February 11, 2025 • 15 mins
AI is transforming how businesses operate, but do you really know what an AI agent is and how to use one? In this episode ofBA Bites, we break it all down:
✅What AI agents are (and how they think!)
✅How they’re designed and how they make decisions
✅Real-world use cases—from chatbots to automation to predictive analysis
✅Examples of AI agents in action
✅How to set one up and start using AI in your business today
Whether you're a BA, product manager, or just AI-curious, this episode gives you the knowledge toleverage AI agents in your work—without the jargon!
🎧 Listen now and take your understanding of AI to the next level! 🚀
#BusinessAnalysis #ArtificialIntelligence #AI #BitesizeLearning
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:00):
AI agents are everywhere, from chat bots to automatic tools,
but many BAS don't fully understand their design, use
cases, and how to actually use them.
By the end of this episode, you'll know exactly what an AI
agent is, how it works, where it's used, and how you can
leverage it in business analysisand beyond.
(00:22):
And a Better Business Analysis Institute presence, the Better
Business Analysis Podcast with Kingsway and Walsh.
So what exactly is an AI agent? An AI agent is a system that
perceives its environment, makesdecisions and takes actions to
(00:49):
achieve a goal. So that's important.
So we're talking about perception, we're talking about
decision making and we're taking, we're talking about
actions to achieve something, which is a goal.
So this is great because we lovethis kind of talk in BA.
We love it's very process focused and it's very goal
focused. Now there are four types of a A
(01:13):
agents, generally speaking, number one is reactive agents.
And that's probably what you're used to or if you haven't really
swore the word of AI too much. So that's responding to inputs,
but they don't have any memory. OK.
And if you get into AI, you'll you'll realize that a lot of
these AI tools don't have what we call state.
(01:34):
So they don't remember what you asked it last time.
And ChatGPT is you use you the the one that we always talk
about. And that's got better at that.
And we'll know about your preferences, but only if you're
logged in. So an example of that would be a
a simple chat bot when you go toa website, OK, and that will be
dumping around for a while and everyone should recognize that
(01:57):
type. Reactive agents, the second type
are called goal based agents. So they use decision making
models to achieve a defined goal.
So that might be a recommendation system.
You want to have recommendationsfor a book that you enjoyed
based on other books you've read.
(02:19):
So Amazon might use that for Goodreads, which Amazon owns,
might use that to generate recommendations for you.
Then there's the learning agent.OK.
And these are ones which improveover time using data.
So that's really machine learning based automation and
check GBTS got better at that. I know that it remembers a few
(02:43):
aspects of me at more the more Iuse it scarily about what but
the length and the amount of information it stores on your
previous chat history and therefore your profile if you
liked is it it depends on the onthe kind of chat bot you're
using. And you'll usually find that a
(03:03):
lot of them, a lot of these chatbots aren't set up that way for
public use. And then the last type is
autonomous agents. So they work independently with
minimum minimal, I should say, human intervention.
So this is like AI powered customer support all the way
through. And so you've seen that again on
(03:26):
call centres that are kind of set up for that and the human is
really not involved unless there's an escalation point.
So they're those four base typesof AI agents.
Then we jump to the design of anAI agent.
So the key components in BA Talkare inputs, decision making,
(03:48):
outputs, and then learning an improvement.
OK. And that's the important one
here if you want to know how IAIis expanding.
So input we call perception. So it's how the agent gathers
data. And now I want to you to open
your eyes and, and look a littlebit wider here.
(04:08):
So we're talking about sensors, OK?
So the whole Internet of Things,IoT devices, APIs, text
learning. So that input is really
important. So it's not human necessarily
doing the input now. So this is where AI agents are
changing how they interface withthings, which is really
(04:29):
important from an architectural point of view.
And then there's processing, which is our decision making
point, and that's algorithms, machine learning models or rule
based logic. So a lot of early AI which we've
had for many, many years is kindof rule based.
