The Difference Between AI and Automation with Tim Lewis | 115

What is the difference between AI and automation, and

how can you use AI and automation in your day to day business?

In this episode, we are talking about AI and automation.

You'll learn the difference between AI and automation.

The simple way that AI and automation can help you day to day.

Plus, you learn the benefits of locally hosted AI over cloud-based AI tools

and stick around to the end where we discuss the fear of AI replacing you.

Welcome to Architecture Business Club, the show that helps you build

a better business in architecture so you can enjoy more freedom,

flexibility, and fulfillment.

I'm your host, John Clayton, and if you're joining us for the first time, don't

forget to hit the follow or subscribe button so you never miss another episode.

We we're joined by Tim Lewis, a long-term podcaster and content creator.

Tim's background is in books and publishing, but he's recently deep

dived into AI automation and AI image and video generation, and how it

can be used to help small business owners in their day-to-day work.

So head over to tim lewis ai.com to learn more about Tim, or

click the link in the show notes.

So Tim, we are going to talk about AI and automation so that architects and other

small business owners can start using it in their day-to-day work and hopefully

make their life a little bit easier.

I think a good place to start would be what's the difference

between AI and automation tools?

They are not the same thing.

And people have, they are related, but they are not the same thing.

So in the simplest way to think about AI is it's basically a computer

program that acts in this kind of a simple, intelligent way, like a human.

So this is not your arithmetic.

Add two numbers together kind of task.

This is like, here's a photograph with a dog and a cat in each tell tell which

one is a dog and which one's a cat.

Kind of simple humanlike tasks that AI can do.

And this is a relative, well, it's not that new, but in terms of since chat,

GBT came in, which was the text-based one for doing conversations and creating.

Like messages.

So that's what AI is.

Now.

Automation is just has been around for a while as well, and that is

where you create what, depending on the program you use, they're like

a flow or a set of instructions.

So you could have something like, I receive an email from this person,

put a entry into a spreadsheet, send an email to a different person.

Uh, and that kind of technology's been around for years.

If you've ever heard of Zapier or NA 10 or any of these solutions, where it's

got superpowered is obviously you can put in an AI task in the middle of that now.

So a very simple example I give where AI has improved things is, let's say

you were creating a, you wanted to create an automation flow to create.

Like images for this show, and you have the picture, you are sent a picture

by the guest and you want to just put the text or thing on the side.

Now, that actually sounds quite straightforward, but where AI helps is

that a lot of the time the guest won't send you a very good picture, or they

might be over here on the picture, right to the right or right to the left.

So you need to crop it, and all of this kind of stuff you

would do manually yourself.

But now an AI process can do that.

Without you having to do anything as part of an automation flow.

So AI is kind of superpowered automation by a lot of the times.

It would be nice because computers generally work in a nice, logical way,

but there's always been this problem where you would need a human intervention and AI

allows you to kind of bypass those steps to a certain degree, if that makes sense.

Okay.

So there is a difference there between AI and automation.

So you, you said that the, the ai, it's software that can do it,

can do more sophisticated tasks than what were possible before.

So it can do thinking type tasks.

Um.

With the quality being, I suppose, dependent on the quality of the

information you put into it, but then the automation side of it, you said that

that's been around for quite a bit longer.

But there is a way now that we can, we can use these AI tools as part of

automation sequences or processes or workflows, I suppose might be some of the

terminology that, that people would use to basically superpower those automations.

Yeah.

Well, I mean, previously automations are fantastic, have always been,

have been fantastic in a way, but.

They are very limited in terms of what you could do because you'd

always hit a roadmap where you need somebody to do something.

Um, and like if you run a big business, I mean, um, probably most of your

listeners are not gonna be in a big business, but maybe you have processes

between people in the company and it might be like, well, I need to send this

invoice off to Janet in accounts, and then she would send it to Barry in HR

Potentially those sort of situations you could put and I, if it's a very simple

task and that's important thing with ai.

It's a very simple task that Janet was just moving something from one

spreadsheet to another, uh, but needed a little bit human intelligence.

That's where AI could potentially be a replacement and that, but similarly,

because of the way that AI kind of replicates the way that human brains work,

this is like the whole neural net thing.

Ais make mistakes like humans do.

So it's very easy to fall into the trap.

And I think a lot of people have done this of thinking that ais

will replace humans totally.

But they are, in a way like humans, in that they make mistakes.

And as we know, you would never necessarily trust people to do

everything all the time in a particular way, if that makes sense.

In reliably.

So Tim, what's the simplest way, do you think that AI

or automation can help day to day?

I mean, I ended up down this enormous rabbit hole last, early on this

year where I thought, I'll get into ai.

