AI to Automate Boring Work
Key Points
- The podcast “AI in Action” introduces IBM’s AI experts, Jessica Rockwood and Morgan Carroll, who discuss how AI can take over repetitive, time‑consuming tasks that most employees dislike.
- Jessica explains that automating data‑preparation and pre‑processing with AI frees up hours each week for strategic, high‑level thinking and decision‑making.
- Morgan shares that even routine activities like drafting emails can be streamlined by AI, turning mundane work into “super‑powers” for productivity.
- The conversation highlights that successful AI adoption requires not just new tools, but also changes to business processes, infrastructure, and organizational culture to make AI work for the organization today.
Sections
- Turning Boring Tasks into AI - The hosts introduce an IBM podcast episode that explores how AI can automate tedious work and guides listeners on implementing AI solutions in business processes.
- AI‑Powered Drafting Saves Time - The speaker explains how AI can quickly generate email drafts, code snippets, and research summaries, speeding up work and helping meet deadlines.
- Beyond Chatbots: AI Assistants for All - The speaker argues that AI assistants now perform actions—not just converse—and that foundation models enable even small companies to adopt powerful intelligence with minimal proprietary data.
- Unified AI Assistant Integration - The speaker explains how AI assistants aggregate data from work and personal calendars, email, and other services to provide daily summaries and proactive task handling, and discusses building such integrations using tools like Watsonx Assistant.
- Choosing the Right LLM - The speakers discuss how model size and parameter count affect performance, emphasizing task‑specific suitability, cost‑benefit trade‑offs, and the need for human judgment when selecting AI tools.
- AI Assistants and Eliminating Grunt Work - The speakers argue that AI will become a universal, supportive assistant—requiring new educational awareness, automating repetitive “grunt work,” and freeing humans for creative, strategic tasks—illustrated through IBM’s design‑thinking practice of empathy mapping to pinpoint replaceable duties.
- AI Insights: Time, Models, Future Careers - The speaker highlights that leveraging AI to reclaim time, selecting the right model for each task, and acquiring AI fluency will be essential for all future careers.
Full Transcript
# AI to Automate Boring Work **Source:** [https://www.youtube.com/watch?v=DGO60Zrdle8](https://www.youtube.com/watch?v=DGO60Zrdle8) **Duration:** 00:19:33 ## Summary - The podcast “AI in Action” introduces IBM’s AI experts, Jessica Rockwood and Morgan Carroll, who discuss how AI can take over repetitive, time‑consuming tasks that most employees dislike. - Jessica explains that automating data‑preparation and pre‑processing with AI frees up hours each week for strategic, high‑level thinking and decision‑making. - Morgan shares that even routine activities like drafting emails can be streamlined by AI, turning mundane work into “super‑powers” for productivity. - The conversation highlights that successful AI adoption requires not just new tools, but also changes to business processes, infrastructure, and organizational culture to make AI work for the organization today. ## Sections - [00:00:00](https://www.youtube.com/watch?v=DGO60Zrdle8&t=0s) **Turning Boring Tasks into AI** - The hosts introduce an IBM podcast episode that explores how AI can automate tedious work and guides listeners on implementing AI solutions in business processes. - [00:03:04](https://www.youtube.com/watch?v=DGO60Zrdle8&t=184s) **AI‑Powered Drafting Saves Time** - The speaker explains how AI can quickly generate email drafts, code snippets, and research summaries, speeding up work and helping meet deadlines. - [00:06:05](https://www.youtube.com/watch?v=DGO60Zrdle8&t=365s) **Beyond Chatbots: AI Assistants for All** - The speaker argues that AI assistants now perform actions—not just converse—and that foundation models enable even small companies to adopt powerful intelligence with minimal proprietary data. - [00:09:12](https://www.youtube.com/watch?v=DGO60Zrdle8&t=552s) **Unified AI Assistant Integration** - The speaker explains how AI assistants aggregate data from work and personal calendars, email, and other services to provide daily summaries and proactive task handling, and discusses building such integrations using tools like Watsonx Assistant. - [00:12:22](https://www.youtube.com/watch?v=DGO60Zrdle8&t=742s) **Choosing the Right LLM** - The speakers discuss how model size and parameter count affect performance, emphasizing task‑specific suitability, cost‑benefit trade‑offs, and the need for human judgment when selecting AI tools. - [00:15:25](https://www.youtube.com/watch?v=DGO60Zrdle8&t=925s) **AI Assistants and Eliminating Grunt Work** - The speakers argue that AI will become a universal, supportive assistant—requiring new educational awareness, automating repetitive “grunt work,” and freeing humans for creative, strategic tasks—illustrated through IBM’s design‑thinking practice of empathy mapping to pinpoint replaceable duties. - [00:18:31](https://www.youtube.com/watch?v=DGO60Zrdle8&t=1111s) **AI Insights: Time, Models, Future Careers** - The speaker highlights that leveraging AI to reclaim time, selecting the right model for each task, and acquiring AI fluency will be essential for all future careers. ## Full Transcript
We all have something we hate about our job.