And then we had algorithms and now we've got machine learning
models. Then there's the action or the
(04:52):
output. So the the what are we deriving
from this process? And that's how the AI agent
interacts with it's environment,EG responding to users or
triggering work flows. So both the input and output
have been kind of standardized now with AAI agent such that it
(05:13):
is an independent body and it can be kind of added on like
Lego blocks to a greater architecture for building these
kind of steps in our process, which are AIA agent steps which
do a certain thing. And they can be to do say an
Uber could have an AI agent, so it could manage the inputs.
(05:36):
So not human inputs and outputs to ordering an Uber
automatically for you. So like if you think we could do
that via humans or APIs, well, now the AI agent can do that
based on inputs and outputs. So it can connect.
And so it allows us to do scalable workflows across our
day-to-day life. And then the last bit is the
(05:56):
feedback loop. So learning and improvement and
that's how the agent improves through reinforcement or
retraining. And this is really, really
important. So we've talked about customer
satisfaction and we've talked about the fact that we take data
points from our customers. Now that was all instead of our
whole team doing looking at thatdata, we're feeding that back
(06:18):
into the AI agent. It's the idea is that it's going
to improve and get better. So your experience with the AI
agent last week may be differentto this week because it's
improved some factors based on real live data and the
architectures behind that. If you're interested, as there's
rule based versus machine learning versus hybrid.
(06:39):
And then there's agent frameworks like Open AI and GBT,
which is what you know, what we know is ChatGPT in terms of its
interface. There's kind of a, a lag chain
and then there's auto GBT. And there's, there's other agent
frameworks, but they're built specifically for what I talked
about that architecture for being able to connect to inputs
(07:02):
and output. So let's dive in to the use
cases for AI agents. We've talked about the fact that
AI agents does cover some of themost simple use cases that
you'll already be aware of. But I really want you to just
think about the kind of AI agents which is kind of being
termed this term has just come out to think about some of these
(07:27):
use cases where there isn't a human interface and there isn't
necessarily. The output doesn't necessarily
go back to a human, it goes to another AI agent.
OK, another linked kind of architecture.
So we look at BA kind of applications for AI agents
favorite, which is kind of automating the requirements
gathering process through AI chat bots.
(07:49):
So again, we're not talking to ahuman, we're using their
customer feedback. There's the AI powered
stakeholder sentiment analysis which we talked about last week,
which is quite important, which is really sentiment drives
decision making. So we understand psychology
then, you know, it doesn't matter what people say on their
face, the way they write things,language, body language can
(08:13):
generate sentiment and that's really important for us.
And that can be automated. And then of course, it's the AI
driven process automation. So really if you want to think
about AI agents conceptually, think about them in terms of
they are playing the role of each step in your process
diagram. OK, every time there's a, you
(08:34):
can reuse steps. So for example, sending an
e-mail could use Gmail AI agent.You might use that multiple
times throughout your workflow to send notifications or draft
emails and think about the AI agent being one of those steps
or notes. Other business use cases, of
course, are customer service bots, but getting much more
(08:58):
advanced in terms of recommending preferences based
on personal preferences. So it already knows what you
shop before it really knows yoursize, it already knows what your
taste is, and it's going to be, you know it's going to know you
more than you might know yourself in terms of data points
anyway. Then there's the AOI agents in
ecommerce. Again, it's the same kind of
(09:20):
situation. 1 might be your customer profile, 1 might be
your, if you like consumer profile.
And then there's AI in terms of data analysis and reporting.
And we haven't seen that get to that tipping point yet.
But you know, you, you won't be writing your own dashboards,
sorry. You might be, or you might be
crafting them. But AI agents are going to do
(09:42):
the vast majority of that work very, very soon, if not today in
your company. And of course the, we've talked
about this before, the use case in terms of project management,
in terms of risks and delays based on past projects.
