but still we need to have we have so

many more information that we can't

provide to AI such as, you know, knowing

which cost center share data and who

signs off the access policy, all those

things which are quite internal in the

office, right?

Now, third and the final skill that is

there in gray zone is data quality

validation. So, if we talk about data

quality validation, AI can run your

statistical checks, it can find nulls,

outliers, all those things. AI is really

good in this, right? But, what I feel is

AI still still can't do these things and

human must know when a revenue figure

that passes every check is still wrong.

Let's assume you're working on a project

and you don't have any idea about that

particular data set and you create a

report and then you share that report

with the actual business owner and then

that business owner provide you that,

"Hey, these figures doesn't match and

they are way off the correct figures."

Because they know the context, right?

And you don't know the context. That's

where humans come into the picture and

this is where I would say it still comes

in tier two. So, data quality validation

is a important aspect. Though, when I

created end-to-end Power BI reporting

through Cloud, my data quality was 100%

accurate and that was so good, but it's

not necessary that it comes that way all

the time. All right, so now let's go to

tier three of this framework and in my

opinion, this is the most important tier

and we should focus on that.

So, first of all, tier three is human is

irreplaceable as of now.

So, what are the skills? So, number one

is requirements discovery.

So,

business users don't say I need a time

intelligence measure. So, it's not like

your business

owner or stakeholders are saying, "Give

me the time time intelligence measure."

It is us that we understand this is what

is required. So, developer sits in the

room, asks the right questions and

translates chaos into a spec. So, the

most important part is asking the right

questions. So, for example, you are on

Cloud and you mention that I want these

four pages and one page should have,

let's say, overall

overview of the data set. Second page is

on geography, third is product analysis,

fourth is inventory. So, you have

provided all the context and it provides

you those pages. But, it don't you

you've not asked the right questions.

So, it will assume and do

uh behind the scenes for you, right? So,

asking the right questions is still a

very important part and this is what

resonates with the context. And that's

what we call prompting as well. Right?

Second is tribal knowledge and business

context. And

So, for example, if you're working on,

let's say, retail data set, that's what

we did last time as well. So, if you are

predominantly a beginner and you don't

know anything about your business,

right? You will not be able to ask the

right questions, first of all. Then, you

won't be able to understand your data.

So, why that one cost center is

excluded? Why Q3 2022 is anomalous? You

know, which vice president will reject a

report if it uses green? So many

questions are there, so much of context

is there. You can't put in everything in

Cloud or any other AI tool, right?

Third is iterative co-design.

So, what iterative co-design means? So,

seeking a stakeholder's face when a

layout doesn't click. For example, you

are sitting in the presentation room and

you're discussing with the stakeholder,

right? So, you can have different

questions and you can pivot and, you

know, brainstorm and come to a

particular conclusion that this is what

we want, right? So, this kind of

brainstorming session

is not there in AI. I mean, you can do

it yourself, but this kind of

brainstorming is missing. Prototyping on

a whiteboard. So, this is what a

traditional way of

designing a thing was there, right?

Reading in the room. No AI can attend a

discovery workshop and adapt to what the

group energy is telling it. So, this is

something I feel like this is what we

call people element is there and it is

irreplaceable.

These are four more skills which I feel

are irreplaceable. So, which is on

trust, governance, and the judgment

call. So, first is stakeholder trust

adoption. So, for example, you have

created a report and you have been

working with your boss for so many

years. Now, he trusts you that you have

been doing good and whatever you deliver

would be accurate. Now, the same thing

has been done through AI. Now, there

would be some credibility issue. Boss

might not be able to trust what has been

generated through AI. It will take some

time, but definitely this is something

which is important. Same goes with

cross-team orchestration. So, aligning

with different departments and there

would be so much of politics, patience,

and, you know, interpersonal trust. So,

for example, you want to, you know,

adapt on this particular methodology,

but the other department wants to go

with the other one. So, it is so hard to

achieve. So, that's the reason humans

are irreplaceable in these aspects as

well.

