Latest / How I Work / Agents explained in 10 minutes (no jargon, no hype)
Transcript
- 0:00Speaker 1: Hey, just quickly from me before we get into it.
- 0:03Speaker 1: If you're using AI but still doing too much manually,
- 0:06Speaker 1: Neo and I have built something for you. It's called
- 0:08Speaker 1: the AI Agent boot Camp, and it's a four week
- 0:11Speaker 1: virtual program and you'll walk away with agents already built
- 0:14Speaker 1: and working. Spot's limited and all the details are in
- 0:18Speaker 1: the show notes. You've probably heard the word agents thrown
- 0:22Speaker 1: around a lot lately, and if you've been nodding along
- 0:26Speaker 1: while quietly having no idea what it actually means, you're
- 0:30Speaker 1: not alone. The AI companies have done a genuinely terrible
- 0:34Speaker 1: job of naming things agents, agentic, GPTs, projects, gems. It's
- 0:39Speaker 1: a lot of words for concepts that, once you understand them,
- 0:43Speaker 1: are actually pretty straightforward. In this episode, Neo and I
- 0:47Speaker 1: cut through all of that. We cover what an agent
- 0:50Speaker 1: actually is and why it's different from agentic AI, which
- 0:54Speaker 1: is a distinction that's important to know, and we get
- 0:57Speaker 1: into how both individuals and teams can start aren't using them,
- 1:01Speaker 1: from automating meeting minutes to building a knowledge agent that
- 1:04Speaker 1: fields your team's repetitive questions so that you don't have to.
- 1:08Speaker 1: By the end, you'll know exactly what an agent is
- 1:11Speaker 1: and have a clear sense of where to start building
- 1:13Speaker 1: one for yourself. Welcome to how IAI with me Doctor
- 1:23Speaker 1: Amantha Imba and Neo Applin, head of Inventium AI. Each
- 1:27Speaker 1: episode we share one practical way to use AI better
- 1:31Speaker 1: at work and in life. No fluff, no tech jargon,
- 1:35Speaker 1: just things you can use straight away. Neo, I want
- 1:39Speaker 1: to talk about agents today, and I'm sure there's lots
- 1:45Speaker 1: of people that know what an agent is, but then
- 1:47Speaker 1: there are an awful lot of people that are probably
- 1:50Speaker 1: quite confused as to what is an agent? What is
- 1:52Speaker 1: a gentic? Is there a difference? Is that the same thing?
- 1:55Speaker 1: How does it relate to my work life? So let's
- 1:58Speaker 1: start off with some definitions. What is an agent?
- 2:04Speaker 2: The first thing I'm going to say is the way
- 2:06Speaker 2: that these companies have named these things is terrible. There
- 2:09Speaker 2: are so many confusing names out there as projects as gpds,
- 2:12Speaker 2: there's agents, there's agentic and all those kind of things.
- 2:15Speaker 2: So here's how I like to look at it. An
- 2:18Speaker 2: agent is generally a thing that we will use. I'm
- 2:22Speaker 2: talking now about an agent, a GPT or a project
- 2:25Speaker 2: to help you to do something as a reusable expert.
- 2:29Speaker 2: In other words, help me to do the thing in
- 2:31Speaker 2: the same way every single time. And what it's not
- 2:35Speaker 2: is to do the entire thing. So it's not like
- 2:37Speaker 2: an entire workflow. It's just to do one thing. So
- 2:41Speaker 2: you might have an agent that maybe checks your work.
- 2:44Speaker 2: It might be an agent that potentially you drop in
- 2:46Speaker 2: the meeting transcript and will give you the meeting action
- 2:48Speaker 2: items and notes and all those kind of things the
- 2:50Speaker 2: way that you want your action items and notes to
- 2:52Speaker 2: be done.
- 2:53Speaker 1: So it's an agent, okay, And is that different from AGENTICAI.
- 3:00Speaker 2: To go to. Yes, So agentic is where it does
- 3:04Speaker 2: the things not in the same way every single time.
- 3:07Speaker 2: It's kind of got a goal and it may have
- 3:10Speaker 2: some tools and we'll figure out how it's going to
- 3:12Speaker 2: reach that goal using the tools, and you leave it
- 3:15Speaker 2: to that thing to decide what to do. So, for example,
- 3:18Speaker 2: you might have an urgentic thing which has an email
- 3:21Speaker 2: tool so it can email in and out and read
- 3:23Speaker 2: emails and things like that. It might have file access
- 3:25Speaker 2: so it can then access your files and so you
- 3:27Speaker 2: can then give it a goal. Hey, you do this thing,
- 3:29Speaker 2: and it might read emails, it might read files, it
- 3:31Speaker 2: might figure outs and things, might get attach a file
- 3:34Speaker 2: to an email and send it off, so it's like
- 3:36Speaker 2: it will decide what to do. Now, with agents, it's
- 3:39Speaker 2: more about you needed to be bound to do the
- 3:41Speaker 2: same thing the same way every single time, and it's
- 3:44Speaker 2: often a smaller thing to achieve in type for example,
- 3:48Speaker 2: like a Riunning style guide checkup. You know, it just
- 3:51Speaker 2: does one thing, does it really well? Does it reliably
- 3:53Speaker 2: every single time?
