February 11, 2025 • 47 mins
AI is reshaping the AEC industry. Join us for a conversation with Bentley about how AEC firms can harness one of the most impactful technologies in decades to unlock the full potential of their data, streamline operations, and solve our most complex challenges.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
(00:04):
Making sure the data is digitalized,that it's well-structured from day one
so it can then be handed overto different people keeping that lineage
is going to be the breakthroughthat we need to be able
to expand the value of AIand those digital twins
into the complete lifecycleof the infrastructure.
You have a lot of firms out therethat are smaller firms
(00:25):
that do not have a lot of the capabilitiesof the larger firms
and I think the capabilities of Bentleybrings through that
and other AI applicationsactually is that great equalizer.
When we approach our design, thinkingabout our design holistically, and I think
for Bentley to build in toolswhere sustainanability
can directly be measuredas our design evolves is important.
(00:54):
I'm excited to be herefor another Viewpoints podcast episode
in the Building for Tomorrow seriesabout AI in the AEC industry.
I'm Dave Mulholland, VHB's ChiefTechnology Officer,
and today I'm joined by Julien Moutte,Chief Technology Officer
for Bentley Systems, a world leaderin the infrastructure intelligence,
and also my VHB Technologycolleague, Ryan Noyes.
(01:18):
So, Julien, we are really looking forwardto talking to you today
about AI's impact on Bentleyand how you are using AI right now
to address the evolving needswithin the AEC industry.
So, let's get right into it.
Can you start with telling our Viewpoint
listenersa little about your role at Bentley?
Well, first, Dave and Ryan,thanks for having me.
(01:39):
I'm really excitedto have that conversation.
So, I've joined Bentley four years ago.
I'm a software engineer by training,
but like infrastructure
engineers, I think it's all aboutfinding a better way to do things, right?
And we all are on the transformationjourney.
Very excited to see how the portfolioof products that Bentley is transforming
(02:03):
to integrate more and more AI capabilities
and to allow our
users, the engineering firmsand the engineers
to do more work,
to create better designs,to make better decisions
because they have the right tools
and the right, they have the right dataand context at the right time.
(02:24):
So, thanks for having me. Yeah.
Oh, absolutely.
And I know AI is not new for Bentley.
I know you guys have done a lot
through the assets and computervision and inspection.
So I look forward to hearing a little bitmore about that.
Ryan, let’s let the Viewpoints listenersknow a little bit about your role
and how you were helpingclients actually leverage AI.
(02:45):
Sure.
First of all, thanks for the invite, Dave.
Happy to be here with a couple of thoughtleaders in the industry.
I’m Ryan Noyes, I’ve been in the AECindustry for a little over 25 years now.
Started my careerdoing transportation design for a state
DOT here in the US and then moved to
VHB and, originally doing transportationwork.
(03:09):
Now, I manage a teamthat is focused on digital transformation
and digital delivery of projectsis one of the areas we're focused on.
For us, advancinghow our teams build models and develop
models, looking towards the futureof not only models as the legal document,
but also leveraging that data downstreamfor us where plans are purely
(03:33):
just a byproduct of our designis really what we're focused on, is
how can we move the deliverable forwardtoday,
with the ultimate target being towardsthat digital twin of the future.
Thank you.
Thank you both.Looking forward to the conversation.
So, Julien, as we get moving into this,
(03:54):
AI is not new for Bentley, right?
So, when you're really doing thisand, or sorry,
when you're looking forwardand talking about AI at Bentley,
what are some of the impacts that you'reseeing relative to the AEC industry?
So, well, there's a lot of differentkind of impacts.
I think we can talk about
(04:15):
the positive ones,but also some of the some of the concerns.
As you've mentioned,AI is not new for Bentley.
There are a lot of differentkind of technologies in AI.
And some of the ones that have been usedalready
extensively, for quite some time is aroundcomputer vision.
So this idea that we can use AI to analyze
(04:35):
pictures, videos on clouds,and then try to detect things,
in those, in those photos and videos,which can be useful for a lot of different
things, like, inspecting a bridge,inspecting a dam,
understanding the kind of deficienciesyou might have on the road,
potholes, signals and things like thatthat has been used for a while.
(04:57):
I think it's just accelerating now.
And basically this helps
engineers be much more productive,try to find things much faster.
So that's one.
Now there's a wholenew family of technologies around the AI
that is appearing now, which is generative
AI, the idea that theAI is creating new content and,
(05:17):
that is opening a lot of new doors,but also, a lot of new concerns.
Right? Right.
So, I mean, if we talk about concerns,
I mean, GenAI has a lot of potentialto improve the productivity of engineers,
but naturally, a lot of them are concernedfirst about their jobs.
Is AI going to replace me?
(05:39):
Is AI going to do my job?
And then there's anothergroup of engineers that are not afraid of
AI doing their jobs and replacing them,But as they work together
with AI as an assistant, as a copilot,then they become more productive
because of thisor maybe smarter in some way.
(06:00):
They are concerned about being ableto make the best use of it,
and that AI is providing them datathey can rely on, insights
that they can trust as they create designsor as they make decisions.
They want to make surethat the AI they use is,
is going to allow themto make the right decisions.
(06:21):
So, a lot of opportunities and concernsat the same time.
And, and there is a big responsibilityon the software vendors, us included,
in making sure we address those concernsand then open
all those doors to new opportunities.
So, Julien, I think that's a great point.
And, you know, I've heard numbersin the industry of 30 to 50 percent
(06:44):
of the work that we do in a typical sitedesign is just mundane tasks.
That's planned production, it is work
that really does not involve higher-levelthinking.
You know, how do you see AI and AI tools
One, giving time for engineers to do
(07:04):
other tasksthat, as we deal with a smaller
workforce, are really more criticalto a project team.
And then, how do we get our existing data
sets ready to help as we make thatAI assistant more intelligent?
You know,how is Bentley prepping for that?
That's, that'sa good point and I think we're hearing
(07:26):
the same, we're sharing the same numbersand reading of the market.
Right?
Our engineers, the one we're talking to,they're telling us that between
like 30 to 50 percent of the timeis spent on producing drawings, 2D
drawings from 3D digital,let's call them intelligent models.
And those are being compressedand flattened into 2D drawings,
(07:47):
and that is taking a lot of time.
So, I think AI here can help a lot by,
this is generative AI,this is generating that content.
It can help by automatinga lot of that mundane tasks.
And it can do thisby looking at previous drawings
to understand how you usually describeand convey
that information from a 3D digital designinto 2D drawings
(08:11):
and accelerate the work of those engineersto freedom to do more
rewarding tasksor to explore the source space More
extensively.
Right.
Because as they spend 30 or 50 percentof the time preparing those drawings,
they might not have enough timeto consider alternatives in the way
the designers site, for instance,or anything related to the infrastructure
(08:34):
to potentially look at alternativesthat would have a better outcome,
maybe by being less
intense on the impact that they doon the environment, or maybe reducing
the time and the energy that is reducedthat is used to construct
that specific site.
So doing all of this
additional explorationis something that will require more time
(08:56):
and that time is just not available.
So, automating a lot of that work withAI is going to free
a lot of time for those engineersto be able to work on other tasks.
I've heard some engineering firms
being concerned because they've
specialized some parts of their teamsto do that, that production of drawings.
So there indeed,there might be some situations where
(09:19):
if the AI is doing a good job at doingthose drawings well, they will need to
retrain or reorganize those resources tothen spend some time on,
on those kind of additional tasksthat are maybe not considered today.
And to the second part of your question,how do we prepare your data
sets to make sure you make the bestuse of AI.
That is a very interesting topic because,
(09:40):
as I've mentioned at the beginning,the concerns that the engineers
and also the engineering firms have ingeneral is about can we trust the data
that the AI is providing to us, but we canwe also trust AI
to respect our data, to protect it
so that it's not leaking out there,
and that our designsthat are informed by AI
(10:01):
models are not contaminatedin some way by the data from someone else.
And that's been a very importanttopic for us since we've started
integrating generative AI in our products,which is how do we ensure that the data
of our accounts of the engineering firmsremains their data,
and that they get to decidewhen it's being used for AI
(10:23):
and how this usage is appearing in
the decisions they take
by adopting some of the recommendations,
and the way we do that is by creating someAI models
that are useful without any datacoming from the engineering firms.
So, a generic model if you'd like.
And then you can specialize and retrainthat model with your own data sets,
(10:46):
but then that model is private to you.
It can only be used to
accelerate the workof your engineers on your projects.
And then you understand what data was used
to specialize and inform that AI model,
so that if you need to make any changein the future, like, for instance,
the data sets you've used for a specificproject was originally authorized,
(11:09):
or you thought it was possibleto use that data to train AI models
to inform new projects that you're workingon, What if that permission disappears?
How do you know?
How do you have the traceabilityto know that the five previous projects
you worked on afterhave been influenced by
that dataset that youoriginally authorized for training?
And if you remove that data set,you need to retrain the model
(11:31):
and you potentially need to look atwhat is the impact of those,
service projects because they were somehowinfluenced by data set.
