S2. Ep39 - OpenAI Dev Day Recap + Your Automation Questions Answered

Katie (00:27)

Hello and welcome back to another episode. Hi, hello, I'm Katie. And as always, I am joined by Noel. Hi Noel, how are you doing this week? You doing okay?

Noel (00:38)

time.

and yeah, having a great week as always. How are you doing?

Katie (00:42)

Yeah having a good week it feels like autumn has landed I feel like it

Noel (00:51)

Yep.

Katie (00:51)

wasn't very long ago that we were complaining about how hot it was here in the UK and now autumn has arrived there is a chill in the air

Noel (00:59)

Yeah, we're

both sat here with jumpers on that's it

Katie (01:05)

all is good

and all is good again.

So we always start the podcast with any updates, any news that are relevant to people using AI and automations for their business. But last week was OpenAI's big conference, correct?

Noel (01:31)

It was, yeah, it was their, they called Dev Day.

Katie (01:34)

Dev Day. And there's been lots of updates.

Noel (01:42)

Plenty. Yeah, there's all kinds of things, yeah.

Katie (01:47)

So I think we're going to spend a bit more than we do usually going through these updates. We've got some questions as well that we've been asked and yeah, all sorts. So let's get straight into it.

Noel (02:04)

Yeah. So the OpenAI Dev Day was last week and they've covered lots of different things. So usually with the Dev Day, what I would say is it's kind of quite nerdy, quite techy talking about code and APIs, but I wouldn't let that put anyone off because the Dev Day now is kind of like their product release kind of conference. So it's a bit like where Apple would release their new products.

Katie (02:31)

Yeah, do you know what I was just

about to say it's a bit like when Apple have their big like conference and it's like the new iPhone or something like that.

Noel (02:43)

Yeah, it is. However, I would say this event went as well as Elon Musk throwing a steel ball at a cyber truck and smashing the window after claiming it was an undestructible. So it didn't go all too well.

Katie (02:56)

Interesting.

Noel (02:58)

Yeah, it kind of reaffirmed my fear of doing live demos because a lot of their live demos didn't actually work and then the live stream crashed. But anyway, we got around it. We got the she got some updates from it.

So

the big, big update is they've released something which is called DOTS. Now DOTS is essentially a AI agent, which would be your personal assistant and follow you around, not physically,

Katie (03:29)

Okay.

Noel (03:32)

follow you around on

Katie (03:32)

Let's get to know.

Noel (03:34)

like ChatGPT It would help you with your tasks, your scheduling and all those sort of things.

I would say it's more of like your personal assistant. Yeah, it looks quite cool, although it does look a lot like the characters from Grokbot. So they use the same sort of like shape agents and things that and yeah, and basically on the back of that, the guys at Grokbot, they basically went out and bought

dots.com or whatever, and then repoint it to their Grokbot. So yeah, there was a bit of marketing to and fro and after that. like I said, essentially it is just that AI agent you can connect to your emails, you can connect in calendars, any application that you would normally connect into ChatGPT this dot will have access to. And it will also learn about you over time as well. So let's say you kind of like

update on what's going on at a certain time every day you'll find that this dot will go off and just automatically do that anyway it will grab all that information just give it to you so

Katie (04:45)

Okay.

Noel (04:46)

yeah it's really really quite cool yeah but this part of their demo was probably the worst bit because it just wouldn't work but it does work currently I think I believe it's on the two higher tier plans so

Like the pro plan, so like the $90 plus plan is where they sit currently. I'd imagine they'll come down eventually to the lower plans in the future. But another big thing about this as well is they'll also have a feature coming soon where you'll be able to send a text message or give it a phone call to your dot. And it will

Katie (05:24)

Hmm.

Noel (05:25)

also call you as well. So if you were like, could you call me at this time? It would go off and it would call you back.

So yeah,

Katie (05:32)

Okay.

Noel (05:34)

so that's kind of interesting, interesting how that would work.

Katie (05:37)

Interesting,

I'd like to kind of like think about how that could be used, you know, in a business sense or like how it might be helpful.

Noel (05:49)

Yeah, because I guess you can access it through the ChatGPT app. So you're already on the phone.

Katie (05:56)

Mmm.

Noel (05:57)

yeah, it's something they've just added in, but who knows.

Katie (05:59)

Yeah,

it'd be really interesting to have a little think about actually how that might be helpful.