Those time sucking tasks that take you away from actually getting your work done.
I know you're thinking about them right now.
But what at the tasks that you dread could be handled not just by someone else,
but by AI.
Welcome to AI in action.
Brought to you by IBM.
I'm Albert Lawrence.
I'm here because I'm a learner. I'm a doer.
I look at a big picture, and I can't help but start asking exactly
how does it work on the inside?
On this podcast, I'm going to be joined by AI experts,
technologists and business leaders alike who are really going to help us
to get beyond the fury of AI and into how we actually put it into practice.
We're all starting to see more and more stories about AI powered outcomes,
but exactly how do I get to those outcomes for my business?
What are the actual steps involved in changing not only my
IT tools and infrastructure, but also my business processes and culture?
Mainly, how do I make AI
I work for me right now.
So let's do it, shall we?
Today I'm joined by Jessica Rockwood and Morgan Carroll.
Jessica is VP client engineering at IBM.
Welcome, Jessica.
Thank you.
Thanks for having me.
Glad that you're here.
And my other guest is Morgan Senior AI engineer in client engineering at IBM.
Welcome, Morgan.
I'm excited to be here.
I bet you're wondering how I chose you for today's episode.
Well, it's not because your work is boring.
Because it's not.
But it's because you work to make the boring stuff easy.
And IBM client engineering is all about making your AI dreams a reality.
So let's take a look at a few places where attention to detail, empathy,
and responsiveness could make or break us if we're being for real for real.
So first off, I'm very curious.
Look, people already don't have enough time to do everything
at the quality and the speed that we expect every day.
Work just seems to have gotten more complicated.
And I'm not alone in this.
So how do you use AI to do the boring stuff that you'd rather not?
Jessica, let's start with you.
For me, it always comes down to
what are the things I hate doing because they're repetitive.
Quite frankly, I don't have to use that many brain cells to do it.
It's all the preparatory work.
So every week I take a look at how are we doing as a business,
what are we doing in the team?
Are we making progress?
I can spend hours just getting data together
and trying to do, let's say, the first pre-processing.
If I can have I do that in a few minutes, that actually gives me
a few hours to do the critical thinking, to think strategically.
What are the next steps to take?
I was telling Morgan earlier, I would love that every time I look
at some data and I think, ooh, I wonder if there could be AI that goes and finds
the right data, does the processing, and gives me back some analysis.
I now have like super powers.
And that time is that's like the most valuable.
invaluable, yes.
And what about you, Morgan?
Okay, I hate to admit it, but writing emails.
I have a I'd rather be writing code.
I have got, like, five points.
I need to make it an email, and I'm like, okay, now how do I phrase this?
So it sounds appropriate?
No, I'm just going to let I do it. For me.
That's going to save me so much time.
And in addition, code generation actually, which is really interesting.
Typically I'm like, how do I write this function in Python?
Maybe
I don't want to go to the documentation, I don't want to read the documentation.
But with code generation I could say, hey, how do I do this?
How do I write this function in? But there it is.
I look like I already have a very clear idea of who I'm speaking with today. Now.
You just admitted you'd rather be writing code than writing emails. So.
Okay, we love it. You're in the right gig, then.
I think of Morgan's example about getting a first draft,
whether it's email or code or, let's say, a history project.
I was working with my daughter.
She needs to come up with a lot of different facts
to support a thesis, and she hates drafting anything
and leveraging a search engine, going out, pulling back
all sorts of forms of information really accelerated her progress on it.
She still had to assess what the search engines
found, because we all know there can be some fake news out there.
You need to kind of do a bit of analytics to understand the source of
what's come back, should you trust it?