And that isthat is very important in the way we are,
designing those AI modelsand integrating them into our products.
You know, Julien,that was actually one of the questions
we had coming in here was, the companyhas made a commitment to actually
(11:53):
drive the protection of IPand drive the protection
of the data in there,so I appreciate you talking about that.
Ryan, same similar question to you, right?
You know, with VHB actually partneringwith Bentley, what are some of the areas
that you're seeing the benefits as a,as a AEC user of the Bentley product?
(12:14):
You know, it's been excitingto kind of see how AI has,
has changed and evolvewith the Bentley tools that we've used.
Julian,we have been a SITEOPS user for years.
So that generative AI has been partof our toolset in standard practice
and to see it evolve into now
(12:36):
what we're seeing through OpenSite+,
you know, as, as one of the early testerswith that tool,
it's been great to see a modelthat is now, you know, that
large language model part of
AI where we can nowspeak to the design and say,
these are our concerns and get feedbackbased on that,
(12:57):
I think is really a game changerfor how we approach a design.
You know, thinking of buildingthat assistant to prompt you
for "I'm getting readyto submit environmental documents;
can you please check this for compliancewith my client requirement?"
Is really an interesting spaceto see come into our world.
(13:18):
You know, you think of that time savings
that you mentioned, Julien,and we think about it purely from
the "what are the drawing production,plan production?
What are the things we do today
that are time consuming on a project?"Well,
think about the ability to actually havea higher-level review take place.
So that entry-level reviewon a project, Dave, we just do with our
(13:42):
AI assistant, you know, your assistantsays, "did you remember to check this?"
Think of how many of usnow use some sort of home assistant
to turn on our lights, toturn off our lights, to turn on our music?
Like, I look at this and saythis is kind of the commercialization
of those kinds of tools.
You're going from a consumer modelto how we use these in the industry.
(14:05):
I don't knowif you have any thoughts on that, Dave.
Well, I like to flip my light
switch sometimes myself,but I do think the I do think that
the industry is moving in that direction.
The questions when I get into it isthe adoption of and the acceptance of it.
Right.
You know, is the industry willingto actually move in this direction?
(14:27):
And Julien, you mentioned, you know, thenot displacing some of those and I agree
100 percent it's going to be an upskillingof some of the engineers out there.
The question I
have is,
you know, as a software providerthat's in this space
and one of our strategic partners,are you seeing a lot of the adoption
(14:49):
of your tools using the AI,or are you getting a significant pushback?
And I think it's almost
a three-part questionbecause you've got the private sector,
you've got the public sectorfrom the engineering side,
and then you got the public sectorfrom the review side.
Right.
And I think they're allon different levels in terms of maturity.
Right.
(15:10):
So, are you seeing anyone move fasterrelative to the private sector versus
the public sector in the adoption of theAI using your tools?
I think it's still a bit earlyto know about that
because OpenSite+was announced in October.
There is an early access programthat is running right now, so
(15:31):
we're starting to get our feedbackfrom the early adopters.
But I think those are the onesthat are like at the bleeding edge, right?
I mean, they are interestedin testing and experimenting with it.
So, I don't know yet if we
have the feedback from the ones that arepotentially more inclined to push back,
to understandwhat are the kind of concerns they have.
(15:53):
What we've done, though, is
that we've tried to make sure that theAI is not in the way.
Right?
I mean, it's really here to assist them.
And, Ryan, you've mentioned the copilotthat is
allowing you to talk to your model.
You don't have to, right?
I mean, you can use all the modeling toolsand do all the kind of queries
and review yourselfthat the model is really here to try to
(16:16):
work as a copilot and helpthe engineers do that faster.
And as we look at automating the production of drawings that we do, as we look at,
maybe doing some
kind of collaborationwhere the design that is proposed,
there is a, there is a tool in OpenSite+to optimize your layout.
And that is a toolwhich we envision to be collaborative
(16:39):
in some sense that you proposesome options, And then two engineers
is speaking one option,and then the model learns from this.
And then throughsome kind of a reinforcement
learning technique, try to understandexactly what you have in mind,
which might not be somethingthat AI would calculate immediately,
but that is somehow carrying your savoirfaire, your know-how, into the design.
(17:01):
So, if AI is not in the way,if you can trust what you're doing,
we expect that pushbackthat we touched on here to be minimal.
But, I still think we need to confirmthat as we get more adoption.
So, I think we'll. Don't you agree?
I do. We'll park that for a futurepodcast, Julien.
We will bring you back
after we start to see some of the resultscoming in from that.
(17:22):
And you opened it up, and I appreciate youdoing that with the OpenSite+,
having seen some of the
presentationson that extremely powerful tool.
Right.
So, thank you To Bentleyfor pushing that forward.
Maybe for the Viewpoint listeners,maybe a high-level overview of
the OpenSite+ tool,what it is, what it can
(17:43):
do, and where do you see itgoing in the future?
Because I looked at it, I've been watchingsome of the roll outs of this,
And VHB is obviously,you know, looking at this
as one of those opportunities and
seeing what Bentley has doneand seeing the opportunities
is just amazing and game changing.
(18:05):
Yeah. So,
there's a lot
we're hoping to dowith that new family of products.
Right.
Because I think OpenSite+is the first of a family of products
that are digital twin native,meaning that instead of generating funds,
as we've done for the past,for the past decades,
you're starting from somethingthat provides you the context.
(18:28):
So, you don't start from scratch.
You start with the context of the map,the terrain.
And you probably heardthat we acquired Cesium recently.
So Cesium is bringingthat geospatial context,
to allow you to start with
the real conditions that are capturedfrom the digital twin of the world
and then from this, you layermore and more, additional layers of data,
(18:51):
so that the work that you start withis already
influenced by all of that contextcoming from the requirements
that are passed by AI,the geospatial data that we bring together
with the terrain, with the, the,the, the, the, the not to say that,
the zoning ordinancethat is provided by different can of tools
of GIS layersthat all of that are brought together
(19:14):
and then the engineersare able to start the work
much faster because all of that contextis provided here.
And I think that's something we're goingto see in lots of different products.
OpenSite+is the place we're starting from.
But when you think about it, when you'redesigning a bridge, when you're
designing a highway,when you're designing a railway
(19:35):
or even a power plant,you need that context always.
Right?
So that's what I think is truly inspiringis this vision
of starting with a digital twinof something that is yet to be,
bringing
as much data as possible in contextfor the engineers
to work in the most efficient way,
And then thinking the productwith an AI-first mindset.
(19:57):
But because it's very difficult,I think, to introduce AI in the,
in an existing product because there areso many things you would do
differentlyif you would have access to AI models,
if you would be able to talk to the data,well talk between quotes.
Right?
But I'm not always meaning this vocally,although
it's coming.
(20:18):
It's possible.
But this idea that your applicationcan understand the smart
objects and answer questions about those
and that you can automatea lot of the work by just.
I think a model understanding what is theintent that you're trying to have
and then execute a lot of those tasksare automated for you or speeding up.
So, I really envision that
(20:40):
as we are bringing
together the data that we have access tonow with our partnership with Google,
the data that Cesium already had accessto, bringing all of this into a
replica of the world and having a productthat is designed for that.
It uses AI as a wayto accelerate the work of engineers
from the first force,from first principles is going to create
(21:02):
a completely whole new family of productsfor engineers, which are going
to supercharge their capabilitiesto create the best possible designs.
And I mean about these sustainable,resilient designs,
but also being more productiveon addressing the backlog, which is huge,
and what that is a prime right nowis the challenge of shortage.
And deliver digital twins that will alsounlock a lot of potential outcomes,
(21:24):
all over the lifecycleof that infrastructure.
So, the part I
just I was so enthralledlooking at the OpenSite+
was the capability to go from pure concept
development to final design plans.
And then digging into it
and understanding how do you actually uselocal codes by, you know,
(21:45):
and you do it through your Copilot accessthat you're doing.
So actually, localizing it,which is one of the biggest challenges
when you think about rolling outan enterprise
across the entire globe of whatyou're doing, to go to optimization. And.
Right, and that's, that's key, right.
So, you know, as an engineer,and I am an engineer,
to be able to have the ability to do siteoptimization at your fingertips.
(22:07):
Simply amazing.
And that's where those younger engineers
or the engineers could sit thereand look and say, what
and why do I have these three or fouror five choices of site optimization?
What makes it better?
So, it's the pluses and minuses of it.
So yes, I'm a fan.
I'm all in.
Ryan, you know thatwe started talking about the OpenSite+.
So, how do you see from VHB's perspectivethe use of the OpenSite+,
(22:31):
you know, with what you're doingand driving,
the model-based designacross the footprint.
Sure.
So, you know, with our focus of buildingintelligent models,
you know, as being kind of our visiontowards the end of 2025.
That any project we start,will be started with the vision of it
being an intelligent model that we canextract data from throughout the process.
(22:53):
You know, the plans are just a byproduct.