And of course you never think ofpast projects, but AI agents can
look out and see what mistakes have been made in previous
(10:04):
projects and predict what risks and and costs you might incur.
So that's kind of some good use cases right now for you as ABA.
Let's just talk about some examples of specific AI agents.
And this is not a definitive list and we did touch on this
(10:25):
last week, so I'm not going to labour the point, but there's
obviously the chat bots and virtual assistants, which is
like ChatGPT, we can move on from that.
But automation agents like UI Path, RPA robots, they've been
around for ages to do repetitivetasks in your workflow.
So that's like an AI agent. That's a really good way of
looking at it. They're much more advanced than
(10:47):
just that one tool that does that Microsoft Copilot now on
top of Azure and Microsoft Fabric will do your business
insights. OK, and Microsoft really pushing
that. So that might be where we see
some major changes in dashboarding in the future.
(11:08):
And then there's kind of auto GPT baby AGI, which are a non
autonomous. So they will just problem solve
that was problem solved. You gotta give them a problem
come back. I'll give you some solutions to
some of those problems. So they're pretty cool.
(11:28):
But I have seen in addition to that, I think the idea is that
you know how we have an app model now.
And this is how I've this is my interpretation.
So please come and, and, and comment on this post if you had
a different interpretation. But the way I've looked at AI
agents or I've convinced myself this is the model in my brain is
(11:50):
that, you know how we had the app model, everything was an app
and we still kind of have that. So you know, Uber, you go to one
app to do 1 task, one goal. Gmail, you go to do your e-mail.
It's the same concept but without the UI aspects.
So you've still got the UI obviously, for those use cases
where we want and tracked, but the AOI agent can send can have
(12:14):
inputs and outputs to that use case to that one app if you
like, quote UN quote, and that'sthe agent.
So the AOI agent is can take in information and output things
without having necessary the UI.So what do what would that mean?
We talked about Uber before. So in an example today, you
(12:35):
might travel for business. So you might travel to a country
through a travel agent. And then when you get to the
other side, you then load the Uber app, you go into your
business account because you want to charge it back as an
expense and you call the Uber because you've just arrived and
finish your trip and it gets charged back to your business.
(12:55):
So that point where the I had todo the Uber steps, if you like,
log in to the app, you know, to the rest of it, that could all
be automated and there wouldn't be any steps for me.
So I would just land. And maybe there would need to be
a trigger point, of course, which is why it's BAM process.
(13:16):
The trigger point could just simply be I've landed in that
country. Maybe there is simply an order
now step and the AI agent would automatically know where I'm
going based on my based on another input, which would be my
inventory. So it knows my hotel details and
it would automatically charged so I wouldn't have to do the
(13:37):
interface side of things. Maybe it's just order Uber now,
for example, and it's just confirming you were ordering
Uber now. I might even do a push
notification and say, oh, we've seen you've arrived in San
Francisco. Would you like to order the Uber
now? Or when would you like to order
the Uber? You could go and it already
knows the details about where you're going.
So it plugs those in. So what I mean, So it might use
(13:57):
APIs to do that or connectors, achain connector to other AI
agents. Maybe there's one that manages
the whole trip, which is the travel agent, AI agent.
So there we go. So that's the good example there
of how it may be used. Now finally, I'm just going, if
(14:18):
you're interested in playing around with AI agents
specifically, not just AI, then you can load some of these tools
up locally. So you can load them up on cloud
hosted for $1020.00 a month US general generally, or you can
load them up locally. And I would suggest that you do
(14:40):
try. And if you're a technical BA,
maybe you should try and configure an agent so you just
understand that to do one step, maybe it's to respond to your
emails or draft your emails for you responses so you can check
them. And just we'll, we'll just
you'll just end up with a whole lot of draft emails that you can
then check. So my final thoughts on this is
(15:01):
that AI agents specifically, notjust AI, are powerful tools for
automation, analysis and decision making.