But by being able to pull that together in seconds,
you actually can make a deadline.
And that's what matters.
We love it when they can make a deadline.
Awesome. Congratulations to your daughter.
Thank you.
Now, but when we're thinking
about these kinds of solutions that both you and your daughter are using.
I'm curious about the build behind them.
Can you take me a little bit into that, Morgan?
Yeah, definitely. Everything starts with the user, obviously.
So we need to think of what is the user experience going to be like.
So what we want to start with is sort of a conversational flow
like hello user, how are you come up with a persona maybe for the bot,
which is my favorite thing like Barry the bot, a little alliteration.
and then we want to gather data from the user, obviously.
And at some point we're going to call out to a large language
model, we're going to take all of this data and say, hey, here it is.
Do something with it.
Summarize this for me, get an answer, it's going to come back.
And then we present it to the user.
So it's overall a relatively simple process.
Okay I mean you made it sound really easy okay.
But but I know that it really can't be as easy as as you made it sound.
So I'm curious about what problems though the companies are coming up against
when they are trying to figure out how to make these a reality.
Well, so I think one of the biggest ones I see is how do you customize?
So I think what Morgan took us through is what I would call the standard. Right.
Like let's say 80% of the time, yeah, we can have just this back and forth flow.
But what happens when you have a problem or what happens when someone
maybe is going to ask a question of a virtual assistant It's never seen before.
We all like to think we're a little unique.
We're a little different than everyone else.
And that's, I think, where the challenges come.
It's like, how do you figure out for the edge cases to get the same response,
the same experience as if it was the common interaction?
That word interaction is really ringing for me right now,
because I notice that both of you are speaking of these as virtual assistants.
Nobody's saying chatbot. What? What?
Is that an intentional thing?
Yes, absolutely.
The reason we don't say chat bot anymore is because
we're not just chatting with the virtual assistant.
This technology is assisting us with various tasks.
So it's not just like, hey, give me this information.
It can, look up account information, maybe fill a prescription.
The options are endless, honestly.
And especially that taking actions associated with it.
So I think most of us would think about an assistant helps you do things.
They don't just talk to you.
Okay, so look, I think we can all agree that we could all use an assistant,
use some additional assistance. Right.
And that's whether we're a big company or whether we're a smaller company.
But what is the data foundation that you need in order to make AI I work for you.
Can the small companies access it just as much as the larger ones?
That's probably the thing I'm most excited about.
With generative AI, we're able to leverage foundation models,
which are built off of an incredibly large corpus of data.
And so now when you're the smaller company and you have a smaller amount of data to start with,
you take advantage of everything
that was already trained into the foundation model.
So we've lowered the bar.
It doesn't take terabytes of data.
In fact, it might be something as simple as 3 or 4 questions and answers.
And that alone can get you started.
So in thinking about that, if even the small companies can really tap
on in and use a lot of the intelligence that's already been formulated, the LLMs,
why is integration a relevant thing?
Can somebody just kind of like,
grab it all from one source and plug it on into their system?
Not necessarily.
So there are and integrations are my favorite topic to talk about.
So there are a number of different sources.
Maybe you have a database that has all of your customer information
or you've got an ordering systems.
I always like to use the flower shop example,
so maybe we have our inventory listed
in a certain place, customer information in a certain place, location information.
If we're doing deliveries,
you can't just kind of copy and paste that into your virtual assistant.
You have to work with integrations which connect your virtual assistant
to all of this data so that you can use it in your virtual assistant.
okay. Okay. So
you're reaching into different sources,
all in order to still connect with your one main task.
So even if we're just asking maybe one question,
the response might be pulling from several sources. Yes.
And I think it's partly about customizing.
So when you interact with a virtual agent, you want it to know who you are.
And to have all the right context.
And then I think the other integration
Morgan spoke to is when you're going to take that action, that workflow
might be in a different system or a different application.
And so it's really it's not just integrating the data,
it's integrating the tools or the actions you're going to take as well, making it
a seamless journey throughout.
So when I look at my phone, when I look at my devices,
I see that I've got all sorts of different like virtual assistants
I've got, whether it's like my calendar app over here
or whether I've got another scheduling app on over here.
How in the world can all of those different spaces
be woven together and integrate to just be one super duper virtual assistant?
I mean, again, as your assistant,
I've got my work calendar, I've got my personal calendar.
