Transcripts
The challenge with that is AI is happening to you.
Speaker A:How do you ban AI?
Speaker B:I don't think you can anymore.
Speaker B:It's built into everything.
Speaker A:Hello, and welcome to another episode of Design Under Influence.
Speaker A:I am Alex.
Speaker A:She is Liz.
Speaker A:And this is.
Speaker A:We're going to come at you from two perspectives.
Speaker A:One is IT provider for architects.
Speaker A:And Liz does BIM consulting for architects, right?
Speaker B:That's right, yeah.
Speaker B:And structural engineers.
Speaker A:Oh, and structural engineers.
Speaker A:Well, today we're going to address a source subject or it's becoming more and more important and it's reality of AI and its adoption in AAC firms.
Speaker A:And I think I'm going to open with something, Liz, and then you just kind of poke at it.
Speaker A:So you're going to be helping me, keeping me honest and potentially, you know, asking questions that help our audience understand if I get carried away.
Speaker A:Sounds good, but I think what's happening, and even if you're listening, you're running your C company architecture, you're thinking, okay, maybe we're not ready to adopt AI.
Speaker A:We don't want to kind of deal with that reality right now.
Speaker A:Let's just sort of see how it goes.
Speaker A:It's too early, blah, blah, blah.
Speaker A:The challenge with that is AI is happening to you.
Speaker A:And I don't know if you have this in your firm, Liz, but everyone here use different tools for different things and keep playing with stuff all the time.
Speaker B:Yeah, same for us.
Speaker B:Absolutely.
Speaker A:Hey, let me put that proposal into perplexity and see what would be RFI or hey, here's the.
Speaker A:Here's my plan.
Speaker A:Floor plan.
Speaker A:Can you see if we build this, you know, what happens, whatever.
Speaker A:So a lot of times this could be like, again, this is all new.
Speaker A:Right?
Speaker A:But, but let's just look at the AI infrastructure right now, this far.
Speaker A:Right.
Speaker A:And maybe in a month it's going to be different.
Speaker A:But for now, you have a lot of agents that are being built based on LLMs that originally were designed where in China.
Speaker A:Okay.
Speaker A:They're very good.
Speaker A:They're very, very good.
Speaker A:And they're open source.
Speaker A:So a lot of times people think, well, this agent here, I could just use it personally as a personal assistant.
Speaker A:Problem is, you're going to have to authorize access to all that.
Speaker A:Not even talking corporate yet.
Speaker A:Right.
Speaker A:Personal.
Speaker A:So access to my Gmail.
Speaker A:Here we go.
Speaker A:Access to my project management tool.
Speaker A:Here we go.
Speaker A:Yes.
Speaker A:And it becomes super productive.
Speaker A:Now, the challenge is it's built on that open source LLM.
Speaker A:But who's to say that there were not some code snippets in There snuck in to share your data.
Speaker A:Yeah, it's so early, nobody knows.
Speaker A:So I think what happens is your team members start using different LLM products.
Speaker A:You know, between chatgpt, Perplexity, using all these Nana Banana for graphics and I mean, it's irresistible to do so.
Speaker B:Sure.
Speaker A:And I think what is your favorite, like, what is your favorite use case personally, for AI when it comes to.
Speaker B:Work, the work that we do, it's the recorders.
Speaker B:Because I can be more present in a meeting.
Speaker B:I can ask questions.
Speaker B:We can just say, okay, we'll do that thing.
Speaker B:I can go back afterwards and query the recording about what's going on.
Speaker B:I can also collect all of that into one place and query like, okay, what was said about sheet setup, what was said about acc, what was said about like in a big overall to collect that information instead of having to write it down and remember everything and forget something or didn't understand.
Speaker B:And that's my favorite use case.
Speaker B:And then my second favorite is probably whenever I need to create any kind of document or email because I know what I want to say, but I don't.
Speaker B:I get so stuck with starting with a blank sheet of paper that, you know, you can, you can use that in combination with.
Speaker B:So let's say that you have a meeting about how you want to build a specific part of their template and now you need to go document that in a manual.
Speaker B:You can kind of brain dump into it and then ask the AI to compare what you wrote back to all the meeting data and ask where have I forgotten something?
Speaker B:And now compare it to the industry standards.