I used to be a software developer many years ago.

I'll try and understand it.

And you've falled out.

And I spent many months looking into all these different AI tools and things, and

I, like most people, scratch the surface.

But the problem is that a lot of people have characterized AI by things like

chat, GBT and these people writing documents and putting M dashes.

And to me that's basically one of the most irrelevant ways to use ai.

If your writing is terrible, then maybe get AI to write stuff.

Much more useful from a day-to-day basis is if you've got access to

something like a local AI version of a smaller model or chat GBT or

whatever co-pilot, whatever you've got available, you can use it to run.

I ideas against and summarizing is AI is fantastic for, in terms of, um, people

who, who are creating visual things.

And I think this is something that architects.

From my understanding what architects do, I'm not an architect, so apologies

in advance if I offend anybody by it.

some of the visualization and image and video generation tools are fantastic,

and you can run these on your local machine or you can use something like

Google's nano banana, which, um, it's actually called Gemini three, but

everybody calls it like Nano Banana Pro.

Even they've started calling it Nano Banana Pro Now.

So for example, if you've got, you can give it like just a sketch of

a building and tell it, like, given this photograph of this particular

material, give me a rendering of that building in this particular style.

Uh, and that's the kind of thing that if you want to create images and

visualizations, um, it's fantastic for, and especially now because you

can create videos from those images.

And there's also tremendous ability to edit videos and images with ai.

So it's not a hundred percent there, but pretty much anything you can

conceive of, you can create an image or a video of probably for

about five seconds at the moment.

But that is the, that is the way the going.

We're very near to being able to produce full length feature films purely

with AI from commercial grade GPUs.

That is pretty crazy, isn't it?

The way that this, um, the pace of change with this technology,

it's, it's pretty incredible.

So just to recap there, you mentioned that.

One of the, the common use cases that people first think of is

writing tasks using something like chat, GPT, for example.

And if you're a terrible writer, that's not your thing.

Yes, maybe that's a good way to use it to begin with.

But you also said that there were much better ways to use it, and particularly

if it's for people that, um, you know, if you, designers, if you're using anything,

um, creating graphics or anything visual.

There are lots of ways that we can use AI in those day-to-day type of tasks to

be able to more easily produce graphics and videos and that sort of thing.

can, I mean, can it really make drawings or renderings look better if

we've got a sketch or, or a drawing that we've got already, if we were to

kind of feed it into one of these AI tools, can it, can it really do that?

it's not necessarily, it's not something where you are,

the first time you do it, it's gonna work and it's gonna look exactly.

But once you start learning about what these tools can do, um.

There's one, there's a few tools I wanted to mention today.

One's called Comfy ui, which is not really a comfy UI at

all, but it's a fantastic tool.

If you've got a fairly high powered, um, machine, you can run images, uh,

image generators and video generators on your own machine, but it can do

so much more because you can mask out particular parts of the image.

You can create 3D model from a particular static photograph.

There's lots of stuff you can do and it takes learning and time, but.

Really there is, there are ways to do everything you can think of.

Um, it's not just a case of, oh, I want a picture of a dog or something like

that, which is what people think of.

That's the text rendering.

But you can say there's things like something called a control net.

And so you can have a, like a picture of your, your sketch of a building

that can be a wire frame and you can say, create a building using

this as a control net for, and that.

Basically the way to think about it's, it adds a different little routing

end as well as the text you give it.

So you could say a tall building in the skyscraper setting, and then

you could give this control net of your rendering of the building and

it will create a cityscape with that building in the middle of it.

And if you can think about how powerful that is potentially for just communicating

your ideas with somebody who might not.

Be able to look at a plan of a building or something and understand how it is, and

then if you take that image and then feed it into a video model and you can say,

have people moving around in front of it, then again, that's another way that you

could potentially, I mean, that's probably more relevant if you are designing like

a station or something would have people moving around in front of it and you can

kind of show what it would look like.

So that, that's where it's really powerful I think, for people.

What about organizing other things?

So like if we've got.

Like project notes, reference materials, um, how, how good is AI at the moment

at assisting with those sorts of tasks?

Really good.

Generally.

There's another, another tool that I made a note of and, uh, even

though I'm not putting my notes on the right screen here, but.

Note, uh, notebook, lm, it's a free product from Google.

Um, if you subscribe to Google, um, business, I think you get a better plan.

But what that is, is basically you give it a whole list of blog posts.

You can upload articles and Word documents and.

Then you can just query that information in the text space way.

It will also create a podcast summarizing it.

You can ask questions about like, what's in that information.

You give it YouTube video links because it's a Google product.