One of the things that excites mewith with OpenSite+ is to think of
not losing datathroughout the evolution of a design.
So, so, Dave,you think about our process today.
Okay.
We've done our preliminary designin, in maybe we did it in concept station.
Maybe we did it, on a pieceof tissue paper or whatever we did it on.
(23:16):
We did that first concept.
Then you have that loss in datawhen you start to move it into real design
and then, oh, well,now you need to do traffic analysis
or you need to do drainage analysis.
Oh gosh.
There's another point where you'relosing data because you're extracting.
And even when we do our bestto connect those models together,
(23:38):
unless you're truly in one environment,you do lose data
through those steps in the process.
And there's reworkand potential for error.
So I think the thought of having that,that process just evolve and oh,
I now need to do my pipe sizingand having that just be a function
within OpenSite+ where all those toolsare accessing that model.
(24:01):
You know,I think, Julien, when I think of AI,
I think of data being
the fuelthat actually drives that AI engine. And
one of the struggles we have is
what is the quality of the datawe're putting into that AI engine.
So, having that dataall be kind of curated from the start,
(24:22):
Dave, in a platformI think is really, really powerful versus
this dispersed data that, you know,we fight the battle every day in the AEC
industry of is that the right sourcefor that information?
You know, that single source of truth,I think is really powerful for teams.
I don't know, Julien.
Do you have any thoughts? Yeah.
(24:43):
Well, I was about to ask you a question,which is I mean,
I think that continuum of datain one environment,
and then making this available to AI.
So, we've talked about AI relievingthe engineers from mundane tasks
like creating the drawings, et cetera,but what is the kind of next
big milestone
that you see for the engineersthat are working on those site designs,
(25:04):
with all that data available to AI,what are the kind of big
accelerators that you believeshould be implemented next?
You know, in OpenSite+?
Well, around
OpenSite+ I am excited to see it injust started to see
some of the drainage componentsrecently added into that.
And I think that that disconnectthat, that
(25:24):
when we talk about sustainabilityand we talk about impact
to our environment, it'snot just about our,
our site layout,it's about considering things like,
"okay, what is my embodied carbonin this design?
What am Iwhat can I do to reduce my footprint?"
You know, I have children, Julien,that are run around the house,
(25:46):
and I want to leave the world in a betterplace for them than where I got it.
And part of
that is when we approach our design,thinking about our design holistically.
And I think for Bentley to build in tools
where sustainabilitycan directly be measured as our evolve,
our design evolves, is importantin that time
(26:08):
that we're gaining that 30 to 50 percent,whatever that number ends up being,
shouldn't be timethat we don't spend on our design.
It should be time that we spendon improving our design.
And to do that, what we need to dois have those measures presented to us.
Like at any point during my design,I want to be able to go into it
and say, okay,
(26:30):
you know what, let's change my pavementtype to all be impervious pavement.
What does that do, Not only to my cost,but what does that do to my,
you know, sustainability index or whateveryou want to call that.
So, that excites mefrom really getting our teams to
have the time to produce the best designthat we can,
(26:50):
not just produce the first designthat meets our requirements.
I'm really happy to hear you say that,because that's,
I think, where AI can bethe most helpful because,
one of the things I likeabout the portfolio of tools we have
is that we have a lot of tools that arevery good at doing some specific things.
(27:12):
But it's difficult for an engineer to runall of those tools, right?
I mean, doing a simulation for flooding,doing a simulation
for electricity, doing a simulation
for mobility of pedestriansor cars and things like that.
I mean, we have all of those tools,but for an engineer, it's
still a lot of time to goand do all the simulations to make sure
they have the right informationto take the right decision.
(27:34):
Now, AI is, potentially
a concern for some because they say,oh, AI might be creating something
that is a bit of a destinationand that could be wrong for us.
But if you use AI not to kind ofcreate content out of the blue,
but instead run all of those simulationengines for you in the background
with your designs,and then provide the feedback
(27:56):
and tell you, well, by changing thisand this and that, we get a better score
on Drainage for flooding or by doing thisand this change, we construct
20 percent faster and we use 20 percentless concrete, for instance.
Well, then you have you have all the toolsto make the right decision
to achieve those sustainabilityresilience goals that you have.
(28:17):
So, I fully agree.
If you have a client who's interested in
how do I capture solar energy on this siteas I redevelop it,
that's information that using thatAI model you can, or the AI assistant,
you can quickly go through and providea few iterative design alternatives where
how much do we weighcapturing solar energy back versus
(28:40):
how much do we balance that versus costof the overall development of the site.
So, I think it's interesting.
It has so many differentareas for potential gains.
You know, a question that I havefor you around that,
you know, it's it's very clearthat that that digital twin
is how Bentley is looking at leveraging
AI so that iTwin environment.
(29:04):
You know, how do you see thatas it goes outside of design like,
and goes to either constructionor into maintenance of operations?
Is Bentley positioning other tools to,
to access that same iTwin framework, and
you know, AI to leverage that data?
Well, our vision is that,
(29:26):
and this builds on what you wereexplaining before, there is a lot of data
loss and all of ours are happeningbetween one stakeholder to another.
And I think a lot of engineering
firms are looking at expandingtheir portfolio of services,
not only to design, but thento construction, and then to operations.
And that is going to address some of this.
(29:46):
But there is still quite a bit of,and all of ours happening
between different stakeholders right now,
including the final ownerand operator of the infrastructure.
And we kind of have a responsibility
to make sure that there is aslittle data loss as possible.
And that's where we
this concept of bringing the data togetherin a common data and augment
(30:06):
or a digital twin of the infrastructurethat keeps evolving together
with the infrastructure itselfis so important to us because instead
of having handovers of papersthat are then remodeled by someone else,
and there is a lot of data loss there,We believe that, well, what we can
handover is actually access to the datawith an audit trail,
(30:28):
a traceability back to the originso that we can
then start to unlock a lot of the outcomesthat are becoming possible.
If AI has access to the whole datalineage,
of the project,
unlocking new possibilitiesin terms of the operation
phase or the construction phase,or even like when you need
(30:50):
to do a renewal
of the infrastructure, I think access toall of that information is precious.
And, we still not there yet,
but I believe that making surethe data is digitalized, that it's
well structured from day one, so it canthen be handed over to different people
keeping that lineageis going to be the breakthrough that
we need to be ableto expand the value of AI
(31:12):
and those digital twins into the completelifecycle of the infrastructure.
And that that is going to also supportso much of those resiliency
and sustainability goalsthat you mentioned before.
So, Dave, I want you to add to that one.
You know, as, as we or VHB
are focusing on things like AIand model-based design, you know, we've
(31:36):
heard Julien mentioned data several timesand, you know, data format and data loss.
How do you, as someone who's dealingnationally with data standards,
how do you see national data standardscoming into play with AI?
Well, Ryan,
before we get to the national standards,because I knew that was going to come up.
(31:57):
You did ask a question.
And this is, you know, Julien,it is an interesting when you're talking
about the digital twin technologyand the virtual representation
of the real worldof what we operate in and live in.
Right?
I think initiallysome of the percent efficiency
that you're referring to, the 30,the 50% become even more profound
(32:18):
when you start to combine those tools,those modules, into one full digital twin.
So, when you think aboutintegrating the stormwater
and with the design and withthe transportation, does a couple things.
It actually eliminates a lot of the errorsby not communicating those tools, right?
So, as the engineers are looking atmarginalizing risk,
(32:40):
that is one area that you'remarginalizing risk by doing that.
And I think that's great to actually bringthem into one centralized environment.
And then, you know, the second part is
when you think about clash detectionor risk management, right?
Having the ability to say, okay,we're going to do this
design and is against whatRyan wants to do with the sustainability,
but it triggers something else
on a different area or a different modulethat's phenomenal, right?
(33:03):
In terms of what it does.
But the models, Ryan, like you say,are only as good as the data goes into it.
And there are not a lot of good, there
not a lot of standards,across the US based on this.
Right. So, what what are we trying to do?
You know, at VHB, we're trying to come upwith our own governance and standard
(33:24):
to put into it,but it's more of an industry challenge.
Right. Of where it is. Right.
So, what is Bentley doing also to help
with standardizingthe data goes into the models
so that it is transferable from agencyto agency, state to state,
on what it isor, you know, is that an issue?
(33:48):
It is, it
is an issue for multiple reasons, right?
The first one is first agreeing onwhat do we handover
from one party to anotherto make sure that the data
is transferred
without loss of fidelity,without loss of information.
(34:08):
And, that I think has been addressedin some sense with
quite a few efforts on creating standards.
Building smart,for instance with IFC has been helping
in defining standardsfor data exchange between stakeholders.
But we believe we need to goone step further
in the sense that,
(34:29):
as engineering firmsare creating that data
in the tools that they use,
we also need to make surethat they build all of that
data structure,that digital twin, in some sense,
in a way that also gives them the safety
that they are not dependingon one single vendor for that.