Noel (06:05)

Hmm.

Katie (06:05)

If you're listening, you already know how that would be helpful. Please let us know. Either email us hello at makeautomations.ai and tell us all about it. Or you can come along and tell us in our free LinkedIn group, is called AO Automations for Business. And we'll leave the link below for that because we know

LinkedIn groups can be a bit hard to find at times. So yeah, we'll make sure we leave the link to that. And if you can't find it, feel free to like email us or message Noel on LinkedIn.

Noel (06:41)

Okay. The next big release after Dots was a thing called Spaces. it's kind of, yeah, it's kind of hard

Katie (06:49)

spaces. Okay.

Noel (06:52)

to describe, but within

Katie (06:54)

Mm-mm.

Noel (06:55)

Spaces, you can like work on documents, you can work on slides, and you would work collaboratively with your team through ChatGPT. So pages almost look like

notion pages sort of thing that that sort of vibe and then obviously slides are just like your presentations but basically you and your team members using ChatGPT could all collaborate and work within this space you can manage tasks as well as that and

Katie (07:24)

get.

Noel (07:25)

if you have meetings using ChatGPT and i'm not i've ever come across anybody that does yet but that is possible

but you will also get all the transcripts and things that which are automatically bought in. So, yeah, kind of useful, I guess.

Katie (07:43)

Hmm.

Noel (07:44)

But the also good thing as well is you can bring your own dot along to it as well. so

Katie (07:49)

wish you can love it

Noel (07:53)

if your team of five could bring your five dots along to, you know, collaborate

Katie (07:57)

Great. Yeah.

Noel (07:59)

and do stuff so you could delegate things to somebody else's dot or

your own and yeah

Katie (08:04)

Yeah?

Noel (08:05)

it all gets a bit a bit messy i guess from that point you know but

Katie (08:09)

Yeah. Yeah.

Noel (08:11)

yeah it's too too many cooks and all that i guess

Yeah, but I thought it

Katie (08:18)

Good one.

Noel (08:18)

was it was kind of interesting to see how that all went to links. But I guess with most

Katie (08:22)

Yeah.

Noel (08:23)

people and companies, we've already got all of our, you know, documents and slides and collaboration spaces. We've already got that already set up. So

Katie (08:33)

Mm-hmm.

Noel (08:34)

that's going to be quite a hard push to, you know, get a company to change and just use chat. I would have thought. know, it's there as an option.

Katie (08:41)

Yeah, yeah.

Yeah, it's an option. We like options.

Noel (08:47)

Yeah,

yeah. Well, I suppose you could connect it into your current apps that you're using anyway, I suppose, and then just share it through a connector. So,

Katie (08:53)

Yep, yep, that's a good option.

Noel (08:57)

yeah, so yeah, kind of kind of helpful. The next big one after this is they really they had to release a new model. Of course they did. So they released GPT 6.1 Sol. So that is just behind six Astra.

And

yeah, just as cheap but pretty clever.

Katie (09:20)

Okay.

Noel (09:21)

But as we mentioned last week, Claude, they bought out sonnet 5.5. I guess they already kind of knew OpenAI would be releasing a model.

Katie (09:31)

Yeah.

Noel (09:31)

they'd have to, there's all a bit of tit for tat, isn't there? When it comes to model releases.

Katie (09:34)

Mm-hmm. Yeah. It's kind of like keeping

up with the Janes' sort of thing, isn't it? Well, whatever you can do, I can do better. Yeah. Yeah.

Noel (09:40)

It is, yeah, yeah. Definitely. Lots of that.

So yeah, not really that much to really talk about with that particular model. They did also release something called Ultra Fast. Now, I use Opus 5.5 a lot, and that is pretty darn quick. Now,

ChachiBt Ultra Fast is like insanely quick. It can

Katie (10:09)

Okay.

Noel (10:10)

do stuff in like light speed. They even showed like a demo where it was like controlling a game in real time where it could figure out how to move around and do stuff in a game or whatever. You know, very, very niche test, but Ultra Fast just means it will be able to go off and do tasks quicker, whether they be longer tasks, maybe it might be.

know, coding tasks and things like that. So really, really clever how it all works. However, there is a catch and it's an expensive catch. if you want to use ultra fast, you need to subscribe to their brand new five hundred dollar per month plan to get access to it, which is ridiculous. So they've called it pro five hundred. Yeah, I'm not sure the marketing.

