How AI Is Reshaping Operations at Small and Mid-Sized AEC Firms

Hello and welcome to another episode of Design Under Influence.

I'm your host Alex.

I am co founder of ARC it.

We help architecture design engineering companies with their IT so they can focus on just running their business, creating beautiful buildings, running amok with AI, whatever.

We take care of it, security, safety, that kind of stuff.

Here with me, my co hosts Liz and Megan.

They're with Aurora bim.

Say hi.

Hello.

Wow.

Maybe you can do a quick like 2 second pitch of what you guys do.

Sure.

We help small to medium architecture and structural firms navigate the BIM world.

Perfect.

On the spot.

There it is.

Okay, so we, we have been doing a lot of shows or multiple shows on various subjects along the bim, innovations around BIM and transitioning into new systems that will make design more efficient and effective for everybody, help you make more money and stuff like that.

But today we're also, I would say, on the bleeding edge of experimentation with AI and how artificial intelligence can help us continue to improve and grow our businesses.

And I'm learning a lot from you guys when we have these conversations off camera and we're conducting a lot of experience here internally experiments.

So what I wanted to do is put this show together with you where we can share what we are doing right now, what's working and what, what I guess what we wish would be working, but it's not yet working.

So maybe Megan, do you want to like start us up?

I know, it's just, just give us like what have you done lately with AI that you've been impressed with?

Oh yeah.

Well, I guess I would kind of like to lay the foundation a little bit.

Go back in time to a year and a half ago when, when like oh my God.

So long ago like that that I think AI has really basically released the floodgates in a way for entrepreneurs to be able to be entrepreneurs.

If I had not had ChatGPT or Gemini in the beginning of the creation of Aurora Bim, I would have been stuck under and, or had to hire someone to do a bunch of writing of our proposals, our contracts, all of that stuff.

In the beginning that was an evolving like huge, huge long conversations with Chat GPT, trying to get it perfect, evolving it as we got new clients with new needs, writing emails.

All of the things that just take time really have been expedited to a point where like I think it saved us years.

Like I think in an pre AI we would be where we're at in about five years from now.

So I think that's quite a report card.

Wow.

I mean it's huge.

It really has impacted our ability to be business owner.

Liz, would you agree?

Totally.

Yeah, completely.

And also just to like.

No, I would just build on that to consume as well because there's so much info coming back at us from clients, from contractors from the world.

And to be able to also use the.

The AIs to help us consume that, like summarize it and just give us the points we need is.

Yeah, priceless basically.

And like the other.

I mean the big.

So jumping into the AI tools we use, like the big one that we implemented pretty early on was a note taker, actually, I think we saw your note taker.

We saw another note taker from another company that we were working with.

She was also a small firm and she.

It was Fathom AI.

And so now it's a note taker that follows us into every meeting.

It provides summaries and notes and it's a database that we can go in later and ask questions to.

And if we forget, you know, something that was done in the meeting, we can always go back and ask or click on a clip, basically like a summary item.

And it takes us right back to that part of the meeting.

Those we found our clients really like those as well.

You can take the summaries and put them into Chat GPT and create like an agenda for the next meeting or action items or all of those things.

None of this we could.

We didn't do any of this like four years ago.

And with as many clients as we have now and all the calls, like there's just no way that like, I don't know, there's no way that I could keep it all straight between all of the different clients and all of the different information and setups.

And every client is unique.

So that's a big one.

I mean ChatGPT, Gemini, email assistance, what else?

So those are the stuff that we're currently using.

Let me go through the what we are using.

And so people are watching this and listening to this may be able to relate.

I have two businesses and I want to talk about both because it's important where that AI comes in handy.

From your main street retail.

I have an outdoor retail shop, fishing shop, where that comes in and ARC it, which is, you know, IT management firm.

Right.

MMSB provider, managed service provider.

So on the managed service provider stuff in Fireflies is what we use.

I think it's similar to Fathom and besides just recording calls and summaries and stuff like that, which is super useful.

You're right.

You have a thread of documentation leading all the way to your client relationship.

What you and I talked about is how do we connect all these threads together?

Because right now we have that info scattered everywhere.

Right?

Your Fathom has client nodes, your CRM has notes, your Excel spreadsheet has some new entries.

Like, right.

It's just everywhere.

And then your invoice is going out.

That's completely different system and they're not all talking.

So that's, let's put a peg on this one because that's, that's the ultimate.

But to connect something, you have to have it.

And I think we're already on step one collecting all that data.