- 3:55Speaker 1: And today and for the next few weeks, we are
- 3:58Speaker 1: going to be focusing on a gents as opposed to
- 4:01Speaker 1: agentic AI. So when it comes to agents, depending on
- 4:06Speaker 1: what AI tool you're mostly using, can you took us
- 4:10Speaker 1: through like what the equivalents are where they we're in
- 4:13Speaker 1: co pilot versus Claude versus Chattipit versus Gemini.
- 4:17Speaker 2: Yeah, Before I do that, I just want to quickly
- 4:20Speaker 2: say you do not need to be a geeky rocket
- 4:23Speaker 2: scientist AI a kind of person to be able to
- 4:26Speaker 2: build an agent AGENTIC Yep. Maybe that's something very different workflows, yep. Maybe,
- 4:32Speaker 2: but this is all reasonably easy for just normal regular
- 4:35Speaker 2: human beings to be able to do so in the
- 4:37Speaker 2: different tools they are called in Copilot's called an agent.
- 4:41Speaker 2: That's one I'm going to start there because they've actually
- 4:43Speaker 2: called it that. In Chichipitt it's called a GPT. In
- 4:47Speaker 2: Gemini they're called gems, and in Claude generally one ones
- 4:53Speaker 2: we're talking about where we're talking about actually projects. Interestingly enough,
- 4:57Speaker 2: projects or another kind of term. Chechipt also has projects
- 5:00Speaker 2: as a similar kind of thing, and also there are
- 5:03Speaker 2: notebooks in a copilot as well. So these are all
- 5:07Speaker 2: different flavors of agents, and the thing that they've all
- 5:11Speaker 2: got in common is you can give it an instruction
- 5:13Speaker 2: in other words, a prompt, a small or a large prompt.
- 5:16Speaker 2: Here's what I want you to do. And you can
- 5:17Speaker 2: also give it some files so that it has some knowledge.
- 5:20Speaker 2: Maybe it's like a here's a template to use, or
- 5:22Speaker 2: it might be here's some knowledge that I need from you.
- 5:24Speaker 2: So it's instructions and a file or files, and that's
- 5:28Speaker 2: all you are giving your agent so it can do
- 5:31Speaker 2: the thing. So it doesn't have a bunch of tools
- 5:32Speaker 2: and it can't just go things by itself. And it's
- 5:35Speaker 2: also very responsive like you. It can only do things
- 5:37Speaker 2: when someone asks it to do the thing. It doesn't
- 5:40Speaker 2: do things at three in the morning. But yeah, so agents, GPTs,
- 5:44Speaker 2: projects and gems are generally the names of these things.
- 5:48Speaker 1: Let's talk about how individuals can use them, and then
- 5:51Speaker 1: I want to talk about how teams can use agents.
- 5:53Speaker 1: And for the purposes of today, we're just going to
- 5:56Speaker 1: use the term agents. So what are some of the
- 6:00Speaker 1: way that individuals can get value from creating an agent to.
- 6:05Speaker 2: Do the things that you do all the time just
- 6:07Speaker 2: to make your life better. So if you are that
- 6:10Speaker 2: unfortunately I've done this before, and it sucks to be
- 6:12Speaker 2: that person who has to get the meeting transcript and
- 6:14Speaker 2: then turn out the minutes and then check the minutes
- 6:16Speaker 2: and then email them out to people. If that's something
- 6:19Speaker 2: you do every single week or you've got multiple meetings,
- 6:21Speaker 2: it's great to build an agent to be able to
- 6:23Speaker 2: do that because it then speeds up your time and
- 6:25Speaker 2: all you got to do is check what it's done.
- 6:27Speaker 2: So it's like a really helpful small assistant there. You
- 6:31Speaker 2: might have something like a voice and style checkert if
- 6:36Speaker 2: you'd like. Maybe that's a company one, but maybe that's
- 6:38Speaker 2: even just yours, where you might say, hey, i've got
- 6:42Speaker 2: some text here, I've got to look after the website.
- 6:44Speaker 2: Maybe you're not a marketing kind of person and you've
- 6:47Speaker 2: given a couple of dot points and said, now turn
- 6:48Speaker 2: this into Vivee voice and style and sort of then
- 6:51Speaker 2: do things like that also good for like bulk emails
- 6:54Speaker 2: and things like that, just to make sure it's actually
- 6:56Speaker 2: hitting the right beats. So it's something you just do
- 6:58Speaker 2: small and often. Other ones might be research assistance. So
- 7:03Speaker 2: I've got one which is a research prompt builder. I
- 7:05Speaker 2: think we'll talk about that in another episode. But yeah,
- 7:09Speaker 2: when I'm doing researching, I do a lot of research
- 7:11Speaker 2: just to make sure I'm keeping up to date on
- 7:13Speaker 2: different industries and how AI is going and things like that,
- 7:16Speaker 2: and so I've got a research prompt buildup that helps
- 7:18Speaker 2: me to build those things up. So the things that
- 7:21Speaker 2: you do more than I'm going to say once or
- 7:23Speaker 2: twice a week every single week, ask yourself, is this
- 7:27Speaker 2: something that you can put into an agent? Because your
- 7:29Speaker 2: job is going to be so different to mine and Amantha's,
- 7:31Speaker 2: you'll figure out what you're going to do more often.