Spiel is gonna get many people to to upgrade to $500 a month, but there you go. With that plan also you get more usage, of course But what most people are finding if they turn on ultra fast is they're getting through that usage? As you would expect pretty quickly And then yeah, they're then like left with an account. They can't do anything that they've paid $500 for so yeah interesting that they've gone with that, but I think when you look at

for the amount of usage you get. I think that still beats ChatGPT even with on all of their plans, including the ultrafast. So yeah, really interesting that they've done that.

Katie (11:41)

Yeah.

Noel (11:43)

One of the other updates was a, you we talked about Jev a couple of weeks ago it's an AI model which does decisions, but in like a blink of an eye. So OpenAI have released their own one called Decisions API. Not as catchy as Jev in titles, but it basically does the exact same thing. So I'm guessing in the future, we're gonna see a lot of other,

model

providers provide their own decision API or variant of things like just so we can basically get those tasks and those snap decisions done super fast without wasting a load of computing resource for, know, trying to be used to do a load of reasoning. It doesn't really need to just need to say yes or no. So yeah, that was pretty interesting to see. And they've also started to release

a marketplace for plugins. Now they've tried this a number of times and they failed every single time. And I've had a look at the new marketplace this week and I think this one's going to fail as well because it makes absolutely no sense. But the pitch was is like their enterprise users would you be able to buy their services via the marketplace so you could sign up with them.

And those enterprise customers could then sign up to your stuff. You know, if you had a plugin, you could sell that to an enterprise customer. It all sounds great, but if you go and look at it today, you have to submit a form to, you know, say, I would quite like to go and, you know, work together or buy the product or service from this enterprise customer. I don't see the point in doing that because if you want to work with them, you just go on their website and then

buy it or book in a demo or whatever. I don't, you know, it doesn't

Katie (13:39)

Yeah. Hmm.

Noel (13:44)

seem to make any sense to me, but I'm

Katie (13:46)

No.

Noel (13:46)

sure one day they'll get this plugin marketplace nailed down, but alas, not yet. Yeah.

Katie (13:52)

But until then, no, no, exactly. Okay. That was a lot. That's a lot of updates.

Noel (13:59)

Yeah, there was, there was a hell of a lot. Yes. Yeah. Lots to get through.

Katie (14:01)

Yeah, okay.

Well, we've got our own update as well before moving on to the main part of the podcast, we've been asked a few questions, so I thought we would answer all of those. But before we move on, Noel, do you want to tell people who are listening some of our updates as well?

Noel (14:24)

Yes, so right now...

Katie (14:26)

Which are more exciting, much

more exciting than what Noels talked about for the last 10 minutes, just like to add. Yeah, yeah.

Noel (14:35)

Yeah, very true. It's nothing to do with ChatGPT either. this week we're launching a one-on-one and team sessions to help you get set up with Claude. We're getting lots of people come to us and asking how can I use Claude? I'm just tinkering with it, but you know, I want to be more productive. I want to know how to use it properly.

You know and basically sit down with with me and then figure out you know how we can set up their accounts so they can get more out of it so That is what we're releasing this week We'll leave a link in the show notes so you can go and go off and check that out and book in if you wish but we do have 90 minute one-on-one sessions and We also have a half day session as well these are

Predominantly online sessions unless you're in the Bristol area and then yeah, we can meet up face to face and you know Go through it go through it that way. But yeah, that's I'm quite you know, I'm excited to get into this because it's interesting to see all these You know different people and how they want to use Claude, you know, I know how I like to use it And yeah, I just can't wait to see how these guys want to use it and you know Get get them up and running and you're using it a lot better

than what they currently are.

Katie (15:56)

amazing

Noel (15:57)

So we also have on top of this as well, we are also releasing our courses as well. So we're going to have lots of different courses on all kinds of different topics, but we will have some where we're just going back to the basics, like how you can find your repetitive tasks. What should you automate? So those very early questions that you will have in implementing AI in your business and

We'll then also have some extra courses like on Claude, Claude code. We've got one on Clyde, which is an AI agent system that will be coming soon. We'll also have all kinds of different courses on like make.com and things like how to set up your automations in there, maybe connect them into Claude and things like that. And have those two applications work together with your data. There's lots of different things. Yeah, we're really excited to get those out, but yeah.