But one thing that may not be very impactful to you yet, I don't know how many employees you have, but one of the things we're doing is when our manager has one on one coaching sessions with the team every week, those are recorded.

And what we're building is a database of those conversations on the over quarter.

So when we put a quarterly review for a team member to give them, hey, here's all the great things you've done, here's all the great opportunities that you should do.

Here's the stretch goals, here's the real goals.

Here's your bonus.

Here's your, your next, you know, you know, here's what your career path is going to look like.

So that in those hour long conversations, a lot of that intelligence, a lot of that information that will never like a human person would not go and read 100 pages of conversations.

So that is a big, I think it's gonna be a big addition to improve quality of re.

Quality of feedback to our employees, which will improve our employees.

Okay, well, and also it's an unbiased feedback.

Well, it depends how you ask.

That's an issue.

Depends how you ask it.

That is true.

We ran into that today too, didn't we, Megan?

Yes.

It's all about that prompt engineering.

All right, so if you go to ChatGPT and say, here's my conversations with John over the last 30 days, 90 days, I want to be able to fire John tomorrow.

Give me reasons.

Yeah, it's going to find it.

No, but then it's like you in order.

Because I used to dabble in the HR field too in previous companies and I would say then it would be like, okay, we just set up a company standard prompts to get this.

So it's always the same words you use to get this.

Yeah, the same secrets.

You can get more of a unbiased summary of something instead of Just you know, me trying to fire John and looking for reasons.

So we're not just looking for, for what chat GPT is going to spit out.

What the manager is going to be doing is using his her judgment and have those bullet points or opportunities maybe not really something that hasn't surfaced for the manager's manager has eight people under under.

They can only have so much time but this will help them compile that review.

So typically employee reviews the reason why small companies don't conduct them and bigger companies conduct them.

But it's the most painful process.

This is most painful time in everybody's life.

The manager, the employee, the owners.

It's all waste of time.

Grand giant waste of time.

Everybody hates it.

But it's necessary.

Right.

Because it just sucks and but what doesn't suck is when as an employee, my boss, what knows intimately what my goals aspirations are, how well have I done and gives me actual things to work on that to me is of massive value.

Right.

As an employee.

And so what we'll do is I, I guess it's a very long winded way to say we'll shortcut that the icky part of the process to get to a really effective outcome.

That was a long winded way to say that we use note taker too.

Yeah.

Which note taker?

Fireflies.

Oh, use.

Okay, never mind.

Sorry.

I thought it was a different one.

Yeah, I think it's the same.

I mean it's, I think similar functionality.

What it is just follows you in every meeting and stuff like that.

Yeah.

What I've been talking about is like I and Liz found a solution that apparently there's like a little thing you can wear around your neck because at the end of the day.

So I mean AEC firms, you see this all the time.

And even for you guys, you have to log your day by the hour, if not sometimes by the half an hour in order to get paid.

So all that time needs to get divided up and put into the right buckets.

And like doing that at the end of the day just for me never happens.

Liz has become disciplined and she can do it.

But like I still I've also been enabled by another AI tool called Timely that you install in your machine.

You tell it your, you link it to your Google Calendar, your teams and your Microsoft Calendars and then it basically records during your day all the things that you touch on your computer.

And then you can start to tell it.

You say this is project A B C D and you throw the first couple of chunks of time into Those projects and it learns, and then it actually starts to anticipate your time and put it in the right buckets.

So if you do it right and you program it right, it creates your time card for the end of the day, which is really nice.

Does it work?

Yeah, it works pretty well.

Yeah, but it's.

It's, you know, crap in, crap out, so you have to spend the time.

I need to organize it.

Yeah, I've got to organize it now based on, like, more like I just had clients, so it would be like, okay, I worked for this client, but now it needs to be client and project.

So I've got to go in there and like, organize so that it's got the right buckets.

But then we were looking for, like, there is a way to export data from there to go directly into our invoicing software.

Software, which is.

Would be amazing.

But the part that's missing on that is that it is not listening to my conversation right now or integrated with Fathom.

If there was like, some way to tie those two together.

So, like, I'm in a meeting, it knows what I'm talking about.

It remembers all of the things I said I was gonna do throughout the day as, like, my action items.

Right.

Like, it's kind of like a personal assistant that I'm envisioning that's following me, that helps me log my time.

But also it reminds me about all the things that I talked about on all these calls.

Because I start I get up at like, you know, 6am By 7am, I've had like two or three calls, and then I'm on calls until like 3pm, 4pm every day.