(34:50):
Right? So, one of the things
we've
talked about recently a lot in our yearin infrastructure is that with Cesium
and the iTwin platform and bringing themtogether in one common platform.
We also try to have an open approachin everything we do,
so that not only are we talking aboutthe standard for the industry by
(35:10):
how do we exchange datawhen it comes about, design files, etc.
but how do we make sure that all the datais collected in an infrastructure
project, which is not only about models,it's also about the IoT data,
the GIS data,
the reality meshes, the surveys, etc.
How do we make sure that all of that data
that you gather in one placeso that you can leverage it with AI,
(35:33):
or that you federate from differentsystems into one common data construct?
How do you make sure that this data
is going to livefor as long as the infrastructure itself,
that is described using openstandards that everybody can understand.
So, one of the open standards we've talkeda lot recently is the 3D Tiles,
which is an open, an OGC, Open Geospatial
(35:54):
Consortium standards
that is making it possiblefor everyone to read that data.
But there's many more like the featureservice, from OGC,
which is exposingGIS data and interesting,
interesting data being surfaced
on the 3D visualization canvas.
(36:16):
But also, how do you store the BIM dataabout your model?
So, we're looking at creating standardsfor all of those
that are going to facilitatethe interoperability of the data
between vendors,
but also making surethat you can take your data out
and move it to another platformif you want to,
and then continueto work in this with this.
Because if you try to bring all of thatdata in one common environment
(36:39):
as we've described before, but all thatdata is in some proprietary format,
then you're locked up with a single vendor
and the interoperabilityis going to be complicated.
And as you want to use that datain ten years from now has
you need to do some workon this infrastructure project,
or if that tool has disappeared, orthe vendor is not servicing you anymore,
then that datathat all of that value is lost, right?
(37:02):
So, the standards are important onboth fronts.
How do your national standardsfacilitate the exchange
between parties, but also for youas you make that investment in your data?
How does the standards helpyou ensure that this is open
and that you will be ableto leverage that data in the long run?
What's your view on that?
Ryan,
I turn to you on this and say,what's your perspective?
(37:24):
You and I had some conversationaround IFC. Right.
And where the standards are.
For me, Julien,
there has not been a common
standard across,and we operate within different platforms.
Right.
So, we operate in multiple states,which creates a challenge
in terms of the uniformity of the datathat comes out. Right.
(37:45):
So that creates an issue for us,you know, as a firm
that operates in multiple states.
To basically try and come upwith the national standards
that you're referring to that operatewithin
any platform, the full interoperability
is going to take a Herculean effortthat's not there today.
(38:06):
Right.
But it is actually under the focusesas we all talk.
You can't do good AI, you can't dogood business models, you can't do
good digital twins unless you havesome standardized format working with,
you know, agencies and vendorslike yourself.
So, this is one of the biggest pushesthat VHB is actually making.
We have differentareas that we're talking to,
(38:30):
agencies, partners,
policymakers, relative tohow do we actually create these standards?
What's what does it meanfor the governance side of it?
And then what do we do within our own datathat we control,
but also some of the datathat we're actually working with
state and local agenciesrelative to how they process their data
(38:50):
as part of their datalake and repositories.
Well, Dave, to build on that,I think it's exciting
that we're actually startingto see some momentum around
that with different DOTsacross the US right now
looking at what is their new digitaldelivery standard going to be?
You know,
(39:11):
we're working with a number of clientshere on the East Coast where that is
a specific project that they have.
It's not arounda particular design project,
but it's aroundwhat is our digital data standard.
And I think thatwill help in having AASHTO,
you know, Julien being in Europe,
(39:33):
your standards are a bit more evolved
than the US standards as of todayaround horizontal construction,
but having AASHTO actually start
to release this is some levelof development guidelines.
This is what our data standardfor the industry we recommend.
I think those are really powerfuland helping us
(39:56):
have a target
so that we can ensure that we havehigh quality data in these environments.
Like to me,
level of developmentneeds to be a discussion on every project
so that the expectation is set onwhat is the quality of that data.
You know, Julien,I can bring in data from any partner
(40:18):
but understanding what the intended
quality of that data is is really criticalas we're doing design work.
Absolutely.
Yeah.
So, I agreeAASHTO has as a big role to play here.
I think that we see the marketevolving in that direction, right?
I mean, we've seen multiple DOTswork on kind of research projects.
(40:41):
There's been some public funding aroundthis, how to create the digital
delivery models of tomorrowto be able to manage the tracking
of progress, to manage the deliverablesas a digital model.
So, I think there's a lot of,there's a lot of activity
that we're startingto see that in that direction.
It would be a lot about the right rhythm
(41:03):
in adapting to that moving landscape.
And in the meanwhile,we need to also make sure we provide
all the foundationsfor that technology to exist. So,
that's all we have.
We have engaged in multiple openstandards bodies or Bentley
is part of the Khronos, an institute
(41:23):
for the standardization of gITF,or 3D geometries.
We are part of OGCas an important member right now
to discuss all of those standards aroundhow do you represent today's data?
How do you represent 3D tiles?
We also talking with Metaverse
Forum and the Alliancefor OpenUSD to discuss how do we,
(41:44):
how do we make sure that we have openstandards to represent that data across
vendors in an open way so that as thosenational standards are evolving,
you always have the warranty that you havedata in the format that you understand.
That is the standard that is open.
And that you canthen massage and transform
into those national standardsas they start to emerge.
(42:06):
I have one broader question.
Put the crystal ball out there,
5 to 10 years. Crystal ball, right?
5 to 10 years.
Where do we see the GenAIin the design services in 5 to 10 years
from now?
(42:27):
So, I think that GenAI is goingto be connected
to more and more toolsthat are not artificial intelligence,
but actually human intelligencethat has been ingrained
in softwareto explore more and more alternatives.
So, I think right nowwe're seeing tools like OpenSite+
(42:50):
and others in the marketuse techniques like large language models
to find information faster, maybe to talkto small objects in some sense.
And I really envision that those toolsare going to become smarter and smarter
by being informedby many other kind of data sources
and algorithms that they can then employ
(43:12):
to become the best possible assistantfor the engineers.
So, we talked about embodied carbon,for instance,
being able to look at the designas it's being designed by the engineers
and then providing the carbon footprintis something
that AI assistants should be able to doin the coming year or two.
And that will already have massive impacton the quality of the designs
(43:33):
that are being produced,
but then bringing all the perspectivearound structural analysis,
mobility simulations,but also informing that mobility
simulation with datathat comes from the world.
Right?
I mean, Google has a lot of informationabout mobility.
You have millions of peoplerunning on the road every day with Waze
(43:53):
and there is a lot of datathat is being collected by dash cams
informing the design decisions,With all of that crowdsource
data is also somethingthat is going to unlock massive potential.
So, in 5 to 10 years, I'm envisioninga world where the engineering tools
that you're using are not just a toolthat is helping you build 3D shapes
(44:14):
and to put some metadataaround those structures,
but actually see a complete what-ifscenario, right?
I mean, here are the traffic flowson that highway for the past five years,
the stormsthat have been flooding that region
and how those would potentially impactyour design right now
and how you couldpotentially make some changes
(44:36):
to adapt to this,so that your design is more resilient.
And I think engineers are goingI mean, it's going to be very exciting
times to be an engineer,because having all of that context
and becoming much smarterabout your designs is actually going
to be very fulfilling, I believe, for,the engineers at the entering firms
because you will have so much contextavailable
(44:56):
at the at your fingertips,
to create so much
of better designs, thatI think is going to be truly exciting.
So, 5 to 10 years, I really see thatconvergence of data and tools,
elevating the engineers
to capabilities that were not thoughtto be possible before.
(45:18):
You know, Julien, how exciting will it bewhen we do get to that world of,
okay, we've had our designed digital twinthat went to construction and then,
you know,you had mentioned visual learning models
and tools like Blyncsythat you have out there today,
you know, okay, these dash camsnow are providing feedback to the owners
around, hey, you know what?
(45:38):
There's a maintenance issueyou need to address over here,
whether it's a drainage issueor whether it's,
a safety issuewith a guardrail missing or whatever.
But, you know, you think about those typesof tools really
being able to inform inand advance our design process.
Prioritizingsomething from, hey, AI noticed
(45:58):
there's a high number of near missesat this intersection.
Maybe this should be a priorityaround construction
for pedestrian safety or whatever.
Like that to me is,
to the crystalball, Dave, that's not sitting here today
and we're doing it on every projectbut five years, ten years down the road.
(46:19):
I think with good data,good digital twins,
and using the tools as they evolve,we really have a chance to like
make a significant differencearound safety and
around the experiencethat your traveler has.
You know, self-driving cars,If all of a sudden they're making us aware
of problems in the road through AI,
(46:41):
fantastic to get to that point.
So, Ryan, I think that AI in the work
that Bentley and Julien is doingis actually the great equalizer.
We have a lot of firms out therethat are smaller
firms that do not have a lotof the capabilities of the larger firms.