- 7:34Speaker 2: Maybe that's something you can build into an agent.
- 7:36Speaker 1: Now let's look at teams. How can teams get benefit
- 7:41Speaker 1: from agents similar.
- 7:44Speaker 2: Kind of things in order to make their lives better?
- 7:46Speaker 2: And so some of them is about the things that
- 7:49Speaker 2: you do that you have to do all the time
- 7:52Speaker 2: as a team. So maybe meeting minutes is really handy thing.
- 7:56Speaker 2: Maybe multiple people are doing meeting minutes. There's that, but
- 7:59Speaker 2: maybe there's all things like reports. So there's a way
- 8:02Speaker 2: that we produce those types of reports. And sometimes what
- 8:06Speaker 2: you want to do is have that same report being
- 8:10Speaker 2: kind of built the same way, even though it may
- 8:11Speaker 2: be built by different people. And here's the thing that
- 8:14Speaker 2: I might have a different style the way I'll do
- 8:15Speaker 2: my report, you'll do your style to the way you
- 8:17Speaker 2: do your report. It would be great to have an
- 8:19Speaker 2: agent that has like the approved way that it grabs
- 8:22Speaker 2: these data in and then crafts it and then puts
- 8:25Speaker 2: it into the right headings and things like that. So
- 8:27Speaker 2: the things that you and your team do often I
- 8:30Speaker 2: think they're really good things to do to build as
- 8:32Speaker 2: an agent. The other thing is the things you wish
- 8:35Speaker 2: you wouldn't do, So how can you offload some of
- 8:37Speaker 2: those things to AI? So that's another one. And the
- 8:41Speaker 2: third is the questions that you often have or that
- 8:45Speaker 2: you get as a team. So if you're in HR,
- 8:48Speaker 2: I imagine you get an awful lot of questions on
- 8:51Speaker 2: how do I change my bank details? When's the next
- 8:53Speaker 2: pay run? Those kind of questions. If you're in like
- 8:56Speaker 2: a customer service role, you're going to get lots of
- 8:59Speaker 2: questions on this thing is broken, how do I fix it?
- 9:01Speaker 2: And then you'll probably know a lot of that in
- 9:03Speaker 2: your brain. But wouldn't be better if there was an
- 9:06Speaker 2: agent that did a look up of your company procedures
- 9:09Speaker 2: or the FAQs and things like that. And of course
- 9:12Speaker 2: here's that the other thing, which is why are you
- 9:15Speaker 2: answering this questions? Wouldn't be possible to build an agent
- 9:19Speaker 2: that answers those questions for those people, so you don't
- 9:21Speaker 2: even have to get involved. So it's really useful for
- 9:23Speaker 2: those kind of things. So yeah, if it's often regular
- 9:26Speaker 2: you do it the same way as a team, or
- 9:28Speaker 2: people are pestering you and your team, then maybe what
- 9:31Speaker 2: you can do is create an agent feed at the
- 9:33Speaker 2: knowledge with the files and things like that so that
- 9:35Speaker 2: people can self service those answers.
- 9:38Speaker 1: So what now and I are going to be doing
- 9:40Speaker 1: over the next few weeks is we are going to
- 9:42Speaker 1: be digging deeper into agents. So we are going to
- 9:45Speaker 1: be unpacking things like, if you're a knowledge worker, what
- 9:49Speaker 1: are the must have agents for all knowledge workers. We
- 9:52Speaker 1: are going to go into some of the tasks that
- 9:54Speaker 1: you should never do manually again, we are going to
- 9:58Speaker 1: go into Neo's favorite AG, the one that he uses
- 10:02Speaker 1: pretty much every single day. And we're also going to
- 10:05Speaker 1: get into how can you build a knowledge agent that
- 10:07Speaker 1: answers every one oh one question like there was just
- 10:11Speaker 1: talking about, so that you don't have to So if
- 10:13Speaker 1: you want to go deeper into the world of agents,
- 10:16Speaker 1: make sure you have hit subscribe or follow, because we
- 10:19Speaker 1: will be sharing a lot more info about agents over
- 10:22Speaker 1: the coming weeks. How i AI was hosted by me
- 10:26Speaker 1: Amantha Imber and Neo Applan. A big thank you to
- 10:29Speaker 1: Martin Imba who does our sound editing, and Jim Rubio
- 10:32Speaker 1: for production support, and thank you to John Kilby who
- 10:36Speaker 1: composed the theme music.