By the end of that, my brain is mush and I can't remember what I said.

I can't remember who I told I was going to do what for.

If I have time in between, I'll log tasks and I'll log action items and things.

Or Liz and I will get on a call and we'll just have like a working call, which we decided works really well to just get stuff done.

But, like, it's becoming.

The volume is becoming.

The volume of work is starting to surpass, I think, like a single person's ability to handle it.

And I think it's because we have the AI that we have this problem.

But I think AI also needs to be able to come in and help us solve it.

Yeah, yeah.

And we talked also, you know, like prompts or even if it's the recorder you wear around you or if it's just every so often, if timely, was like, hey, you're not in a meeting.

What are you doing?

And then you can answer it like, I'm working on this.

You know, that it maybe once an hour, every half an hour.

So we have lots of ideas for improving it.

And then also with our colleagues that we work with and our contractors, we don't want to chase them down and get their status update.

But if they were prompted every day at a specific time where it was like, hey, what'd you do today?

And what are you doing tomorrow?

And they just answered it instead of us, and that gets recorded somewhere, it's.

Then we can just go look at that.

That.

See, that would be amazing.

Like, can you imagine if we just said, what are you working on today?

What do you plan to work on tomorrow?

And then it fills in this, like, time sheet what every employee is working on.

Are we working on the things we need to be working on and making progress on the priority items?

Why is so and so working on this when we need all hands on deck on this?

Right?

Yeah, exactly.

Yeah.

You know, let me run an analogy by you.

You know how I love those.

And then I'll.

So it's just to go back real quick to what you were just talking about.

I just finished utilizing new utilization policy.

Time utilization policy.

Because we are, you know, IT firm, we're required to work on IT issues.

And so 80% of our tax time available time per day.

Engineers time is required to be.

Not required, but it's.

It's the upper goal target for them to get their bonus.

They have to use 80 of their time working on customer tickets or internal tickets that are related to customer projects.

No, the.

The point being is that they have to go into the ticket, right?

Open the ticket, the clock starts, they doing their work, they close the ticket, the clock stops.

And that's kind of easy to do.

That sounds nice.

But they still have to fill out their time card at the end of it, which is a painful.

A painful process, and we're seeing some challenges there.

But let me run an analogy by you.

So I feel like starting a business is like running, you know, in a bathing suit, in a beautiful Hawaii, on the beach.

Okay, so it's like running on the beach, but you're running towards water.

Okay, so like, imagine like maybe a year and a half ago.

That's where you, like.

All right.

Running towards water.

Yeah, it's so exciting.

Then you're starting to get into the water, and this is where all of the stuff begin.

Like, oh, you need accounting oh, you need to pay taxes.

Oh, oh.

You know, you need to build this client.

Great conversation with the client, but they promised me to introduce me to the next client, so I got to remember not to follow up with them about that introduction.

But.

So you're just entering the water, right?

So you're entering the water, you're still kind of hopping, it's still fun.

And then the surf starts coming in, and then eventually you go neck deep and eventually you start swimming and it's super inefficient, and that's it.

And then you're not having fun anymore.

You just kind of fighting the waves.

It's.

You get thrown around.

You're like, man, I just want to be an employee.

It was easy.

It used to be easy.

What the heck happened?

I'm here in all this BS and I feel it.

I feel it every day.

And I'm almost angry at AI its inability to solve this right now.

Yeah, it's pretending to be so intelligent.

And I think it is.

I think it's got a lot of that capability in it.

But the way it fails on basic things is still frustrating.

I'm still, you know, wobbling in the water.

I'm not like, I'm not seeing it.

Like, I think it's doing great job.

So for marketing, like, I can talk about what it's doing.

And it's like you said three years ago, I had to have a full time, full time person doing this or what it's doing right now in a matter of me prompting for 10, 10, 15 minutes a day.

But yeah, outside of that, it's just all this disjointed, all these jointed systems.

Like, I just want to spend my time on figuring out how to grow the business, Test, experiment all these ideas, talk to customers, be out on trade shows, be out in customer offices figuring, you know, I mean, that's what I want to do.

What I end up doing is clocking my time, you know, just documenting conversations, writing policy, making sure the policy is being followed, writing specific documentation on what we'll do when people don't follow the policy and escalations, you know, and what, what happens.

First write up, then we're gonna do, you know, second write up, whatever.

Right.

So it's like hr, finance, accounting, customer relationship, all this stuff, right?