And this is one of the most important

element, I would say, governance and

data strategy. Security becomes so

important when we are talking about AI.

So, now it's a brutal truth that if even

if you're working on Power BI reporting,

you have to use AI. I mean, if you're

not using AI, then your job is on stake.

So, if you're using AI plus Power BI,

then security

is very important. Security becomes so

important that every organization, you

know, whenever they are interviewing

someone on mid-level

data analyst or senior level, so they're

looking for someone who knows how to

work on governance and security.

And finally, the judgment on done. You

know, knowing when a report is good

enough to ship versus when it is

confused, when it will confuse the

audience. So, this requires the existing

context and knowledge. So, this is where

humans are still irreplaceable. All

right. So, now that we know all these

three tiers, then what should we focus

on?

So, this is where the real shift AI is

creating. So, if you see on the

left-hand side,

these were the things that we used to

focus on, like ours writing DAX from

scratch, manual theme JSON editing,

scripting, deployment pipelines,

debugging M code. So, we still have to

do it, but, you know, start with AI. Let

AI create these things for you and then

you have to debug. So, more time is

spent on reviewing AI-generated DAX,

describing the look and iterating,

reviewing and approving generated

scripts, specifying the transformation

and validating the output. So, I believe

this is the shift that is happening.

Now, the future is clear. Even if you're

working as a Power BI developer, you

have to use AI to enable yourself so

that you are more productive, you are

faster, you are able to create better

reports. So, now when you're using AI,

this becomes very important.

AI is only as good as the prompt it

receives,

right? Writing a good prompt requires

the same skills the developer always

had. So, what it means is So, whenever

you're working on Power BI reporting,

you're working on basic skills, right?

Like understanding whatever is required

by the stakeholder. So, that is the

translation.

So, let's say you're sitting in the

board room and you're discussing with so

many stakeholders what kind of reporting

that they want, what kind of overall

structure they want in data strategy,

right? So, when you are discussing with

them, what kind of questions you're

asking them, what kind of analysis you

they expect from you and then what you

understand, right? So,

you both should be on the same page.

That's where the same concept applies in

prompting, right? So, when you're using

AI, the prompt is really important. So,

it can be very small. You don't need to

have a very very big prompt, but

even small prompt can give you better

results, but it should be really precise

and good.

Right? Same concept over here. Bad spec

is bad output. Same concept. So, this is

where I believe uh we can derive more

value. Domain expertise compounds. So,

let's say you're working in the finance

sector

>> [snorts]

>> and you're working on creating Power BI

reports. So, I believe if you have good

understanding of how finance works, how

accounting works, then you can become

successful Power BI developer.

Otherwise, it is really hard to just

continue with your technical skills.

Same goes with any other

domain. Even if you're working on mining

sector, for example, and you're working

on Power BI reporting, you need to have

in-depth knowledge and you have to have

those skills so that you can improvise

on your reporting structure. Finally,

this is my opinion. So, I mean, you must

have seen so many memes on this. AI

won't replace people, but people who

would be using it will replace you. So,

this is actually true. So, if you're not

using AI as a developer and you're just

working as a traditional developer, just

opening Power BI report, dragging

dropping the visuals, creating data

model, that's great. I mean, you should

know the concepts. If you don't know the

concepts, you won't be able to survive

with just AI. Right? AI can help you if

you know the concepts. So, my two cents

would be understand all the concepts,

work on your data modeling, work on

concepts of DAX. You don't need to learn

everything in DAX, but at least the

concepts. Work on the concepts of how

Power Query is done. So, how how to

utilize that, how cleaning is done, how

manipulation is done, all those things,

right? So, once you have understood the

concepts, then start utilizing AI to

enhance your reporting. So, that's how I

would say this approach should be.

All right. So, this was it in this

particular video. If you have any

questions, please let me know in the

comments below and I would be trying to

help you out. Thanks a lot for watching

this video.