And you can chain them together.I think you should experiment
with AI and you'll be a work anyway.
But now think about AI agents for doing one thing.
Try and build a simple AI agent using free tools.
(15:22):
Chat DBT will do open AI you cando run it locally as well and
just do one thing like automate your replies on your e-mail.
I hope you learned something about AI agents this week and I
will see you next week.
AI agents are everywhere, from chat bots to automatic tools,
but many BAS don't fully understand their design, use
cases, and how to actually use them.
By the end of this episode, you'll know exactly what an AI
agent is, how it works, where it's used, and how you can
leverage it in business analysisand beyond.
(00:22):
And a Better Business Analysis Institute presence, the Better
Business Analysis Podcast with Kingsway and Walsh.
So what exactly is an AI agent? An AI agent is a system that
perceives its environment, makesdecisions and takes actions to
(00:49):
achieve a goal. So that's important.
So we're talking about perception, we're talking about
decision making and we're taking, we're talking about
actions to achieve something, which is a goal.
So this is great because we lovethis kind of talk in BA.
We love it's very process focused and it's very goal
focused. Now there are four types of a A
(01:13):
agents, generally speaking, number one is reactive agents.
And that's probably what you're used to or if you haven't really
swore the word of AI too much. So that's responding to inputs,
but they don't have any memory. OK.
And if you get into AI, you'll you'll realize that a lot of
these AI tools don't have what we call state.
(01:34):
So they don't remember what you asked it last time.
And ChatGPT is you use you the the one that we always talk
about. And that's got better at that.
And we'll know about your preferences, but only if you're
logged in. So an example of that would be a
a simple chat bot when you go toa website, OK, and that will be
dumping around for a while and everyone should recognize that
(01:57):
type. Reactive agents, the second type
are called goal based agents. So they use decision making
models to achieve a defined goal.
So that might be a recommendation system.
You want to have recommendationsfor a book that you enjoyed
based on other books you've read.
(02:19):
So Amazon might use that for Goodreads, which Amazon owns,
might use that to generate recommendations for you.
Then there's the learning agent.OK.
And these are ones which improveover time using data.
So that's really machine learning based automation and
check GBTS got better at that. I know that it remembers a few
(02:43):
aspects of me at more the more Iuse it scarily about what but
the length and the amount of information it stores on your
previous chat history and therefore your profile if you
liked is it it depends on the onthe kind of chat bot you're
using. And you'll usually find that a
(03:03):
lot of them, a lot of these chatbots aren't set up that way for
public use. And then the last type is
autonomous agents. So they work independently with
minimum minimal, I should say, human intervention.
So this is like AI powered customer support all the way
through. And so you've seen that again on
(03:26):
call centres that are kind of set up for that and the human is
really not involved unless there's an escalation point.
So they're those four base typesof AI agents.
Then we jump to the design of anAI agent.
So the key components in BA Talkare inputs, decision making,
(03:48):
outputs, and then learning an improvement.
OK. And that's the important one
here if you want to know how IAIis expanding.
So input we call perception. So it's how the agent gathers
data. And now I want to you to open
your eyes and, and look a littlebit wider here.
(04:08):
So we're talking about sensors, OK?
So the whole Internet of Things,IoT devices, APIs, text
learning. So that input is really
important. So it's not human necessarily
doing the input now. So this is where AI agents are
changing how they interface withthings, which is really
(04:29):
important from an architectural point of view.
And then there's processing, which is our decision making
point, and that's algorithms, machine learning models or rule
based logic. So a lot of early AI which we've
had for many, many years is kindof rule based.
And then we had algorithms and now we've got machine learning
models. Then there's the action or the
(04:52):
output. So the the what are we deriving
from this process? And that's how the AI agent
interacts with it's environment,EG responding to users or
triggering work flows. So both the input and output
have been kind of standardized now with AAI agent such that it
(05:13):
is an independent body and it can be kind of added on like
Lego blocks to a greater architecture for building these
kind of steps in our process, which are AIA agent steps which
do a certain thing. And they can be to do say an
Uber could have an AI agent, so it could manage the inputs.