Yeah.

Have you ever heard of like, those.

There's like, they are selling those AI where it's like your AI assistants and you have different.

I was just about to bring that up.

Yeah.

Have you?

I haven't heard of anyone who's worked with them.

If it like can you train them to actually do that?

And then they hang out over here and listen to your day and then chime in when you need to answer a question like I don't understand how that like actually even works.

Oh no.

So this is the whole concept of right now.

The functionality of AI right now is what they call agentic, which is agents.

And this is exactly so timely is an agent.

Right, right.

It sits there.

It sits there and just does one particular thing.

But it's still very far from this being useful.

And I'm sure maybe some larger companies have little bit better way for them to have AI sit on their server.

You know the LLM sit on their server and then being trained on their data.

But larger companies are too slow and too risk averse.

Right.

So they're probably not the job either.

I think that though, because it's such a, it's some fire right now and everyone's like get on board or you're screwed kind of thing.

I think that the large companies are putting a lot of money into these initiatives and I think that is why we are seeing AI grow so fast.

Because I think the large companies are on board.

But on board meaning they dedicate resources.

They don't expect much, you know, but the customer service chats become.

Became a lot more smarter.

They did.

How much did you hate talking to automated chat?

Terrible.

A year ago, right?

Zero.

Zero.

I just put zero.

Give me a person agent.

That was a doornail.

If you need a company's address, it can give it to you.

Yeah, you need a company's even.

But if you need it, contact somebody, you know, it won't let you.

So anyway, it was dumbest thing ever.

I don't understand how people even thought this was useful.

But now it's much better.

You.

Yeah, it actually can search its knowledge base, it can make some inferences and you can write in plain English with mistakes of what's going on with your particular problem and it actually understands and either points you the right direction.

Right article like that, that's very impressive.

So I think that's one thing that that's working right now.

Yeah.

In customer service I mean I think so we're also finding ways to like for us like being consultants, like we do a lot of teaching and training as well as part of like our deliverable for clients.

And I think the other part that I'd like to see gain abilities is the ability to make like images and infographics and like slide decks and things like that and I, I've tried out a couple different softwares, I forget what the one was called, but like they could create a presentation for you really quickly or like, you know, or even a PowerPoint, like a slide deck or a PowerPoint.

And it would just be like you put in your ideas, you tell it whether you want generic images or clip art or whatever.

And it would just, it would push this like beautiful presentation out for you to then tweak and edit.

Which I think is.

That's something that I've always really been.

It's been difficult for me even in like I remember in college to come up with like my own outline and like my own document versus come in and help someone else and tweak something that's already there is like a whole other ball game.

So I think that is something that, like having these tools to get you 90, 85% of the way there and then you tweak it and make it what it needs to be and then let it do a little finesse at the end.

That's doing.

I think that they're doing really well with that and I'm excited.

I started trying out Nano Banana.

Oh yeah.

Oh yeah.

That's pretty killer, man.

That's what made me sign up for the Pro.

Really?

Gemini?

Yeah, it was so.

Yeah.

So like Liz, it will do.

That's the one that like just got released that does like way better infographics than what like Dall E or just like regular ChatGPT has been able to do.

If you ask ChatGPT to build you a slide deck, it's crap.

Yeah, it's like text like not aligned in the center of the page.

It's like, it's a mess.

But if you add in Nano Banana and ask it to do it, it's pretty good.

The text that goes on, it still has those weird blocked out pieces.

The little black.

The text in the image still is messed up, but it's much better than it ever was.

Okay, so they're getting there.

Yeah, it's impressive.

But I, I can tell you my.

Let me just convey my particular observation with how limited current platforms LLMs are with data processing.

Okay.

Because at the end of the day, if you want remember our dream, right?

You actually articulated the dream really well.

Last time we spoke, Megan, you said I want something.

I want a tool that just has everything in one place and right.

All of my business, from invoices to HR files to customer, you know, conversations captured to my calendar, everything right in one place and everything's running and Everything's searchable and.

Okay, good.

I want that too, very badly.

But the data part is not there yet.

And I think that's where a lot of things right now fall apart.

And let me give you an example.

So I've put my shop in Napa.

I've uploaded 12 months of sales history line by line.

It was 5,000 times.

It was about 40,000 lines of what was sold UPC date it was sold.

What's the cost?

What's the price margin?

And it's like a lot of data.

And it was so happy taking it.

It's like, give me more.

I'm like, okay, great.