I'm going to be spending a lot of time talking for the next few weeks, getting all this content ready. But yeah, that is all being launched this week.

Katie (17:05)

Very exciting and

Noel (17:06)

Yeah.

Katie (17:07)

if there's anything as well that you think or I'd like a course on this or You know, would like one-to-one help with this and you don't see it. Please email us And we'll see what we can do

Noel (17:21)

Yeah, absolutely.

Katie (17:23)

Okay, so we've been sent in some questions and some we have already answered in different podcasts. things like, what is the first automation that I should set up? We've done podcasts on this and then like the difference between something

like Zapier and other automations, again, we have done a podcast episode on this. So please just look back through our back catalog and have a look because we would be able to explain it much better in a full episode rather than kind of like trying to answer these questions quite quickly now. So the first question that we have got is actually

is time relevant? So it's a long-winded question, but I'm just going to shorten to it to what is a realistic time saving from automations? Because this person said that they keep seeing save like X amount of hours a week claims. So this one was

it was I keep seeing 10 hours a week claims.

Noel (18:40)

Yes, this is a lot of things that we go through with different companies and stuff. it's kind of, it's all that return of investment. It's all kind of linked into that. So if you save X amount of time, then that would then be, you know, a pot of money or time back or whatever. So I guess the way, the way you need to start looking at this is working out, you know, how

how long are these repetitive tasks taking you to do?

Katie (19:12)

Mm.

Noel (19:13)

And we have this in the free course, which will be coming very soon. And in that worksheet that we provide in the course, it basically details all of this out so can scribble it down and say, look, on Monday, I do this particular task for two hours a day. Could I hand that off to AI? Yes or no. Could I automate it? So it's all kind of

What I would suggest is to kind of sit down, you know, and give yourself a week. And then as you're doing stuff and you just think, you know, I just don't either don't want to do it or I feel as if AI could just do this for me, then just write that down with how many hours that would take. And that would then give you your weekly hourly total that you would save. So yeah, when people just put out random numbers saying you're going to save 10, 20 hours a week, like great, but

You know, we all work and do things differently. So, you know, that number is always going to shift. like I say, if you write it all down, then you'll soon get to your number very quickly.

Katie (20:20)

So where do you put human in the loop checkpoints and how do you decide what steps an agent can do alone?

Noel (20:33)

Yeah, that's a good question. So what I would always say is if a agent is going to go off and do something where it would be published, anything where it'd be sending anything outside of your organization, then that is the point where you say, no, put that human in the loop at that point. So let's say you're, it's drafting an email then

you would always, you wouldn't let it just go off and send that email. You would always put it in your drafts. And then it's then that human in the loop is then you to check it and then send it. So it's always

Katie (21:12)

Hmm.

Noel (21:12)

at those points and especially where it might be sending, you know, we could be a message or data to a customer that that's always a big touch point for human in the loop.

If it's sending anything to a customer, I'm like, I need to see this before it goes anywhere. Because the last thing you need is for that customer to come back to you and go, what's this about? It doesn't make any sense. You're like, no.

Katie (21:39)

Yeah.

Noel (21:40)

We don't want those moments. Yeah. So that's kind of where I would look at it. If it's going to send anything outside of your organization, that would be where it would definitely have a human in the loop

Katie (21:52)

Okay, good to know So the next question is actually directed straight to you,

Noel (22:03)

oh no.

Katie (22:04)

Yeah.

Noel (22:06)

You

Katie (22:08)

What's an automation you built that failed or backfired and what did you learn?

Noel (22:15)

failed or

Katie (22:16)

Yeah and there's actually two parts

to this question so they've also asked another question which I am going to ask you as well once you've answered this because these were good questions.

Noel (22:29)

Yeah, that is a great question. I mean, I've got to rack my brains a little bit for something that really went wrong.

I guess one thing we did do a while ago is I basically built the backend of a software product using make.com. So there was, I think it was in the region of 78 scenarios that were all interlinked together. And silly me, I not fed in the right parameters.

when people were using it, they were like, this just makes no sense. And I was like, what you

Katie (23:07)

Mm.

Noel (23:08)

want about it makes no sense. And then, yeah, I'd completely mucked up some of the parameters that I was feeding through the make scenario. So yeah, they were just getting stuff that was like completely generic and not on brand whatsoever. Whereas the whole point of it was it was supposed to give, you know, it was trained on context and brand when that brand for that company. And they were just like, this just makes no sense.