They've got like access to all of that sort of stuff.

So that's probably the easiest to use, uh, tool for organization, especially if we're

doing research and that kind of stuff.

Um, there are more advanced ways of doing it.

You can actually feed this information into your own chat, GBT and other things,

but there are, uh, well there were issues with both Notebook, LM and Chat

BT, which come on later about privacy.

Obviously this information's up in the cloud somewhere in

Google or chat GBT Service.

But yeah, it is really amazing for that kind of stuff.

I've heard a lot of good things about Notebook

lm. Um, I've yet to try it.

That's on my, my to-do list of, um,

I, I don't really use it as much as I should

do, but I should use it more.

Yeah,

Yeah, well at least if we know about these things, if we know, you

know, have an idea of what they can do and um, how we can access them, then um, you

know, that's a good starting place, isn't

it?

For sure.

So, um, why would an architect or a small business owner generally bother?

Running AI locally instead of just using these cloud tools.

You, you've mentioned this a couple of times while we've been talking about

the, you know, there is this option of being able to run certain ais locally.

So kind of when we say locally, we're talking about your own computer that

sat in your office or your home, um, why would we wanna do that?

Two reasons.

Um, one is the good old reason of.

Uh, usually free or cheap.

Um, I mean, obviously there's power costing money in the machine and doing

all the rest of it, and at the moment you need kind of a really reasonably

good spec machine to run it on.

But they're starting to, they're starting to introduce like AI

machines and specialized one for a couple of thousand now that you can

just have sitting in the corner.

Have you got an office?

Three or four people?

That's not an awful lot, so yeah.

And you'd be surprised, certainly for the image generation and uh, video

generation side, the models are not.

That far behind.

It's kind of weird in that the open source world, most of these, some of

these models come from companies like Snapchat and Facebook and even Google

have got some open source models.

But a lot of the Chinese companies, um, like Alibaba, and there's another

one that I've forgotten, think of a big Chinese company and they've

got an AI model that's open source.

Generally speaking, the, the a the, they tend to release new

AI models every week or so.

Um, and the bigger models, so in terms of video image generation, they will

release a new version every three months, six months, the commercial ones.

And by the end of that three month period, the open, the open source models

are basically caught up and a little bit ahead, and then they'll be behind again.

So it was kind of like that when pseudo three.

Uh, pseudo when, when open, uh, chat GBTs video model came out, they

were ahead and then about a month later they were out of date again.

Um, so yeah, I mean, certainly for video and image generation, I think

there's a lot to be said for using local AI because you're not, unless

you want to spend a fortune on all these different subscriptions to

different video and AI services.

Um, the.

The cheapness will will help you.

The other reason is privacy.

You are guaranteeing that all of the data that you have is on your local network.

and if you are dealing with really sensitive information, do you want

it to be on chat, GBT servers?

Um, do you guarantee that they're not gonna have a data breach of some kind?

Admittedly, you need to make sure your network's safe thing, but at least

it's your fault, if that makes sense.

So, um, privacy and security is another issue in that you are totally in control

of your data and you can run like, there's G-B-T-O-S-S model, which is a, a

version of chat GBT that's open sourced.

Um, there's obviously deep seek, which is a Chinese model you can run locally.

Gem, there's a Gemini version, um, of Google's one you can run locally,

and they're not actually, and again, those models are catching up.

Like, um, they're probably a month or two behind chat, GPT,

Google, and, philanthropic.

So yeah, the, the problem at the moment is the hardware

you need to run them is still.

Quite elicit.

I mean, the smaller the model, the more stupid it is basically.

So depending on what you are using the model, if you're just using it for

automation flows, then actually having a local model can make a lot of sense.

Um, that comes onto the whole automation tools kind of thing as well.

Potentially the software costs are cheaper with locally

hosted AI because there's open source models available that can

be installed on a machine locally.

They might not always be the most UpToDate thing out there.

You said that that can change, you know, um.

by week, twice

Week by week.

But I think for most people that are getting started with this,

that they're unlikely to need the very newest AI models anyway.

I think, you know, for most of us that, I mean, it changes so quickly we, we

can hardly keep up with it anyway.

Um, but you did say that a disadvantage might be that.

In hosting it locally, you may need a fairly decent spec machine, which

depending on your circumstances or the size of your business, may or

may not be viable for you, which I suppose that could be a factor.

I mean, if it's an architecture practice that's already potentially, they might

already be, um, have been using, you know, doing like 3D rendering and

that sort of thing on machines anyway.

So it might be that they've already got machines within the practice that are.

Potentially capable to run this stuff depending on what the models are.

Yeah.

Well, I mean, I, I, I've done a lot.