And because I couldn't export it because I only have 5,000 fields, you know, 5,000 lines limit for my analytics to export it.

So anyway, I have to do every two months.

So it's like, oh, I'll stitch.

So it's.

It stitched all the data, multiple exports.

Like, I can see all your data.

Here's your revenue, here's your, here's your margin.

And I said, okay, exclude this.

And then it's like, everything's fine.

Great.

I'm like, dude, I'm paying $200 a month for analytics to sit on top of my point of sale system to help me making orders, right?

To help me with intelligence about creating orders.

Because this is very important for retail.

For a lot of businesses, that data is key, right?

So here's the set data.

Help me infer what I need to bring next couple of weeks for us to not run out of stuff, but also not over buying.

It's not that complicated.

But there's a lot that goes into it, a lot of other parameters that can go into it outside of just basic analytics stuff.

Because the way analytics works is like you sold three, you have two, therefore you need one right?

Now you did this last month and I think you might need five more to keep up with the demand from last, right?

What about seasonality?

Right?

Big time, right?

It's Christmas time now.

It's completely different.

So look at last December, not yesterday, not last week.

What about what's called phantom demand?

You know what phantom demand is?

No, not me.

Phantom demand is.

I brought it to.

I sold two in next two days.

And then for a month I didn't have another, didn't have a replenish.

So it's not like I need to order two in a month.

It's like I need to order eight.

Because we could have sold, we could have sold eight more of these gadgets given the sales velocity.

But because we're out of stock.

You can't sell something that's out of stock.

Got it.

Anyway, it's all retail stuff, so I was so freaking excited.

I'm like, that's it.

I'm gonna cancel my analytics.

That's it.

I don't need any of that crap.

I'm gonna just.

Gemini and I, we're gonna.

We're gonna do it all.

So I said, okay, let's put an order together for yeti.

You guys know yeti, right?

Yeah, yeah, yeah.

The cups and the coolers.

Yeah, the coolers.

This is.

And it's like.

I'm like, all right.

Based on the information you have, you have everything in phantom demand.

Like, give me an order for Yeti.

Like, it spits out an order for, like, $40,000.

I'm like, okay.

So that's.

And then it tells me, are you severely limiting your ability to like.

I'm like, okay, is.

Whatever.

So here's my budget.

All right.

Push it to grand.

Okay, Next.

The problem is, like, what it started doing is it started to just give me.

Give me stuff based on my last years.

But some of the stuff, you know, how yeti, like, changes colors and discontinues, And I'm like, that's.

I can't order that stuff because it's last year's.

It wouldn't be.

So then I said, okay, let me upload yeti's catalog so you can see more data, and then let me upload our current inventory on hand.

Anyway, by the time I'm done uploading everything, the thing is like, yeah, ready to go.

So it spits out the order, and it's missing, like, upc.

So then I tell it, okay, you got to put UPCs.

How am I going to order it if there's no anyway?

So it's what you really need to do.

It's like a child that is extremely smart and, like, a genius, but it has no idea how life works.

Yeah.

And so the problem is you have to.

And that's where the programming comes in.

Right.

You have to program in the parameters that you need for it to go and evaluate for you for whatever project it is.

And each project has different parameters, different requirements, and you also need to let it know what to consider as your limitations.

Yeah.

And so after this, like, exhaustive two and a half hours, we got the YETI order done.

But I tell you, you know how long it takes me to put a YETI order together on my own?

Fifteen minutes.

Yeah.

Yeah, but it was two and a half hours.

And maybe I got two products I wouldn't have ordered otherwise.

Maybe.

But now can you do it again?

Yes, I try.

And so that's another answer.

And I asked it like, hey, you have all the data.

You have all my sales data.

I want to do Daiwa order.

I wanted the Shimano order.

And I'm, I'm like, okay, this is two and a half hours.

It, I invested it.

Now it's all figured out.

Now I have the prompts.

And so it says, no, I can't retain the data after you close the chat window.

Wait, was this Gemini or Chat Gemini.

Oh, I need you to memorize this date.

I try to hack it, right?

Like, hey, I don't really need this relationship unless you have this data available for me for any of my chats and conversations, because otherwise we can't be together.

Does that not.

It doesn't work like ChatGPT does.

I guess it doesn't like, because you.

Can have projects in ChatGPT and it retains that information for correct.

Ever.

Correct.

But I don't know if it retains 40,000 lines of data.

ChatGPT I haven't tried yet.

You can upload documents to it and data to it.