Speaker B:What else have I forgotten in this?
Speaker B:So, you know, reaching into all different aspects of the Internet.
Speaker B:But you're right, it is now getting access to a recorder.
Speaker B:It can, I can also give it access to my email.
Speaker B:I can give it access to a contract.
Speaker B:I can give everything.
Speaker A:Let's unpack this.
Speaker A:All right, so you gave two excellent use cases.
Speaker A:So let's say, number one, let's say the value here, I think this makes you.
Speaker A:And like I personally used exactly the same setup.
Speaker A:So I think it makes you almost twice as productive.
Speaker A:And it's not just the speed, it's the quality of the output.
Speaker A:Like your care to clients priorities is exponentially better because the Fireflies or whatever note taker you're using, it can dissect that.
Speaker A:They're that good these days.
Speaker A:Let me give you the real world scenario.
Speaker A:Okay.
Speaker A:I went to my Fireflies and we had conversations about raises for certain people, performance for certain people, and you know, by default, it doesn't gate it.
Speaker A:What do you know?
Speaker A:It doesn't gate anything.
Speaker A:You have to make sure your meeting note taker is structured.
Speaker A:And also you can't bring just random note takers into your meetings if you're doing outside stuff.
Speaker A:Right.
Speaker A:Or if you bring in outside people.
Speaker A:Well, let's say you have a meeting between your team, you, Megan, maybe a couple of the key employees and let's say you brought a consultant who brings in their own meeting tool without telling you.
Speaker A:Right.
Speaker A:And you guys discuss very sort of, let's say, you know, touchy subjects because they're your corporate importance.
Speaker A:But anyway, so.
Speaker A:So that, who knows where that stuff goes.
Speaker B:Exactly.
Speaker B:So well and.
Speaker B:But that's not takers that we can see and we might know about.
Speaker B:But there's also different types of note takers.
Speaker B:Now you saw the new product Apple's going to drop that basically just sits and spies on everything.
Speaker B:And I could have that going now.
Speaker B:And how would you know?
Speaker A:I would hope that we're past, yeah.
Speaker B:We'Re past those tricks because, because I'm like, look at this cool thing.
Speaker B:But I'm just meaning, you know, for those kind of meetings because we do have a big client that will not allow us to bring any note takers in.
Speaker B:And at this point that's a bottleneck for us because now we have to sit in meetings and take notes of the old fashioned way.
Speaker B:We've gotten so used to just relying on the AIs to help us and facilitate that now we're a bit like, I don't know how to work like that anymore.
Speaker A:And we don't want to work like that anymore because I'll tell you what you only have.
Speaker A:This is back in, in pre AI days, you only have so much concentration and focus that you give to your day.
Speaker A:Okay.
Speaker A:If you, if you now taking manual notes and paying attention.
Speaker A:I do that during podcasts because I like it, but I like it.
Speaker A:I don't like.
Speaker A:It's not mentally taxing, but during a call where you're structuring a next three months of design service that is super intense, you have to pay attention.
Speaker A:So you, you're spending that finite amount of focus you have for the day and then something else.
Speaker A:Other things won't be as important or this will fall apart anyway.
Speaker A:Yeah, understood.
Speaker A:But so here's the thing.
Speaker A:How do we deal with this as a business?
Speaker A:Can we ban AI?
Speaker B:I don't think we can.
Speaker A:How do you ban AI?
Speaker A:Say Liz, you want to ban AI?
Speaker A:How do you go about It, I.
Speaker B:Don'T think you can anymore.
Speaker B:It's built into everything.
Speaker A:People will quit.
Speaker A:They'll be like, hey, I have now, I am now have this amazingly high caliber emails.
Speaker A:I know how to deal hold myself with clients.
Speaker A:I'm double as good in my bim, you know, consulting stuff because I can pull up any manual quickly, you know, any line in any manual so quickly that it makes me a lot more productive.
Speaker A:And now you're going to take that away.
Speaker B:You just can't.
Speaker A:So Benny fails.
Speaker A:We, we both agree on that, right?
Speaker A:So what is that then?
Speaker A:Then how do you go about like even you think about your organization?
Speaker A:I think about mine all the time because I'm guilty and you're guilty, right?