And I think the capabilitiesthat Bentley brings through that and other
(47:03):
AI applications actually,is that great equalizer that can help
normalize, I think, the industryand move the entire industry forward.
Making sure the data is digitalized,that it's well-structured from day one
so it can then be handed overto different people keeping that lineage
is going to be the breakthroughthat we need to be able
to expand the value of AIand those digital twins
into the complete lifecycleof the infrastructure.
You have a lot of firms out therethat are smaller firms
(00:25):
that do not have a lot of the capabilitiesof the larger firms
and I think the capabilities of Bentleybrings through that
and other AI applicationsactually is that great equalizer.
When we approach our design, thinkingabout our design holistically, and I think
for Bentley to build in toolswhere sustainanability
can directly be measuredas our design evolves is important.
(00:54):
I'm excited to be herefor another Viewpoints podcast episode
in the Building for Tomorrow seriesabout AI in the AEC industry.
I'm Dave Mulholland, VHB's ChiefTechnology Officer,
and today I'm joined by Julien Moutte,Chief Technology Officer
for Bentley Systems, a world leaderin the infrastructure intelligence,
and also my VHB Technologycolleague, Ryan Noyes.
(01:18):
So, Julien, we are really looking forwardto talking to you today
about AI's impact on Bentleyand how you are using AI right now
to address the evolving needswithin the AEC industry.
So, let's get right into it.
Can you start with telling our Viewpoint
listenersa little about your role at Bentley?
Well, first, Dave and Ryan,thanks for having me.
(01:39):
I'm really excitedto have that conversation.
So, I've joined Bentley four years ago.
I'm a software engineer by training,
but like infrastructure
engineers, I think it's all aboutfinding a better way to do things, right?
And we all are on the transformationjourney.
Very excited to see how the portfolioof products that Bentley is transforming
(02:03):
to integrate more and more AI capabilities
and to allow our
users, the engineering firmsand the engineers
to do more work,
to create better designs,to make better decisions
because they have the right tools
and the right, they have the right dataand context at the right time.
(02:24):
So, thanks for having me. Yeah.
Oh, absolutely.
And I know AI is not new for Bentley.
I know you guys have done a lot
through the assets and computervision and inspection.
So I look forward to hearing a little bitmore about that.
Ryan, let’s let the Viewpoints listenersknow a little bit about your role
and how you were helpingclients actually leverage AI.
(02:45):
Sure.
First of all, thanks for the invite, Dave.
Happy to be here with a couple of thoughtleaders in the industry.
I’m Ryan Noyes, I’ve been in the AECindustry for a little over 25 years now.
Started my careerdoing transportation design for a state
DOT here in the US and then moved to
VHB and, originally doing transportationwork.
(03:09):
Now, I manage a teamthat is focused on digital transformation
and digital delivery of projectsis one of the areas we're focused on.
For us, advancinghow our teams build models and develop
models, looking towards the futureof not only models as the legal document,
but also leveraging that data downstreamfor us where plans are purely
(03:33):
just a byproduct of our designis really what we're focused on, is
how can we move the deliverable forwardtoday,
with the ultimate target being towardsthat digital twin of the future.
Thank you.
Thank you both.Looking forward to the conversation.
So, Julien, as we get moving into this,
(03:54):
AI is not new for Bentley, right?
So, when you're really doing thisand, or sorry,
when you're looking forwardand talking about AI at Bentley,
what are some of the impacts that you'reseeing relative to the AEC industry?
So, well, there's a lot of differentkind of impacts.
I think we can talk about
(04:15):
the positive ones,but also some of the some of the concerns.
As you've mentioned,AI is not new for Bentley.
There are a lot of differentkind of technologies in AI.
And some of the ones that have been usedalready
extensively, for quite some time is aroundcomputer vision.
So this idea that we can use AI to analyze
(04:35):
pictures, videos on clouds,and then try to detect things,
in those, in those photos and videos,which can be useful for a lot of different
things, like, inspecting a bridge,inspecting a dam,
understanding the kind of deficienciesyou might have on the road,
potholes, signals and things like thatthat has been used for a while.
(04:57):
I think it's just accelerating now.
And basically this helps
engineers be much more productive,try to find things much faster.
So that's one.
Now there's a wholenew family of technologies around the AI
that is appearing now, which is generative
AI, the idea that theAI is creating new content and,
(05:17):
that is opening a lot of new doors,but also, a lot of new concerns.
Right? Right.
So, I mean, if we talk about concerns,
I mean, GenAI has a lot of potentialto improve the productivity of engineers,
but naturally, a lot of them are concernedfirst about their jobs.
Is AI going to replace me?
(05:39):
Is AI going to do my job?
And then there's anothergroup of engineers that are not afraid of
AI doing their jobs and replacing them,But as they work together
with AI as an assistant, as a copilot,then they become more productive
because of thisor maybe smarter in some way.
(06:00):
They are concerned about being ableto make the best use of it,
and that AI is providing them datathey can rely on, insights
that they can trust as they create designsor as they make decisions.
They want to make surethat the AI they use is,
is going to allow themto make the right decisions.
(06:21):
So, a lot of opportunities and concernsat the same time.
And, and there is a big responsibilityon the software vendors, us included,
in making sure we address those concernsand then open
all those doors to new opportunities.
So, Julien, I think that's a great point.
And, you know, I've heard numbersin the industry of 30 to 50 percent
(06:44):
of the work that we do in a typical sitedesign is just mundane tasks.
That's planned production, it is work
that really does not involve higher-levelthinking.
You know, how do you see AI and AI tools
One, giving time for engineers to do
(07:04):
other tasksthat, as we deal with a smaller
workforce, are really more criticalto a project team.
And then, how do we get our existing data
sets ready to help as we make thatAI assistant more intelligent?
You know,how is Bentley prepping for that?
That's, that'sa good point and I think we're hearing
(07:26):
the same, we're sharing the same numbersand reading of the market.
Right?
Our engineers, the one we're talking to,they're telling us that between
like 30 to 50 percent of the timeis spent on producing drawings, 2D
drawings from 3D digital,let's call them intelligent models.
And those are being compressedand flattened into 2D drawings,
(07:47):
and that is taking a lot of time.
So, I think AI here can help a lot by,
this is generative AI,this is generating that content.
It can help by automatinga lot of that mundane tasks.
And it can do thisby looking at previous drawings
to understand how you usually describeand convey
that information from a 3D digital designinto 2D drawings
(08:11):
and accelerate the work of those engineersto freedom to do more
rewarding tasksor to explore the source space More
extensively.
Right.
Because as they spend 30 or 50 percentof the time preparing those drawings,
they might not have enough timeto consider alternatives in the way
the designers site, for instance,or anything related to the infrastructure
(08:34):
to potentially look at alternativesthat would have a better outcome,
maybe by being less
intense on the impact that they doon the environment, or maybe reducing
the time and the energy that is reducedthat is used to construct
that specific site.
So doing all of this
additional explorationis something that will require more time
(08:56):
and that time is just not available.
So, automating a lot of that work withAI is going to free
a lot of time for those engineersto be able to work on other tasks.
I've heard some engineering firms
being concerned because they've
specialized some parts of their teamsto do that, that production of drawings.
So there indeed,there might be some situations where
(09:19):
if the AI is doing a good job at doingthose drawings well, they will need to
retrain or reorganize those resources tothen spend some time on,
on those kind of additional tasksthat are maybe not considered today.
And to the second part of your question,how do we prepare your data
sets to make sure you make the bestuse of AI.
That is a very interesting topic because,
(09:40):
as I've mentioned at the beginning,the concerns that the engineers
and also the engineering firms have ingeneral is about can we trust the data
that the AI is providing to us, but we canwe also trust AI
to respect our data, to protect it
so that it's not leaking out there,
and that our designsthat are informed by AI
(10:01):
models are not contaminatedin some way by the data from someone else.
And that's been a very importanttopic for us since we've started
integrating generative AI in our products,which is how do we ensure that the data
of our accounts of the engineering firmsremains their data,
and that they get to decidewhen it's being used for AI
(10:23):
and how this usage is appearing in
the decisions they take
by adopting some of the recommendations,
and the way we do that is by creating someAI models
that are useful without any datacoming from the engineering firms.
So, a generic model if you'd like.
And then you can specialize and retrainthat model with your own data sets,
(10:46):
but then that model is private to you.
It can only be used to
accelerate the workof your engineers on your projects.
And then you understand what data was used
to specialize and inform that AI model,
so that if you need to make any changein the future, like, for instance,
the data sets you've used for a specificproject was originally authorized,
(11:09):
or you thought it was possibleto use that data to train AI models
to inform new projects that you're workingon, What if that permission disappears?
How do you know?
How do you have the traceabilityto know that the five previous projects
you worked on afterhave been influenced by
that dataset that youoriginally authorized for training?
And if you remove that data set,you need to retrain the model
(11:31):
and you potentially need to look atwhat is the impact of those,
service projects because they were somehowinfluenced by data set.
And that isthat is very important in the way we are,
designing those AI modelsand integrating them into our products.