Yeah, but the question is, does it retain that data and integrity of data.

Because the reason why I got excited jump on Gemini is because it was supposed to be better with data.

Because ChatGPT falls apart for me.

Like, Gemini gave me hope.

Chad GPT gives me no hope.

Like the stuff it comes up with, it says this order is 99% you're going to sell through this order in 60 days.

And it's just a bunch of horse, horse crap.

Like, like Chat GPT is not, like, is not lived up.

So I use it a lot for marketing, you know, getting intelligence from large text, that kind of stuff.

Right.

Everything you're doing, but not for data.

Have you.

But it's funny, I feel like I have the reverse experience because I used Gemini in the beginning.

Gemini and ChatGPT side by side when I was doing all the startup like contracts and proposals and stuff like that for my company.

And Gemini was like, they would give back the worst answers.

Like not even like following my prompts, like just complete crap.

And ChatGPT was good.

So I've always gravitated towards Chat GPT just because that was where I started.

But I have heard that Gemini is getting better.

Well, the new version just Nana Banana came out with a new version.

This thing is a week like two weeks ago.

Yeah, when people actually watch this or listen to the show, probably going to.

Be something else, it'll be a new one.

But what people end up doing is.

And I'm listening to a lot of, you know, people who are on the edge of this and investing in AI and stuff like that.

And you know, companies are using anthropic for any kind of communication related stuff.

They use Gemini for data.

Like they use different GPTs or different, I guess AIs for different LLMs for different tasks.

And I was hoping, because GPT failed on data for me quickly because for me it's easy to test.

You know, this is like, and this is maybe the takeaway for the viewers and listeners.

Like I have this retail shop and everything is so clear.

Like either that order is brilliant or it's mediocre or it's crap.

And I can tell, like I know if it, if it inferred like, hey, you should order these because things in Napa, you know, those are the colors.

And it did that originally.

And then we asked to do.

When you see the problem is like I said, hey, give me a reasoning for ordering this sku, right?

The skew this color and put that in the spreadsheet and put along.

And so I said, you know, I told the car to design the spreadsheet, do it again.

So it runs the same or the same, it's the same order.

I just added more parameters through a new prompt and it creates a new order.

So I'm like, okay, so what happened to my last order that I thought was good?

It was almost there.

Yeah.

And then now it's like, Jesus.

And all I did is I asked it to add two more columns and give me the reasoning.

Now I guess it was afraid to give me BS reasoning.

So it had to redo the order, come up with some reasoning that sounded good.

Like, yeah, it's a forever.

It's a forever spiral.

Like it's a, it's a, it's an.

I don't know if it's an adventure like Alison Want Wonderland.

You know, it's like you go, you go there, you, you stay there.

I mean, it's just, you gotta wake up, man, and just go to work.

Because it's.

This thing is absolutely just round logic.

Like it's just this child that spits out answers like, don't know.

Just do what you did.

Just add this one thing.

Just do what you did.

Yeah.

And I've actually said that I've been in this loop.

They know exactly how you're feeling.

It still Doesn't.

And I love the confidence.

99% of the time this solution works.

I'm like, there's no 99% of the time.

And you don't know what that means.

And it doesn't work.

So have you done the.

Like, I do the voice version in my car and I have conversations with it and it like.

I'm like, no, go back to what you told me two.

Two prompts ago.

And it like, can't.

It like can't compute.

It won't go back and get the like, almost perfect ANSW gave me.

Yeah.

And that's still.

So in the car, I carry on.

I have an hour commute to and from when I go to the shop.

And this is for me, same thing.

I talk.

But I.

Last night, last conversation was yes, last night I was going home, I was talking about picking up a new series, the book to read.

And finished with a serious.

I was reading and I said, hey, I like Dick Francis, I like C.J.

box.

Can you recommend something similar serious that?

Blah, blah, whatever.

I describe what I wanted and then we started talking.

It was very useful.

It was very useful from that perspective.

It could do a good job.

Yeah.

Yeah, I.

It was really useful when I was wandering around Sydney and was asking it for suggestions and ideas and architectural insights and.

And so on.

Then it was great.

Yeah, super helpful.

Yeah, that's.

That's really cool.

So, you know, we went to Italy this summer too, and you know, it was just like, without it, I don't know how I'd live.

You know, there's translations, everything's so easy.

But like, when it comes to data, it and like its ability to be certain in its outputs.

Okay, because why do you reprompt something and add one more dimension?