Speaker B:Yeah.
Speaker A:Do we have the controls we need?
Speaker A:Do we have the, the tech stack they call AI stack actually need.
Speaker A:And so I think there's a process what we call secure adoption framework.
Speaker A:Okay.
Speaker A:And that's what sort of.
Speaker A:There's five things in there.
Speaker A:I'm going to mention them and like I'd like you to think about how would you structure that for your business?
Speaker A:Number one is discover usage where AI is being used.
Speaker A:You need to understand how your people are using it in order to aid themselves.
Speaker A:Then you need to the other piece, the second piece is classified data in your org.
Speaker A:Liz in Aurora bim what is considered to be proprietary, what is considered to be, let's say there are levels, right?
Speaker A:Something that can be shared internally on all users, what should be locked down for group.
Speaker A:And then lastly what could be public, like your marketing stuff, blah, blah.
Speaker A:And so I think that work right there, it's not a lot of work, but it's intentional where you classify data and, and then you have.
Speaker A:The next step is to approve enterprise AI apps.
Speaker A:And when I say enterprise, people are like, oh, dollar signs.
Speaker A:It's going to be, you know, $100,000, no 20 bucks a month or whatever, you know, business chatgpt.
Speaker A:But we'll talk about why that's going to be a lot more advantageous down the line.
Speaker A:But you need those tools or copilot or something.
Speaker A:And we can stop here for a second and I can say already people will argue, well, how do you know those are secure?
Speaker A:I don't know from a very detailed perspective.
Speaker A:What I do know is Microsoft's already got access to your stuff, Google's already got access to your stuff.
Speaker A:So it's an American company.
Speaker A:And not to say that innovation only happens in America, but we have laws and regulations that work, policies that work and companies will cease to exist if they fail us as users.
Speaker A:So there's, there's controls, there's guardrails.
Speaker A:So is it 100% safe?
Speaker A:No, but you can't stop it anyway, so you might as well get to where the enterprise grade security lives.
Speaker A:Okay.
Speaker A:That's where it's going to live.
Speaker A:It's going to live on Copilot, it's going to live in Gemini and it's going to live in business Chat GPT.
Speaker A:We're on the same page.
Speaker B:We're on the same page.
Speaker A:On the same page.
Speaker A:Okay, step number four, lock down the environment.
Speaker A:Meaning you kind of have to lock down personal GPTs and other things.
Speaker A:I don't think it needs to be vast or again, it's not something complicated.
Speaker A:But when somebody just types chatgpt because they forgot, it doesn't go right to their personal account, it goes to their business provision account so they can do their work.
Speaker A:And number five, publish simple policy in training staff.
Speaker A:So that's our kind of framework for how to securely adopt AI.
Speaker A:And the first thing we're doing it is to ourselves.
Speaker B:Yeah.
Speaker A:And we've almost gotten through the cycle to go through the steps ourselves as a company.
Speaker A:I mean we have 20 engineers working.
Speaker A:It's not trivial, but it's not that complicated.
Speaker B:Right, right.
Speaker A:And so I encourage you guys to think about those things.
Speaker A:But the next phase of it is, is so this is how it works, but the risks.
Speaker A:I just want to reemphasize some real world scenarios out there.
Speaker A:Right.
Speaker A:If this is not done, things will be done to you.
Speaker A:And you told me a story before we started talking and that sounds so unreal, but it was real, right?
Speaker A:It really happened.
Speaker B:Yeah.
Speaker B:Yeah.
Speaker A:You want to just give us a little like, just like a little snippet of it.
Speaker B:A little snippet with be that a person that I know, friend of mine was in a managerial position and one of their employees was not performing well and they were already considering letting them go.
Speaker B:And this person was on AI on I don't know which platform, let's just say ChatGPT and asking it questions about this situation and leaving it open and unlocked and available for her to see.
Speaker B:And then once termination happened, this AI was still logged in and still like basically mirroring everything that this former employee was writing and asking.
Speaker B:And it turned out that they were going to sue the company and they were for wrongful termination.
Speaker B:It was also having mental issues in it.
Speaker B:It raises also an HR and legal question because yes, we're talking about AI.
Speaker B:It feels a Little bit like reading someone's diary.