You know, Julien,that was actually one of the questions
we had coming in here was, the companyhas made a commitment to actually
(11:53):
drive the protection of IPand drive the protection
of the data in there,so I appreciate you talking about that.
Ryan, same similar question to you, right?
You know, with VHB actually partneringwith Bentley, what are some of the areas
that you're seeing the benefits as a,as a AEC user of the Bentley product?
(12:14):
You know, it's been excitingto kind of see how AI has,
has changed and evolvewith the Bentley tools that we've used.
Julian,we have been a SITEOPS user for years.
So that generative AI has been partof our toolset in standard practice
and to see it evolve into now
(12:36):
what we're seeing through OpenSite+,
you know, as, as one of the early testerswith that tool,
it's been great to see a modelthat is now, you know, that
large language model part of
AI where we can nowspeak to the design and say,
these are our concerns and get feedbackbased on that,
(12:57):
I think is really a game changerfor how we approach a design.
You know, thinking of buildingthat assistant to prompt you
for "I'm getting readyto submit environmental documents;
can you please check this for compliancewith my client requirement?"
Is really an interesting spaceto see come into our world.
(13:18):
You know, you think of that time savings
that you mentioned, Julien,and we think about it purely from
the "what are the drawing production,plan production?
What are the things we do today
that are time consuming on a project?"Well,
think about the ability to actually havea higher-level review take place.
So that entry-level reviewon a project, Dave, we just do with our
(13:42):
AI assistant, you know, your assistantsays, "did you remember to check this?"
Think of how many of usnow use some sort of home assistant
to turn on our lights, toturn off our lights, to turn on our music?
Like, I look at this and saythis is kind of the commercialization
of those kinds of tools.
You're going from a consumer modelto how we use these in the industry.
(14:05):
I don't knowif you have any thoughts on that, Dave.
Well, I like to flip my light
switch sometimes myself,but I do think the I do think that
the industry is moving in that direction.
The questions when I get into it isthe adoption of and the acceptance of it.
Right.
You know, is the industry willingto actually move in this direction?
(14:27):
And Julien, you mentioned, you know, thenot displacing some of those and I agree
100 percent it's going to be an upskillingof some of the engineers out there.
The question I
have is,
you know, as a software providerthat's in this space
and one of our strategic partners,are you seeing a lot of the adoption
(14:49):
of your tools using the AI,or are you getting a significant pushback?
And I think it's almost
a three-part questionbecause you've got the private sector,
you've got the public sectorfrom the engineering side,
and then you got the public sectorfrom the review side.
Right.
And I think they're allon different levels in terms of maturity.
Right.
(15:10):
So, are you seeing anyone move fasterrelative to the private sector versus
the public sector in the adoption of theAI using your tools?
I think it's still a bit earlyto know about that
because OpenSite+was announced in October.
There is an early access programthat is running right now, so
(15:31):
we're starting to get our feedbackfrom the early adopters.
But I think those are the onesthat are like at the bleeding edge, right?
I mean, they are interestedin testing and experimenting with it.
So, I don't know yet if we
have the feedback from the ones that arepotentially more inclined to push back,
to understandwhat are the kind of concerns they have.
(15:53):
What we've done, though, is
that we've tried to make sure that theAI is not in the way.
Right?
I mean, it's really here to assist them.
And, Ryan, you've mentioned the copilotthat is
allowing you to talk to your model.
You don't have to, right?
I mean, you can use all the modeling toolsand do all the kind of queries
and review yourselfthat the model is really here to try to
(16:16):
work as a copilot and helpthe engineers do that faster.
And as we look at automating the production of drawings that we do, as we look at,
maybe doing some
kind of collaborationwhere the design that is proposed,
there is a, there is a tool in OpenSite+to optimize your layout.
And that is a toolwhich we envision to be collaborative
(16:39):
in some sense that you proposesome options, And then two engineers
is speaking one option,and then the model learns from this.
And then throughsome kind of a reinforcement
learning technique, try to understandexactly what you have in mind,
which might not be somethingthat AI would calculate immediately,
but that is somehow carrying your savoirfaire, your know-how, into the design.
(17:01):
So, if AI is not in the way,if you can trust what you're doing,
we expect that pushbackthat we touched on here to be minimal.
But, I still think we need to confirmthat as we get more adoption.
So, I think we'll. Don't you agree?
I do. We'll park that for a futurepodcast, Julien.
We will bring you back
after we start to see some of the resultscoming in from that.
(17:22):
And you opened it up, and I appreciate youdoing that with the OpenSite+,
having seen some of the
presentationson that extremely powerful tool.
Right.
So, thank you To Bentleyfor pushing that forward.
Maybe for the Viewpoint listeners,maybe a high-level overview of
the OpenSite+ tool,what it is, what it can
(17:43):
do, and where do you see itgoing in the future?
Because I looked at it, I've been watchingsome of the roll outs of this,
And VHB is obviously,you know, looking at this
as one of those opportunities and
seeing what Bentley has doneand seeing the opportunities
is just amazing and game changing.
(18:05):
Yeah. So,
there's a lot
we're hoping to dowith that new family of products.
Right.
Because I think OpenSite+is the first of a family of products
that are digital twin native,meaning that instead of generating funds,
as we've done for the past,for the past decades,
you're starting from somethingthat provides you the context.
(18:28):
So, you don't start from scratch.
You start with the context of the map,the terrain.
And you probably heardthat we acquired Cesium recently.
So Cesium is bringingthat geospatial context,
to allow you to start with
the real conditions that are capturedfrom the digital twin of the world
and then from this, you layermore and more, additional layers of data,
(18:51):
so that the work that you start withis already
influenced by all of that contextcoming from the requirements
that are passed by AI,the geospatial data that we bring together
with the terrain, with the, the,the, the, the, the not to say that,
the zoning ordinancethat is provided by different can of tools
of GIS layersthat all of that are brought together
(19:14):
and then the engineersare able to start the work
much faster because all of that contextis provided here.
And I think that's something we're goingto see in lots of different products.
OpenSite+is the place we're starting from.
But when you think about it, when you'redesigning a bridge, when you're
designing a highway,when you're designing a railway
(19:35):
or even a power plant,you need that context always.
Right?
So that's what I think is truly inspiringis this vision
of starting with a digital twinof something that is yet to be,
bringing
as much data as possible in contextfor the engineers
to work in the most efficient way,
And then thinking the productwith an AI-first mindset.
(19:57):
But because it's very difficult,I think, to introduce AI in the,
in an existing product because there areso many things you would do
differentlyif you would have access to AI models,
if you would be able to talk to the data,well talk between quotes.
Right?
But I'm not always meaning this vocally,although
it's coming.
(20:18):
It's possible.
But this idea that your applicationcan understand the smart
objects and answer questions about those
and that you can automatea lot of the work by just.
I think a model understanding what is theintent that you're trying to have
and then execute a lot of those tasksare automated for you or speeding up.
So, I really envision that
(20:40):
as we are bringing
together the data that we have access tonow with our partnership with Google,
the data that Cesium already had accessto, bringing all of this into a
replica of the world and having a productthat is designed for that.
It uses AI as a wayto accelerate the work of engineers
from the first force,from first principles is going to create
(21:02):
a completely whole new family of productsfor engineers, which are going
to supercharge their capabilitiesto create the best possible designs.
And I mean about these sustainable,resilient designs,
but also being more productiveon addressing the backlog, which is huge,
and what that is a prime right nowis the challenge of shortage.
And deliver digital twins that will alsounlock a lot of potential outcomes,
(21:24):
all over the lifecycleof that infrastructure.
So, the part I
just I was so enthralledlooking at the OpenSite+
was the capability to go from pure concept
development to final design plans.
And then digging into it
and understanding how do you actually uselocal codes by, you know,
(21:45):
and you do it through your Copilot accessthat you're doing.
So actually, localizing it,which is one of the biggest challenges
when you think about rolling outan enterprise
across the entire globe of whatyou're doing, to go to optimization. And.
Right, and that's, that's key, right.
So, you know, as an engineer,and I am an engineer,
to be able to have the ability to do siteoptimization at your fingertips.
(22:07):
Simply amazing.
And that's where those younger engineers
or the engineers could sit thereand look and say, what
and why do I have these three or fouror five choices of site optimization?
What makes it better?
So, it's the pluses and minuses of it.
So yes, I'm a fan.
I'm all in.
Ryan, you know thatwe started talking about the OpenSite+.
So, how do you see from VHB's perspectivethe use of the OpenSite+,
(22:31):
you know, with what you're doingand driving,
the model-based designacross the footprint.
Sure.
So, you know, with our focus of buildingintelligent models,
you know, as being kind of our visiontowards the end of 2025.
That any project we start,will be started with the vision of it
being an intelligent model that we canextract data from throughout the process.
(22:53):
You know, the plans are just a byproduct.
One of the things that excites mewith with OpenSite+ is to think of
not losing datathroughout the evolution of a design.
So, so, Dave,you think about our process today.
Okay.