(05:36):
So not human inputs and outputs to ordering an Uber
automatically for you. So like if you think we could do
that via humans or APIs, well, now the AI agent can do that
based on inputs and outputs. So it can connect.
And so it allows us to do scalable workflows across our
day-to-day life. And then the last bit is the
(05:56):
feedback loop. So learning and improvement and
that's how the agent improves through reinforcement or
retraining. And this is really, really
important. So we've talked about customer
satisfaction and we've talked about the fact that we take data
points from our customers. Now that was all instead of our
whole team doing looking at thatdata, we're feeding that back
(06:18):
into the AI agent. It's the idea is that it's going
to improve and get better. So your experience with the AI
agent last week may be differentto this week because it's
improved some factors based on real live data and the
architectures behind that. If you're interested, as there's
rule based versus machine learning versus hybrid.
(06:39):
And then there's agent frameworks like Open AI and GBT,
which is what you know, what we know is ChatGPT in terms of its
interface. There's kind of a, a lag chain
and then there's auto GBT. And there's, there's other agent
frameworks, but they're built specifically for what I talked
about that architecture for being able to connect to inputs
(07:02):
and output. So let's dive in to the use
cases for AI agents. We've talked about the fact that
AI agents does cover some of themost simple use cases that
you'll already be aware of. But I really want you to just
think about the kind of AI agents which is kind of being
termed this term has just come out to think about some of these
(07:27):
use cases where there isn't a human interface and there isn't
necessarily. The output doesn't necessarily
go back to a human, it goes to another AI agent.
OK, another linked kind of architecture.
So we look at BA kind of applications for AI agents
favorite, which is kind of automating the requirements
gathering process through AI chat bots.
(07:49):
So again, we're not talking to ahuman, we're using their
customer feedback. There's the AI powered
stakeholder sentiment analysis which we talked about last week,
which is quite important, which is really sentiment drives
decision making. So we understand psychology
then, you know, it doesn't matter what people say on their
face, the way they write things,language, body language can
(08:13):
generate sentiment and that's really important for us.
And that can be automated. And then of course, it's the AI
driven process automation. So really if you want to think
about AI agents conceptually, think about them in terms of
they are playing the role of each step in your process
diagram. OK, every time there's a, you
(08:34):
can reuse steps. So for example, sending an
e-mail could use Gmail AI agent.You might use that multiple
times throughout your workflow to send notifications or draft
emails and think about the AI agent being one of those steps
or notes. Other business use cases, of
course, are customer service bots, but getting much more
(08:58):
advanced in terms of recommending preferences based
on personal preferences. So it already knows what you
shop before it really knows yoursize, it already knows what your
taste is, and it's going to be, you know it's going to know you
more than you might know yourself in terms of data points
anyway. Then there's the AOI agents in
ecommerce. Again, it's the same kind of
(09:20):
situation. 1 might be your customer profile, 1 might be
your, if you like consumer profile.
And then there's AI in terms of data analysis and reporting.
And we haven't seen that get to that tipping point yet.
But you know, you, you won't be writing your own dashboards,
sorry. You might be, or you might be
crafting them. But AI agents are going to do
(09:42):
the vast majority of that work very, very soon, if not today in
your company. And of course the, we've talked
about this before, the use case in terms of project management,
in terms of risks and delays based on past projects.
And of course you never think ofpast projects, but AI agents can
look out and see what mistakes have been made in previous
(10:04):
projects and predict what risks and and costs you might incur.
So that's kind of some good use cases right now for you as ABA.
Let's just talk about some examples of specific AI agents.
And this is not a definitive list and we did touch on this
(10:25):
last week, so I'm not going to labour the point, but there's
obviously the chat bots and virtual assistants, which is
like ChatGPT, we can move on from that.