Speaker B:But emails, I mean, it's the same thing.
Speaker B:The company, it's a company computer.
Speaker B:If you were writing private emails on company email, they would have right to read them and see them.
Speaker B:It's company.
Speaker B:So it's, it starts to ask those kind of questions as well.
Speaker B:And so I think to your point, it also is good for individuals to think about it and think, okay, if I'm going to be writing personal things to an AI, have I locked, like separated myself from my business?
Speaker B:Because someone could access that and use it to hurt me or know things about me, I don't want them to know because they're personal.
Speaker B:So it's a. I think people have to start thinking of these AIs in the same way we already do with email.
Speaker B:Private email, work email, maybe even work PCs or work environments.
Speaker B:And private PCs and private environments not let them overlap just because they're new cool tools that we talk to.
Speaker A:That is a really good point.
Speaker A:Even from perspective of, okay, data leaks is one thing.
Speaker A:Okay, data IP loss is one thing, and that's very, very important.
Speaker A:But hurting, you know, potentially hurting your employees or employees hurting themselves by using all these tools because they have to, because you're not providing them with the appropriate tool set to do their work.
Speaker A:So they have to log in their personal GPT that has all kinds of other stuff in there, plus work.
Speaker A:That is a very good point.
Speaker A:That's an HR nightmare.
Speaker A:And as an employee, I don't want that.
Speaker A:I don't want anybody looking into my private GPT chat histories.
Speaker A:That's invasion of privacy.
Speaker A:Right.
Speaker A:So at least that's how I feel.
Speaker B:Yeah.
Speaker A:Then I shouldn't use it on a computer laptop or a work laptop.
Speaker B:No.
Speaker B:Right, exactly.
Speaker A:Very.
Speaker B:I mean, it goes both ways.
Speaker B:It's like me protecting myself, but also a company protecting themselves.
Speaker B:Because what if I say, well, the reason you're firing me is because you read my, my private conversation, which had GPT and I was just joking about how much I hated my job and my boss.
Speaker B:We're just joking and like just needed to vent a little bit.
Speaker B:But that's why you fired me.
Speaker B:So it's wrongful termination, you know, like it can go both ways.
Speaker B:Ways in a way.
Speaker A:Yeah, I can see that.
Speaker A:So with that said, let's sort of, let me kind of encapsulate everything with how ARKit helps with this.
Speaker A:This is all new for a lot of people, but we've sort of structured a program, an approach.
Speaker A:And the approach is usage, discovery, data Classification, approve AI stack governance and policy, staff training, that's it.
Speaker A:Those are the five steps.
Speaker A:But each of them need to have.
Speaker A:Each of them needs to be completed in order for this, for your organization to adopt AI.
Speaker A:The ones that are going to win are not going to be the fastest adopters, not going to be the first ones to use all the tools, be the smartest ones who go about it the right way.
Speaker A:I wanted to highlight a couple of things.
Speaker A:One is the history.
Speaker A:I don't know about you, Liz, but the Chat history and ChatGPT, that's why it's so sticky for me.
Speaker A:Like, I'm paying for Gemini Corporate, I'm paying for GPT Corporate, I'm paying for GPT Personal, I'm paying for Claude.
Speaker A:I got them all because we are on the edge and I want to be, and I want to understand the strength and weaknesses of each tool.
Speaker A:But the chat GPT history and its memory and its understanding of my projects and things that I've been working, my home projects or repainting cabinets and sending them down, like seriously detailed stuff that I can pull up at any time.
Speaker A:Say, hey, remember those cabinets we were working on?
Speaker A:One of them has a hole.
Speaker A:How do I patch it in the, in a solid wood and because it has pictures of the cabinets, it knows it's.
Speaker A:To me, it's invaluable.
Speaker A:And a lot of people discount that or don't even think about that.
Speaker A:So for corporate AIs, this is going to be a very similar, very similar thing.
Speaker A:When you work on a project, hey, remember the proposal to Johnson and Company?
Speaker B:Yeah.
Speaker A:So we want to make a provision that does this.
Speaker A:So this eliminates you from what?
Speaker A:Let's go to the folders, documents, let's look for contracts, let's look for past countries, let's look for Johnson and Company.
Speaker A:Oh, it's not there.