We've done our preliminary designin, in maybe we did it in concept station.
Maybe we did it, on a pieceof tissue paper or whatever we did it on.
(23:16):
We did that first concept.
Then you have that loss in datawhen you start to move it into real design
and then, oh, well,now you need to do traffic analysis
or you need to do drainage analysis.
Oh gosh.
There's another point where you'relosing data because you're extracting.
And even when we do our bestto connect those models together,
(23:38):
unless you're truly in one environment,you do lose data
through those steps in the process.
And there's reworkand potential for error.
So I think the thought of having that,that process just evolve and oh,
I now need to do my pipe sizingand having that just be a function
within OpenSite+ where all those toolsare accessing that model.
(24:01):
You know,I think, Julien, when I think of AI,
I think of data being
the fuelthat actually drives that AI engine. And
one of the struggles we have is
what is the quality of the datawe're putting into that AI engine.
So, having that dataall be kind of curated from the start,
(24:22):
Dave, in a platformI think is really, really powerful versus
this dispersed data that, you know,we fight the battle every day in the AEC
industry of is that the right sourcefor that information?
You know, that single source of truth,I think is really powerful for teams.
I don't know, Julien.
Do you have any thoughts? Yeah.
(24:43):
Well, I was about to ask you a question,which is I mean,
I think that continuum of datain one environment,
and then making this available to AI.
So, we've talked about AI relievingthe engineers from mundane tasks
like creating the drawings, et cetera,but what is the kind of next
big milestone
that you see for the engineersthat are working on those site designs,
(25:04):
with all that data available to AI,what are the kind of big
accelerators that you believeshould be implemented next?
You know, in OpenSite+?
Well, around
OpenSite+ I am excited to see it injust started to see
some of the drainage componentsrecently added into that.
And I think that that disconnectthat, that
(25:24):
when we talk about sustainabilityand we talk about impact
to our environment, it'snot just about our,
our site layout,it's about considering things like,
"okay, what is my embodied carbonin this design?
What am Iwhat can I do to reduce my footprint?"
You know, I have children, Julien,that are run around the house,
(25:46):
and I want to leave the world in a betterplace for them than where I got it.
And part of
that is when we approach our design,thinking about our design holistically.
And I think for Bentley to build in tools
where sustainabilitycan directly be measured as our evolve,
our design evolves, is importantin that time
(26:08):
that we're gaining that 30 to 50 percent,whatever that number ends up being,
shouldn't be timethat we don't spend on our design.
It should be time that we spendon improving our design.
And to do that, what we need to dois have those measures presented to us.
Like at any point during my design,I want to be able to go into it
and say, okay,
(26:30):
you know what, let's change my pavementtype to all be impervious pavement.
What does that do, Not only to my cost,but what does that do to my,
you know, sustainability index or whateveryou want to call that.
So, that excites mefrom really getting our teams to
have the time to produce the best designthat we can,
(26:50):
not just produce the first designthat meets our requirements.
I'm really happy to hear you say that,because that's,
I think, where AI can bethe most helpful because,
one of the things I likeabout the portfolio of tools we have
is that we have a lot of tools that arevery good at doing some specific things.
(27:12):
But it's difficult for an engineer to runall of those tools, right?
I mean, doing a simulation for flooding,doing a simulation
for electricity, doing a simulation
for mobility of pedestriansor cars and things like that.
I mean, we have all of those tools,but for an engineer, it's
still a lot of time to goand do all the simulations to make sure
they have the right informationto take the right decision.
(27:34):
Now, AI is, potentially
a concern for some because they say,oh, AI might be creating something
that is a bit of a destinationand that could be wrong for us.
But if you use AI not to kind ofcreate content out of the blue,
but instead run all of those simulationengines for you in the background
with your designs,and then provide the feedback
(27:56):
and tell you, well, by changing thisand this and that, we get a better score
on Drainage for flooding or by doing thisand this change, we construct
20 percent faster and we use 20 percentless concrete, for instance.
Well, then you have you have all the toolsto make the right decision
to achieve those sustainabilityresilience goals that you have.
(28:17):
So, I fully agree.
If you have a client who's interested in
how do I capture solar energy on this siteas I redevelop it,
that's information that using thatAI model you can, or the AI assistant,
you can quickly go through and providea few iterative design alternatives where
how much do we weighcapturing solar energy back versus
(28:40):
how much do we balance that versus costof the overall development of the site.
So, I think it's interesting.
It has so many differentareas for potential gains.
You know, a question that I havefor you around that,
you know, it's it's very clearthat that that digital twin
is how Bentley is looking at leveraging
AI so that iTwin environment.
(29:04):
You know, how do you see thatas it goes outside of design like,
and goes to either constructionor into maintenance of operations?
Is Bentley positioning other tools to,
to access that same iTwin framework, and
you know, AI to leverage that data?
Well, our vision is that,
(29:26):
and this builds on what you wereexplaining before, there is a lot of data
loss and all of ours are happeningbetween one stakeholder to another.
And I think a lot of engineering
firms are looking at expandingtheir portfolio of services,
not only to design, but thento construction, and then to operations.
And that is going to address some of this.
(29:46):
But there is still quite a bit of,and all of ours happening
between different stakeholders right now,
including the final ownerand operator of the infrastructure.
And we kind of have a responsibility
to make sure that there is aslittle data loss as possible.
And that's where we
this concept of bringing the data togetherin a common data and augment
(30:06):
or a digital twin of the infrastructurethat keeps evolving together
with the infrastructure itselfis so important to us because instead
of having handovers of papersthat are then remodeled by someone else,
and there is a lot of data loss there,We believe that, well, what we can
handover is actually access to the datawith an audit trail,
(30:28):
a traceability back to the originso that we can
then start to unlock a lot of the outcomesthat are becoming possible.
If AI has access to the whole datalineage,
of the project,
unlocking new possibilitiesin terms of the operation
phase or the construction phase,or even like when you need
(30:50):
to do a renewal
of the infrastructure, I think access toall of that information is precious.
And, we still not there yet,
but I believe that making surethe data is digitalized, that it's
well structured from day one, so it canthen be handed over to different people
keeping that lineageis going to be the breakthrough that
we need to be ableto expand the value of AI
(31:12):
and those digital twins into the completelifecycle of the infrastructure.
And that that is going to also supportso much of those resiliency
and sustainability goalsthat you mentioned before.
So, Dave, I want you to add to that one.
You know, as, as we or VHB
are focusing on things like AIand model-based design, you know, we've
(31:36):
heard Julien mentioned data several timesand, you know, data format and data loss.
How do you, as someone who's dealingnationally with data standards,
how do you see national data standardscoming into play with AI?
Well, Ryan,
before we get to the national standards,because I knew that was going to come up.
(31:57):
You did ask a question.
And this is, you know, Julien,it is an interesting when you're talking
about the digital twin technologyand the virtual representation
of the real worldof what we operate in and live in.
Right?
I think initiallysome of the percent efficiency
that you're referring to, the 30,the 50% become even more profound
(32:18):
when you start to combine those tools,those modules, into one full digital twin.
So, when you think aboutintegrating the stormwater
and with the design and withthe transportation, does a couple things.
It actually eliminates a lot of the errorsby not communicating those tools, right?
So, as the engineers are looking atmarginalizing risk,
(32:40):
that is one area that you'remarginalizing risk by doing that.
And I think that's great to actually bringthem into one centralized environment.
And then, you know, the second part is
when you think about clash detectionor risk management, right?
Having the ability to say, okay,we're going to do this
design and is against whatRyan wants to do with the sustainability,
but it triggers something else
on a different area or a different modulethat's phenomenal, right?
(33:03):
In terms of what it does.
But the models, Ryan, like you say,are only as good as the data goes into it.
And there are not a lot of good, there
not a lot of standards,across the US based on this.
Right. So, what what are we trying to do?
You know, at VHB, we're trying to come upwith our own governance and standard
(33:24):
to put into it,but it's more of an industry challenge.
Right. Of where it is. Right.
So, what is Bentley doing also to help
with standardizingthe data goes into the models
so that it is transferable from agencyto agency, state to state,
on what it isor, you know, is that an issue?
(33:48):
It is, it
is an issue for multiple reasons, right?
The first one is first agreeing onwhat do we handover
from one party to anotherto make sure that the data
is transferred
without loss of fidelity,without loss of information.
(34:08):
And, that I think has been addressedin some sense with
quite a few efforts on creating standards.
Building smart,for instance with IFC has been helping
in defining standardsfor data exchange between stakeholders.
But we believe we need to goone step further
in the sense that,
(34:29):
as engineering firmsare creating that data
in the tools that they use,
we also need to make surethat they build all of that
data structure,that digital twin, in some sense,
in a way that also gives them the safety
that they are not dependingon one single vendor for that.
(34:50):
Right? So, one of the things
we've
talked about recently a lot in our yearin infrastructure is that with Cesium
and the iTwin platform and bringing themtogether in one common platform.