Well, I guess you had one more dimension.

It goes out and redoes everything and everything like before didn't exist.

Now it's all right, it does do that.

And when you want it to do that, it doesn't.

Sometimes I'm like, forget what we just talked about, try again.

And it can't.

Exactly.

And so to me, this example of, okay, my daughter's learning to play volleyball.

Okay?

So we pass this way she knows how to pass the ball and set the ball.

And so I say, hey, just use three fingers, not five.

And it would just forget everything.

Forget even needs to use the hands and actually catch the ball.

But she'll use the three fingers.

But, like, she'll miss the ball because.

Like.

You know, whatever before we talked about is still relevant.

It's not like, you can't just lose it and completely.

Like, now I have to tell you, like, okay, use three fingers, but do touch the ball and.

Yeah, and it's gonna do that.

Touch the ball, but it's gonna be sitting down.

I'm like, you gotta stand.

Right.

And it's like, never.

And now it stands, but it forgot about three fingers.

Yeah.

That uses four fingers.

This is.

This is what.

This is how frustrating this is.

So this was my report to you guys.

Do not do what I. I spent half a day and my head was.

My brain was buzzing.

I was so excited.

I was so excited.

Here's.

Wouldn't this be.

Here's all my sales history.

I'm gonna just prompt it like, hey, give me a YETI order for next two months.

Done.

Set it to the wrap.

It would save me 10 hours a week.

I wonder if there's.

Is there a GPT out there or a different.

Is there a platform we don't know about that does this better?

I'm sure it's being built and I'm sure it's being improved.

And then when you take it back to the architecture side of the CA side of things, what we're trying to build, what we're starting to build for our clients or offer to build for our clients is what you describe, but sort of small portion of it.

Like, we want to have an agent that actually got, does and organizes all project communication into one space.

That's it.

It's one.

Like, we don't want to do everything, but if we can help our architecture customers organize all the communications for a particular project in one space, that is automagical, so they don't have to go and drop things.

Yeah.

All the time, you know, that's a big win.

So I will tell you, we didn't.

Solve the AI question.

What.

What.

I mean, I also want to throw out there.

So did you.

Do you know what, like, what.

Do you know what they're going to call success for AI?

What they're saying is like, the ultimate goal.

What is it?

That's a.

You just.

You love.

I just opened a can of words.

Sorry.

At the end of the call.

Teaser for the next call.

Doing great work here.

That's a very good question.

Well, so it's like that.

So this podcast I've been listening to, they talk about that.

There's like, is it a bubble or is it not a bubble?

And the whole conversation around whether it's a bubble or not is like, what is the.

Like these companies that are producing all of these different LLMs and software for us to use.

What is their ultimate goal?

Like when are they gonna stop developing it?

And the ultimate goal is for it to take over what humans can do, period.

Take over humanity.

So like, I forget what it's called, but like that is their goal.

So it's like that kind of scares the crap out of me.

That is the goal.

But they're a long ways away from it.

But that's why they say the bubble, it could be a bubble because they don't think that these companies, that they're pouring all this money into it with all these high hopes that this is this like goal that they're not going to be able to achieve.

So that's why they're talking about it in this.

I'm not explaining it very well, but that's, it's like the, the dot com bubble that there was like these big high hopes for that pour a bunch of money into it and then if it doesn't do what they said it would do in the time period that they thought it could do it, that's when the floor drops out financial wise.

Yeah, I don't necessarily.

So I hear very similar.

So what do you guess?

Like a tail?

General intelligence?

AGI?

Yeah, AGI.

And that's like AGI's ultimate goal.

I don't think it is.

It's not.

We already.

You just mentioned, you just said this at the beginning of the podcast.

If AI wasn't there, we'd only get here where I'm here in five years.

Isn't that already like it's successful, mind blowing outcome?

For sure.

I think we're just a bunch of spoiled children running around here wishing for things when we get so much already.

Right.

What do I need it to do to just like run my retail shop for me for 20 bucks a month?

I think it's wishful thinking.

Totally.

I, I just thought it was interesting.

That is like the AGI is that conversation in itself.

It was mind blowing.

When I heard them talk about it, I was like, that is their goal.

Like I would want the goal to be like to still have us involved.

I think the ultimate, the ultimate vision of it that I see that the good side of it, the good side of it would be that the AI agents and the general models will do all of the heavy lifting, all of the, you know, of the, all the nuanced and, and unpleasant kind of work.

Like for example, if I'm really good at math, I don't want to do anything else.