But automation agents like UI Path, RPA robots, they've been
around for ages to do repetitivetasks in your workflow.
So that's like an AI agent. That's a really good way of
looking at it. They're much more advanced than
(10:47):
just that one tool that does that Microsoft Copilot now on
top of Azure and Microsoft Fabric will do your business
insights. OK, and Microsoft really pushing
that. So that might be where we see
some major changes in dashboarding in the future.
(11:08):
And then there's kind of auto GPT baby AGI, which are a non
autonomous. So they will just problem solve
that was problem solved. You gotta give them a problem
come back. I'll give you some solutions to
some of those problems. So they're pretty cool.
(11:28):
But I have seen in addition to that, I think the idea is that
you know how we have an app model now.
And this is how I've this is my interpretation.
So please come and, and, and comment on this post if you had
a different interpretation. But the way I've looked at AI
agents or I've convinced myself this is the model in my brain is
(11:50):
that, you know how we had the app model, everything was an app
and we still kind of have that. So you know, Uber, you go to one
app to do 1 task, one goal. Gmail, you go to do your e-mail.
It's the same concept but without the UI aspects.
So you've still got the UI obviously, for those use cases
where we want and tracked, but the AOI agent can send can have
(12:14):
inputs and outputs to that use case to that one app if you
like, quote UN quote, and that'sthe agent.
So the AOI agent is can take in information and output things
without having necessary the UI.So what do what would that mean?
We talked about Uber before. So in an example today, you
(12:35):
might travel for business. So you might travel to a country
through a travel agent. And then when you get to the
other side, you then load the Uber app, you go into your
business account because you want to charge it back as an
expense and you call the Uber because you've just arrived and
finish your trip and it gets charged back to your business.
(12:55):
So that point where the I had todo the Uber steps, if you like,
log in to the app, you know, to the rest of it, that could all
be automated and there wouldn't be any steps for me.
So I would just land. And maybe there would need to be
a trigger point, of course, which is why it's BAM process.
(13:16):
The trigger point could just simply be I've landed in that
country. Maybe there is simply an order
now step and the AI agent would automatically know where I'm
going based on my based on another input, which would be my
inventory. So it knows my hotel details and
it would automatically charged so I wouldn't have to do the
(13:37):
interface side of things. Maybe it's just order Uber now,
for example, and it's just confirming you were ordering
Uber now. I might even do a push
notification and say, oh, we've seen you've arrived in San
Francisco. Would you like to order the Uber
now? Or when would you like to order
the Uber? You could go and it already
knows the details about where you're going.
So it plugs those in. So what I mean, So it might use
(13:57):
APIs to do that or connectors, achain connector to other AI
agents. Maybe there's one that manages
the whole trip, which is the travel agent, AI agent.
So there we go. So that's the good example there
of how it may be used. Now finally, I'm just going, if
(14:18):
you're interested in playing around with AI agents
specifically, not just AI, then you can load some of these tools
up locally. So you can load them up on cloud
hosted for $1020.00 a month US general generally, or you can
load them up locally. And I would suggest that you do
(14:40):
try. And if you're a technical BA,
maybe you should try and configure an agent so you just
understand that to do one step, maybe it's to respond to your
emails or draft your emails for you responses so you can check
them. And just we'll, we'll just
you'll just end up with a whole lot of draft emails that you can
then check. So my final thoughts on this is
(15:01):
that AI agents specifically, notjust AI, are powerful tools for
automation, analysis and decision making.
And you can chain them together.I think you should experiment
with AI and you'll be a work anyway.
But now think about AI agents for doing one thing.
Try and build a simple AI agent using free tools.
(15:22):
Chat DBT will do open AI you cando run it locally as well and
just do one thing like automate your replies on your e-mail.
I hope you learned something about AI agents this week and I
will see you next week.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.