The reason I had a reasonably good machine in the first place is I

bought one for video editing and it had like a 12 gigabyte graphics

card in it, and that is capable of running most of the smaller image and

small, small list video, um, cards.

So yeah, you're right, actually.

A lot of them, if they've done 3D rendering and stuff, they may well have.

Why?

I mean I, I think probably the smaller, it's probably like, and it's all about

at the moment for video and image, it's all about graphics, card, memory,

and generally speaking it just Nvidia.

Um, gradually the other graphics card manufacturers, their software

is not quite as fast as the Nvidia stuff, so it's a bit weird.

But for something like Deep seek or one of the text-based models,

'cause they're slightly different way they work, you just need a lot of.

Uh, memory.

It doesn't have to be graphics card memory.

So it's, it's funny actually, if you've got a Big Mac machine that's

really good for running something like deep seek or one of the text models,

it's pretty terrible for running image generation and video stuff.

But if you've got a really beefy PC with a 16 or 30, 24 gigabyte graphics card

in it, that'll be fantastic for running.

Image, image generation, and you can run company UI easily.

Um, you could also do a lot of other stuff.

There's like, Voice cloning you can do on a fairly low spec machine actually.

You can clone somebody's voice and create a new thing.

And also music creation, music generation like AI music with lyrics

and stuff, you can do this for me.

There's all sorts of weird stuff you can do with AI and people are

just thinking about, oh, is it gonna create something with an M dash in it?

It's like, well, that was like two years ago.

Really?

Like that's the kinda stuff that's irrelevant.

I, I find quite funny in a way.

there's a lot more we can do with it now.

Um, the other point just to come back to you that you mentioned was around privacy.

So again, if.

If you are a business where privacy is a key concern, it may be you're dealing with

sensitive information in the nature of the work that you do, then locally hosted AI

is, is gotta be the way to go, hasn't it?

I think, am I correct in saying, Tim, that if you are using some of the cloud-based

tools, if you are paying for those tools, do you typically get a better level of

privacy than the, than the free plan?

Like,

I mean.

GPT as an example.

Yeah, I mean they, they've got in their contracts, if you use

the free version, then they can just do what the hell they like with it.

Um, if you're paying for it, they at least give you some contractual kind

of, but even then it's like there's another person in the middle and like

they could have a system outage where weird data they throw around anywhere.

Um, it's also important to say that there is kind of a middle

option you can actually run.

There's things like run Pod where you can hire A GPU in the cloud, uh,

in your local country or whatever.

Um, so that would be like a, a third way potentially.

That's probably the hardest to set up in a way.

'cause you're using a remote PC to run a model that you would run locally, but then

you've got somebody else maintaining it.

And then if they're very high secure environment, I'm sure there

must be people who have got like.

Proper security thing around this in the cloud.

So there are kind of hybrid options in between as well.

What about presenting to clients?

So client presentations, that sort of thing.

Um, can AI make a difference with that for us too?

well, in terms of both the visualization elements, obviously.

Because it's so quick.

Once you've learned these tools, to be able to say, say like, oh, and if a

client say said, well, I don't want that building in that particular material.

I want it in a different material.

You can just say, oh yeah, I'll just quickly change the modern, send you new.

The other thing more generally in terms of presentations is that obviously some

of the models like Nano, well Gemini.

2.0 nano Banana Pro, whatever it's called, that can, that can

create charts and graphics and things and whole kind of report.

Now, I would never just put in like create me a graphic report about

this particular thing, mentioning these things, and then just take the

output you need to go through and check all of the output is correct.

Is, this is the thing.

It's like it will be nine, a lot of the time it'll be 90% and then

it will put some weird thing in somewhere, but at least you're

cutting down that initial time.

Um, so yeah, these AI tools can really help you speed up your processes.

That's the main thing.

Hmm.

Well, anything to save time then?

Uh, yeah, that's a good thing as far as I'm concerned.

Um, okay, Tim, a little bit of a divisive question, I suppose here,

um, just about the fears around ai.

Um, should, should we be worried about AI replacing us?

yes, yes and no.

Um, yes, in as much as if your, if your job is purely.

You are the person who is handed a piece of paper and told to create a render

of it, and that's the only job you are even paid like the whole week to do.

Yeah.

Your job might be, might be screwed, but more realistically, like as you said

in the answer you, your last comment, like there's so much we want to do as

business owners and people we can't do because we haven't got enough time.

AI is gonna reduce a lot of the time.

We do on particular things.

Automation tools can help with that as well.

As I say, like you could probably, I mean, I, I, I still run my book

chat live podcast occasionally, and I. Each show each season.