I just want to do that.

Right?

Or if I'm really good at marketing.

That's all I want to do.

Right.

I want everything else done for me.

And I think if it does, that's gonna be amazing, you know, I think, you know, another amazing thing is humanoid robots that wash dishes, walk dogs.

I need.

Nobody's brushing my dogs.

Like, I am just trying to take them to a damn.

You know, it's like the hair is everywhere around the house.

Like, okay, my little vacuum cleaner robot, you know, it's not robot.

It's like the thing is bumping everywhere like a dumbass.

Right?

Eating core.

And it's like it gets full all the time.

So you have to like, I need a robot.

This goes and brushes my dogs, cleans up after them in the backyard.

This is what I do every day.

I pick up poop.

And so I hope it's two dogs.

You know how much they produce.

Yeah, it's a lot.

A bit.

A bit.

And so anyway.

Yeah, so I think that would be groundbreaking.

Right.

If we get someone another thing.

I'll tell you one last thing.

You guys get me going.

This is what's happens.

I just don't.

Shut up.

But like, I have two girls, right?

I have 14.

I'm sorry.

16.

They were growing up so fast.

16 and 12.

Okay.

I don't feel comfortable sending them to school or elsewhere on with Ubers.

Uber drivers, who knows?

Yeah.

I don't know.

They're vetted, they're checked.

But I would send them away or a driverless car.

It will relieve me to no end.

Right when.

And Waymo's already.

You guys around San Francisco.

There's a lot of Waymos everywhere.

I love them.

They're in LA too.

Perfect.

Yeah.

They haven't made it all the way out to East Bay yet, but they're expanding San Mateo now.

Like they're expanding and it's.

I can't wait.

Like that will change.

That will give us so much more time because now we can send the kids to mall or whatever.

Because then I'm just.

Weekends I'm ferrying kids around, right?

Yeah.

Oh, yeah.

So we're doing.

So I. I would say besides AGI, we have a lot more to look forward to and yeah, all that.

All the money that they're investing in the infrastructure, at the end of the day, a lot of it, or most of it is raised for build out of data processing centers.

And.

It'S hard to call this a bubble in my view, because it's already accomplishing so much.

But you're right, given the investment, like it better do a lot more.

I mean, so it's interesting.

This guy, you got to listen to that podcast.

I told you because he has a good one on this and he's basically like, they need to actually become two separate things because it's kind of like, yeah, the dot com, whatever bubble burst.

But we still got the Internet and it still is like affecting our everyday lives.

Like, it didn't fail.

It just didn't do what they said it would do in the time they said it would do it.

Yeah, yeah.

It's like Webvan.

I don't know if you heard about Webvan.

It was back in the 90s, this was the company.

This was the first company that did free delivery of grocery.

Like this was a most amazing thing.

A web van.

Web van, yeah.

They big.

They became huge.

They built the fascinatingly automated lines to sort groceries and to deliver like those photo sensors back then.

They invested billions.

I mean, it was crazy.

And they crashed so hard and they burned so loud.

Like it burned a lot of other companies.

I think they started the dot com crash Webvan.

Yeah.

And this was like, that's it.

You'll never have to go grocery store again.

Everybody was.

Grocery stores were shaking, like, oh my God, nobody's gonna come in.

Free delivery anywhere in an hour and.

A half or something.

It was crazy good.

Yeah.

I love that.

I mean, Amazon kind of did it eventually.

It took 15, 20 years to actually execute, but it was too early.

So.

Yeah, I mean, we'll see people crash and burn in this space, but everything's moving so fast.

I think we do it in two weeks.

We're already going to have different opinions on stuff.

Totally.

We're going to have a new tool to look at.

Yeah.

I'm going to bring an infographic from Nano Banana.

Yeah.

So let's do.

We can do this episode once every couple of months.

I think two months is long enough to actually get some exciting improvements or talk about some fails that we had with the AI.

I think we should do it every.

Every couple of months just to, you know, just to stay on the edge.

Stay on edge.

Yeah, I think it'd be great.

Yeah.

Cool.

All right, well, all of you who are still with us, we appreciate you.

We wish you happy holidays.

If you are listening or watching this in the time of December, you know, happy holidays.

It's a fantastic time of the year.

I want to thank my co host, Liz and Megan.

So thank you very much.

If you guys need guys and gals, if you need help with it, it's get arcit.com if you need help with BIM, it's aurora bim.com.

all right.

Thank you very much for watching.