I get this out of chatGPT and I'm like, what? And then yeah, yeah. So yeah, that backfired pretty spectacularly and very quickly, thankfully an easy fix.

Katie (23:44)

Yeah, and what do you feel that you learned from that?

Noel (23:48)

So what I learned from that is to always check, check and test and then test and check again. Because I've done a quick pass of it and then yeah, was like, yeah, that's fine. And then,

Katie (24:02)

Yeah.

Noel (24:03)

yeah, yeah, I should have, yeah, I should have done more, but yeah, I definitely test everything lots these days. yeah,

Katie (24:11)

Okay.

Noel (24:12)

big lesson learned there.

Katie (24:14)

Yeah. So the follow on question is what's the one automation you would keep if you had to delete the rest? So you're only allowed to keep one.

Noel (24:29)

Whoa, wow.

So I have one that for anybody that interacts with the make automations website, they go through one automation. So yeah, we can't delete that one. But yeah, I could do without everything else, but yeah, I definitely need that one.

Katie (24:51)

Yeah,

okay. the next one is about who owns an automation. So we had this question asked a few times. So who owns an automation after handover?

Noel (25:13)

Yeah, that's a great one because some agencies will keep it on their accounts. Depending

Katie (25:20)

Mmm.

Noel (25:21)

on what you paid for, some will structure it in a way where you have to pay a monthly retainer and then they'll maintain and keep those automations running on your behalf.

Katie (25:33)

Yeah, or be like a

bit of a support desk maybe.

Noel (25:37)

Yeah. So with those sort of agreements, that's kind of a bit of a gray area when it comes to who owns what. So what I would always see that as is if their make account was deleted, how would that affect you? And if it is, it would all disappear. Then you as the business don't own it. That's something which I personally don't do. I always hand it over. I build everything in, in customers make.com accounts. So yeah.

Katie (26:07)

Yeah.

Noel (26:07)

If I disappear off the face of the earth, then their automations don't go with me.

Katie (26:12)

Yeah. And it's

not just as simple as like, you, you you, you disappear, but things happen, life happens, you know,

Noel (26:22)

Yeah.

Katie (26:23)

and anything can happen. And if it, if it's just you on your own, then yeah, all of a sudden for clients that I don't know, support desk just disappears.

Noel (26:39)

Hmm. Yes.

Katie (26:41)

So yeah, it's good

Noel (26:43)

Yeah.

Katie (26:44)

to be able to hand things over so people can.

Noel (26:50)

Also, actually on this point,

Katie (26:50)

Yeah.

Noel (26:55)

if the scenario is on the agency's account, then you won't probably see when it goes wrong. So if there is an error or anything, then that agency would get that notification to say, this scenario has failed, we've turned it off. But you

Katie (27:12)

Hmm.

Noel (27:14)

that's using it wouldn't probably see that.

If it's on your account, that email will come to you and then you can go, crap, need to, you know, I need to get on the phone or on the email to, know, to get this sorted out or, you know, nip in and turn it back on or whatever. So yeah, that's something also to consider.

Katie (27:35)

Okay. All right. Well, those were all the questions that we've had in. We've been collecting them up for a little while. Like I said, a lot of them were very similar or repetitive. But if you've got a question, please email us. Hello at makeautomations.ai and yeah, we'll answer it on the podcast for you.

Noel (27:57)

Yeah, sounds good. Yeah, always look, I really do like answering those questions as well. Yeah. You get some, it gets me thinking more about different things and how things work and stuff.

Katie (28:10)

well we definitely need to ask you more. Everyone send in a question. Whether you think

you know it might be a bit basic or whether you've got something quite particular please send it in. There's no wrong question. What's it something like the the only wrong question is the one that you didn't ask.

Noel (28:32)

Very true.

Katie (28:33)

or something like that. I've probably

Noel (28:35)

Yeah.

Katie (28:36)

not quite got that but you know what I mean.

Noel (28:39)

Yep.

Katie (28:41)

And that's it for this week's podcast episode. We hope you've really enjoyed it and come in, let us know your thoughts and your questions. Like I said, we'll leave our email and the link to our free group over on LinkedIn AI Automations for Business below so you can come and join us. It's free for everyone to join.

But thank you so much and we will catch you for another episode very soon.