I need to know when to take my dog to his doggy play dates,
but I also need to know when I'm traveling for work.
I want to see everything in one location, so he or she is reaching out
to all of these different places and grabbing all of your data.
So it's connected to maybe my work calendar.
It's also connected to my personal calendar, it's connected to my email,
and it can consolidate all of this in just one place.
So if I say like, hey, what's going on today?
Okay, let's give you a summary.
I just pulled your email, pulled your calendar, checked on Mr. Hubble, the dog.
Now, we're going to put all this together
and say like, okay, here's a summary of your day.
And I think the other thing we're also seeing is
many of the applications
are starting to figure out how do we proactively build those integrations in.
I did a prescription online refill the other day, did it
through virtual assistant. It said it's done.
It then sent a text to me with the time to pick it up, which created an entry
on my calendar and a reminder to say, like, you got to still pick it up.
And so I think to your point, it's either from the assistant outwards
trying to engage and connect and integrate, but also as we build
these applications, we need to be thinking about those touchpoints
and how we proactive send information to be included.
I hear you, Jessica.
So but Morgan, how do you go about actively building these integrations?
Yeah. So let me give you an example for what I do specifically.
I will first build sort of a dialog in a tool that we have called watsonx Assistant.
So that's just the base like, hello, how are you?
How can I help you?
and then I will add something called extensions, which are the integrations.
So for each data source I might have a different extension.
For instance, to reach out to a calendar, I've got a calendar extension.
if I have a data repository with all my documents, that's another extension.
If I want to reach out to a large language model, another extension,
and then in the dialog itself in assistant, that's where we kind of
bring everything together and navigate, you know, how to guide the customer.
But when we are thinking about all of the different sources that are out there,
this is where I can start to get a little bit intimidated
from just keeping it real here.
How can you make sure that you're actually finding the right AI model, and how do you start?
First and foremost, it's about finding a trusted model.
So we know that there's an incredible amount out in the open source world.
There's a lot of proprietary models as well.
So first and foremost you want to find one that you trust.
and that's looking at what information is shared about the model.
What data was it trained on.
The same way that when you go to buy food at the grocery store,
you check the ingredient list, you look at the calorie count.
It's the same kind of thing you're assessing as you look at these models.
And then I think the second piece is to understand what type of actions
are you going to be taking.
So Morgan referenced earlier we talked about summarizing something.
Or maybe you're classifying these different models
typically are strongest at a certain set of activities or actions you might take.
And so it's about finding the right fit.
And then maybe last but definitely not least,
how big is the model and what's it going to cost you to use it.
Because you got to balance all those factors.
I wish I had unlimited funds, but I don't.
And so really trying to find the best mix across those dimensions.
Is it inaccurate to assume, though, that like the bigger the model, the better
that it's going to be for any business, regardless of what the industry is?
Definitely. Yeah.
Like Jessica was saying that different models are good at certain things.
So there's actually one that is really good at code generation.
I'm not going to use that for writing emails.
Otherwise my emails are going to be very weird.
and so there is the concept of parameters and I won't get too deep into it,
but how many parameters go into a large language model?
So I think what you were kind of saying is just because something has more
a larger number of parameters, it doesn't mean necessarily
that it's going to be better for every task out there.
Gotcha.
Okay. Still, it really does matter. It does.
And I think when you think about the size of the model,
sometimes I would say if something is going to cost
you twice as much and it's maybe only better in one case out of ten.
Is it the right balance.
Right. Yeah.
So there's really some real human discernment that needs to happen here.
There's a. Lot of work to be done.
To make sure you find the right fit. Yeah.
So that's of testing
lots of testing.
So let's talk about how this can actually go into practice as well.
Like I'm thinking specifically about so many of the professional tasks
that we do today that we just kind of we automatically think about it.
Right, because it's a part of our routine.
But some of those things are going to look really outdated when our grandkids
think about it. Right. They're going to laugh about it. Right.
So like, what is today's fax machine, for example?
And and what's today's email?
I think we're already seeing it even with the generation of our kids.
And I think it's only going to amplify when we get to our grandkids.
Really simple example.
To give you a sense, most kids today work only in online applications.
So my children do not understand the save button
because they've only ever worked in like Google Docs or some kind of online form.
So like it just isn't even a concept to them.
And I think what we're really going to see is this idea
that you have to talk to people to get something done.
It's going to sound like this really hokey thing, like, what do you mean?