I do.

I've been like working on the automation to make it take less and less time.

Um, so that like, oh yeah, the graphics will think, go out or do

the social media posts automatically and all this kind of stuff That

gives me time to do other things.

So while AI is gonna reduce a lot of the work, it will

give us the opportunity to do.

Both us and businesses, the opportunity to do a lot of the things we know we should

be doing, but we haven't got time to it.

So if that involves you posting more on Reddit than you have done before, or if

it's like checking particular safety.

Procedures or worrying about putting captions on your videos to make

them more accessible for people.

There's lots and lots of things that we know we should be doing,

but we never get round to it.

Because we were taking less time on the things we know we must do because we're

giving them to AI tasks and automation tasks to do it that will give us and

so society as a whole more time to do the stuff we know we should be doing.

Now, I suppose theoretically there's a point where we are doing everything

we possibly could do and we all have become this utopian society, um,

where we all work two hours a week.

I don't think we are gonna be anywhere near that after AI's finished, so, yeah.

Uh, I, I, I don't think we, in general, we need to worry about not having like

a career or anything in the future because of ai, simply because we will

switch the things we know, we, the extra stuff we know we should be doing.

So that's, that's my take on AI at the AI revolution.

I've heard a lot of people talk about this idea that, um, AI's

not going to replace you, but another business owner that's using AI is likely

to be the ones that are gonna replace you because, um, you know, if they're

leveraging these AI tools to be able to work more efficiently, to be able to do

things quicker, to potentially be able to.

Sell their services cheaper at higher profit margins because they're utilizing

all of these tools at their disposal, then potentially your days are numbered.

You know?

But it's, it's only because it's another person that's just

utilizing this stuff that's.

Already available.

Um, so yeah, that's kind of, I think that's kind of my thoughts

on it, that it's more likely to be somebody else that's using the tools

that would, um, replace you rather than the robots just coming in.

Um, I might change my mind though.

I mean, maybe if we we record another one of these, uh, next year, then um,

your robot interviewing my robot, and uh, we'll be just

sitting on the beach somewhere, so

That doesn't sound too bad actually.

No, exactly.

As long as we are sat like sipping pina coladas on

the beach and we're not down to mine or something, you know, like.

Yeah, we are in the human dungeon after the robots are taken

Yeah.

Yeah.

That's it.

Yeah.

Um, Tim, this has been really good today.

So thank you for, um, sharing your experiences, uh, and what

you've been learning about ai.

It's definitely been really helpful for everybody.

What would be the main thing that you would like people to take

away from our conversation today?

Uh, don't be afraid of ai, uh, or automation tools and.

Become curious about what's possible because there's so much that's possible.

I mean, there's tools and things out there that I haven't even, I haven't even

investigated, and I'm sure you haven't investigated, but know that yeah, your

productivity can be like really pushed.

Ahead, uh, with AI and uh, yeah.

If you don't do it, then the person down the road might be the

person who's done it, done it.

And uh, you don't want to be the person who's replaced by the people

down the road who are using ai.

So that's something I would say sounds a bit macab.

But yeah.

No, that's, uh, it's very true.

Um, Tim, was there anything else you wanted to add about AI and or automation

that we haven't already covered?

there were a couple of tools that I didn't mention.

Um, let's scroll down and, uh, yeah.

LM Studio is a good one for running local ai.

As in like deep seek and that kind of stuff.

Um, and I think they've changed the terms.

They used to have like non-commercial use, but now you can use it commercially.

So that's a free point you can download.

It can run these small models.

You wanna just, if you've got a reasonably big pc, you can just download

their underscore studio and you can run a model on your own machine.

Just play around with it.

NA 10 I think is the best automation solution.

You can run that in the cloud or you can run it on your own machine.

They, they, it is an open source product and, um, I've got a lot

of time for an A 10 as a, as a basically coding base automation tool.

So, yeah, those are the two things that I wanted to mention and I forgot about.

where is the best place for people to connect with you online?

LinkedIn probably in terms of the AI stuff on Instagram,

uh, under my stone and press.

Um, brand.

I post a lot, travel stuff and stuff like that, but I'm trying to keep

linked in more the AI side of things.

And as I said, Tim Lewis ai.com is my new domain.

I bought about four days ago from before making this show.

Uh, where I'm gonna put all just focus on the AI and automation side of things.

And hopefully if, if I haven't posted much there, by the time the show comes

out, the listen me to post some examples of some of the, like AI renderings

and things that you can do, um, then hopefully it'll already be there because

I've been diligent, but you never know.

So, uh, yeah, that's, that's, those are the two places to get hold of

me or to find out more about me.