Speaker A:Or oh, that's an older version, or oh, you know, all you have to do is remember those chatgpt.
Speaker A:Can you rewrite that with a new provision?
Speaker B:Yeah, well, and adding on to that, it will help reduce the silo.
Speaker B:Because if you have, and I think Chat GPT calls them projects.
Speaker B:If you have a project that is a client and you, I think now you can add.
Speaker B:So it's not just my project, I can share it with you so we can work on the project together.
Speaker B:So when I say, hey, Alex, can you go add that new provision?
Speaker B:You don't have to ask me a bunch of questions, you can go talk to the chat GPT and it remembers everything I'VE talked to IT about.
Speaker B:You've talked to it about all the stuff.
Speaker B:And I think that's where we need to also go and with the whole lockdown and, and classifying of information and.
Speaker B:But then it's also.
Speaker B:The next question is maybe I have a higher clearance in the folder structure than you do.
Speaker B:Does it know that and doesn't show you stuff you shouldn't see.
Speaker A:Exactly.
Speaker A:So this is.
Speaker A:You hit the point on the head.
Speaker A:There is a governance.
Speaker A:So the, the difference.
Speaker A:So here's here, let me make that statement.
Speaker A:If ladies and gentlemen, you're watching.
Speaker A:Got to listen to this one.
Speaker A:Maybe we'll highlight this one.
Speaker B:Okay.
Speaker A:All right, here's, here's the one.
Speaker A:All right, you want to get.
Speaker B:All right, I'm ready.
Speaker A:If you're not paying for your AI subscription, your data is.
Speaker A:That's what's happening.
Speaker A:This is using your data to train and make it smarter.
Speaker A:How does it, you know, does it go into the main brain and, and dissipates there?
Speaker A:I don't know.
Speaker A:Does it stay there somewhere?
Speaker A:Maybe.
Speaker A:Likely.
Speaker A:And so with that said, when you have enterprise ChatGPT or business ChatGPT, you have admin console and controls over memories and access levels.
Speaker A:We can get granular here, but for the most part for like design company, it's pretty easy.
Speaker A:You know, designers have this, project managers have that, and the rest is exception that can be approved by your IT or your manager, whatever.
Speaker A:Again, not complicated.
Speaker A:But if those controls are not being deployed.
Speaker B:Right.
Speaker A:Yeah.
Speaker A:Anyone can access your project.
Speaker B:Exactly.
Speaker B:And your features competitor.
Speaker B:Yeah, yeah, exactly.
Speaker A:They're using two BIM consultants, you and some other person.
Speaker A:And now you're, you know, your stuff is exposed.
Speaker B:So go and ask the AI like, hey, by the way, what's the fee structure for the other guy?
Speaker A:Yeah, exactly.
Speaker A:Yeah, yeah, let's beat us beat him by a dollar.
Speaker A:So with this is, this is growing need for the product and you know, in my opinion, and I'm heading out this, we call it AI Labs at ARC it and this is the project and what we're going to build out as a larger focus and division for the company because I think if we don't do it, it'll be done to us.
Speaker A:So that's one way to think about it.
Speaker A:The other way to think about it is I think a lot of people need a lot of help with this.
Speaker A:At least this kind of, at least this initial way to govern it.
Speaker A:Not to govern it per se, but to securely adopt and then grow into more tools, more interesting use cases have internal meetings about how everybody's using stuff, knowing that we've done everything we could to stay safe but productive.
Speaker A:And those who are interested in talking to us about this product and other stuff we have for IT support, check us out@getarchitee.com and if you need BIM help, Aurora, BIM is very happy to help.
Speaker A:They're very, very busy.
Speaker A:They'll tell you.
Speaker A:They'll tell you all the time.
Speaker A:We're not looking for new clients, but they are.
Speaker A:They can handle one or two.
Speaker B:We'll make time for the best clients, the nicest people.
Speaker A:Yeah, it's been awesome.
Speaker A:Thank you for poking at me a little bit.
Speaker A:Liz, I appreciate your time.
Speaker B:This is fun.
Speaker B:I learned a lot.
Speaker A:We both did.
Speaker A:And thank you very much for watching.
Speaker A:We'll see you all next time.
Speaker B:Thank you.