You had to line up somewhere and you had to talk to human to accomplish something
They're going to be doing things from a phone.
That mobile device is going to just become central to everything.
And they'll be like, oh, you just press a button
or you just speak to it and everything happens behind the scenes.
I think that concept of manual to accomplish
anything is going to be a little bit of a thing of the. Past,
and I don't think the kids know what a floppy disk is when you refer to the save button,
but another thing is, you know,
if you're typing in a text message, you you've got autocomplete.
That's AI, like that's.
What I mean. But it's funny that you mention that because I feel as though even though
sometimes there can be a fear surrounding those two letters, A.I.
we are all using it in several different ways. G.P.S. like texting. Yeah.
Well, and I think you comment on fear.
I also think generationally it's interesting to watch because I do
think for our kids and our grandkids, those generation
AI is going to be so pervasive that they may not have the same degree of fear
of adopting it,
that maybe some of us do today because we know what it used to be like.
And so there's that.
There is that culture divide that you have to cross.
But as kids come up, as they're learning in school,
I think every career, no matter what you're in,
is going to require some level of knowledge about AI.
That's going to mean
they all come up with a set of awareness and education that none of us have had.
You mentioned career.
So I'm thinking now, how will this growth,
how will this development impact our work?
Oh, it's going to make everything easier for us.
it's that's why we call it an assistant.
It's not going to take an assistant cannot do what I do. Let's be honest.
But it's going to
help me with the things that are either just super repetitive
or maybe the things I'm not good at, like writing emails, things like that.
But it's definitely not going to take over. What I do.
I kind of hope that that phrase, grunt work is a thing of the past.
It becomes a word
that is just not used anymore, because those are the things that I can handle
and we're all going to be doing things that are creative, thought provoking, strategic, and let's go change the world.
Okay, well, you just brought up a word that I'm sure really made people's
ears perk up grunt work.
It's the thing. Like we all run away from that.
So how do you even start to diagnose what some of that grunt work is?
That I might be able to replace?
Yeah. So, we have a concept called design thinking here at IBM.
And something we do as part of that is called empathy mapping.
We put ourselves in the shoes of the user.
We do like interviews with them.
So we're going to see like what is their day to day?
What are the tasks that they're spending the most time on, and can we automate
those like you were talking about your kids earlier, doing the,
the research and all that.
I remember back in my day when we had dial up manually
having the first way for the internet to load
and then having to go and read through all of these different documents.
And I mean, let's think of a, customer service agent.
You know, they're like, how do they update a user's account information
or something like that? They're going to have to go
and read through the manuals, but that's grunt work.
We don't want to do that.
We can use. A.I., hey, go check in within a few seconds, get an answer.
And we know then, though, that people are evolving over time.
Like what you talk about back in my day, you know,
I would like to say I think today is still your day.
I'm not ready to give our days up, you know?
but I'm thinking, how are these
roles also going to evolve over time?
Right. Because I know since we're evolving, our roles have to as well.
What do you think that's going to look like?
So I think we're going to see a lot more time
being able to be spent on engaging with each other.
If we're not having to spend this time, you know, searching on things
or entering things into spreadsheet.
Now, we actually can spend the time face to face and interacting and engaging.
I think we're going to see a big focus on that.
I also think we're going to be having more opportunity to create.
We're going to have things that will help us bring to life what we envision, right?
But we have to spend the time on the big, the big ideas,
the big thoughts, the big visions, and I do.
I think we're going to see a lot of advancements
and a lot of ideas about how to make all of our lives better and more fulfilling.
Okay, now this has been such a rich conversation.
I wish that I could somehow grant us more time, with this, but,
I'm taking a few things away from this, so here are a few takeaways I've got.
Time is the most valuable resource, so using AI to free up your time is absolutely key.
"The bigger the model, the better." It's like, that's not a true statement at all.
So you got to pick the best model for your task.
And then looking to the future, every career is going to require
understanding of AI and honestly it will make them better and easier.
Does that sound about right?
Sounds pretty right.
Okay, cool. I think I passed the test today.
well, thank you both so much for being here.
Jessica, Morgan, this has been a complete blast.
That's it for this episode.
So thank you so much for listening. Thank you for watching.
But hey, look, there's a ton more where this came from.
We're going to be bringing you AI insights all throughout the season.
So stay tuned to this feed and we'll see you here again soon.