We also try to have an open approachin everything we do,
so that not only are we talking aboutthe standard for the industry by
(35:10):
how do we exchange datawhen it comes about, design files, etc.
but how do we make sure that all the datais collected in an infrastructure
project, which is not only about models,it's also about the IoT data,
the GIS data,
the reality meshes, the surveys, etc.
How do we make sure that all of that data
that you gather in one placeso that you can leverage it with AI,
(35:33):
or that you federate from differentsystems into one common data construct?
How do you make sure that this data
is going to livefor as long as the infrastructure itself,
that is described using openstandards that everybody can understand.
So, one of the open standards we've talkeda lot recently is the 3D Tiles,
which is an open, an OGC, Open Geospatial
(35:54):
Consortium standards
that is making it possiblefor everyone to read that data.
But there's many more like the featureservice, from OGC,
which is exposingGIS data and interesting,
interesting data being surfaced
on the 3D visualization canvas.
(36:16):
But also, how do you store the BIM dataabout your model?
So, we're looking at creating standardsfor all of those
that are going to facilitatethe interoperability of the data
between vendors,
but also making surethat you can take your data out
and move it to another platformif you want to,
and then continueto work in this with this.
Because if you try to bring all of thatdata in one common environment
(36:39):
as we've described before, but all thatdata is in some proprietary format,
then you're locked up with a single vendor
and the interoperabilityis going to be complicated.
And as you want to use that datain ten years from now has
you need to do some workon this infrastructure project,
or if that tool has disappeared, orthe vendor is not servicing you anymore,
then that datathat all of that value is lost, right?
(37:02):
So, the standards are important onboth fronts.
How do your national standardsfacilitate the exchange
between parties, but also for youas you make that investment in your data?
How does the standards helpyou ensure that this is open
and that you will be ableto leverage that data in the long run?
What's your view on that?
Ryan,
I turn to you on this and say,what's your perspective?
(37:24):
You and I had some conversationaround IFC. Right.
And where the standards are.
For me, Julien,
there has not been a common
standard across,and we operate within different platforms.
Right.
So, we operate in multiple states,which creates a challenge
in terms of the uniformity of the datathat comes out. Right.
(37:45):
So that creates an issue for us,you know, as a firm
that operates in multiple states.
To basically try and come upwith the national standards
that you're referring to that operatewithin
any platform, the full interoperability
is going to take a Herculean effortthat's not there today.
(38:06):
Right.
But it is actually under the focusesas we all talk.
You can't do good AI, you can't dogood business models, you can't do
good digital twins unless you havesome standardized format working with,
you know, agencies and vendorslike yourself.
So, this is one of the biggest pushesthat VHB is actually making.
We have differentareas that we're talking to,
(38:30):
agencies, partners,
policymakers, relative tohow do we actually create these standards?
What's what does it meanfor the governance side of it?
And then what do we do within our own datathat we control,
but also some of the datathat we're actually working with
state and local agenciesrelative to how they process their data
(38:50):
as part of their datalake and repositories.
Well, Dave, to build on that,I think it's exciting
that we're actually startingto see some momentum around
that with different DOTsacross the US right now
looking at what is their new digitaldelivery standard going to be?
You know,
(39:11):
we're working with a number of clientshere on the East Coast where that is
a specific project that they have.
It's not arounda particular design project,
but it's aroundwhat is our digital data standard.
And I think thatwill help in having AASHTO,
you know, Julien being in Europe,
(39:33):
your standards are a bit more evolved
than the US standards as of todayaround horizontal construction,
but having AASHTO actually start
to release this is some levelof development guidelines.
This is what our data standardfor the industry we recommend.
I think those are really powerfuland helping us
(39:56):
have a target
so that we can ensure that we havehigh quality data in these environments.
Like to me,
level of developmentneeds to be a discussion on every project
so that the expectation is set onwhat is the quality of that data.
You know, Julien,I can bring in data from any partner
(40:18):
but understanding what the intended
quality of that data is is really criticalas we're doing design work.
Absolutely.
Yeah.
So, I agreeAASHTO has as a big role to play here.
I think that we see the marketevolving in that direction, right?
I mean, we've seen multiple DOTswork on kind of research projects.
(40:41):
There's been some public funding aroundthis, how to create the digital
delivery models of tomorrowto be able to manage the tracking
of progress, to manage the deliverablesas a digital model.
So, I think there's a lot of,there's a lot of activity
that we're startingto see that in that direction.
It would be a lot about the right rhythm
(41:03):
in adapting to that moving landscape.
And in the meanwhile,we need to also make sure we provide
all the foundationsfor that technology to exist. So,
that's all we have.
We have engaged in multiple openstandards bodies or Bentley
is part of the Khronos, an institute
(41:23):
for the standardization of gITF,or 3D geometries.
We are part of OGCas an important member right now
to discuss all of those standards aroundhow do you represent today's data?
How do you represent 3D tiles?
We also talking with Metaverse
Forum and the Alliancefor OpenUSD to discuss how do we,
(41:44):
how do we make sure that we have openstandards to represent that data across
vendors in an open way so that as thosenational standards are evolving,
you always have the warranty that you havedata in the format that you understand.
That is the standard that is open.
And that you canthen massage and transform
into those national standardsas they start to emerge.
(42:06):
I have one broader question.
Put the crystal ball out there,
5 to 10 years. Crystal ball, right?
5 to 10 years.
Where do we see the GenAIin the design services in 5 to 10 years
from now?
(42:27):
So, I think that GenAI is goingto be connected
to more and more toolsthat are not artificial intelligence,
but actually human intelligencethat has been ingrained
in softwareto explore more and more alternatives.
So, I think right nowwe're seeing tools like OpenSite+
(42:50):
and others in the marketuse techniques like large language models
to find information faster, maybe to talkto small objects in some sense.
And I really envision that those toolsare going to become smarter and smarter
by being informedby many other kind of data sources
and algorithms that they can then employ
(43:12):
to become the best possible assistantfor the engineers.
So, we talked about embodied carbon,for instance,
being able to look at the designas it's being designed by the engineers
and then providing the carbon footprintis something
that AI assistants should be able to doin the coming year or two.
And that will already have massive impacton the quality of the designs
(43:33):
that are being produced,
but then bringing all the perspectivearound structural analysis,
mobility simulations,but also informing that mobility
simulation with datathat comes from the world.
Right?
I mean, Google has a lot of informationabout mobility.
You have millions of peoplerunning on the road every day with Waze
(43:53):
and there is a lot of datathat is being collected by dash cams
informing the design decisions,With all of that crowdsource
data is also somethingthat is going to unlock massive potential.
So, in 5 to 10 years, I'm envisioninga world where the engineering tools
that you're using are not just a toolthat is helping you build 3D shapes
(44:14):
and to put some metadataaround those structures,
but actually see a complete what-ifscenario, right?
I mean, here are the traffic flowson that highway for the past five years,
the stormsthat have been flooding that region
and how those would potentially impactyour design right now
and how you couldpotentially make some changes
(44:36):
to adapt to this,so that your design is more resilient.
And I think engineers are goingI mean, it's going to be very exciting
times to be an engineer,because having all of that context
and becoming much smarterabout your designs is actually going
to be very fulfilling, I believe, for,the engineers at the entering firms
because you will have so much contextavailable
(44:56):
at the at your fingertips,
to create so much
of better designs, thatI think is going to be truly exciting.
So, 5 to 10 years, I really see thatconvergence of data and tools,
elevating the engineers
to capabilities that were not thoughtto be possible before.
(45:18):
You know, Julien, how exciting will it bewhen we do get to that world of,
okay, we've had our designed digital twinthat went to construction and then,
you know,you had mentioned visual learning models
and tools like Blyncsythat you have out there today,
you know, okay, these dash camsnow are providing feedback to the owners
around, hey, you know what?
(45:38):
There's a maintenance issueyou need to address over here,
whether it's a drainage issueor whether it's,
a safety issuewith a guardrail missing or whatever.
But, you know, you think about those typesof tools really
being able to inform inand advance our design process.
Prioritizingsomething from, hey, AI noticed
(45:58):
there's a high number of near missesat this intersection.
Maybe this should be a priorityaround construction
for pedestrian safety or whatever.
Like that to me is,
to the crystalball, Dave, that's not sitting here today
and we're doing it on every projectbut five years, ten years down the road.
(46:19):
I think with good data,good digital twins,
and using the tools as they evolve,we really have a chance to like
make a significant differencearound safety and
around the experiencethat your traveler has.
You know, self-driving cars,If all of a sudden they're making us aware
of problems in the road through AI,
(46:41):
fantastic to get to that point.
So, Ryan, I think that AI in the work
that Bentley and Julien is doingis actually the great equalizer.
We have a lot of firms out therethat are smaller
firms that do not have a lotof the capabilities of the larger firms.
And I think the capabilitiesthat Bentley brings through that and other
(47:03):
AI applications actually,is that great equalizer that can help
normalize, I think, the industryand move the entire industry forward.
Popular Podcasts
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Therapy Gecko
An unlicensed lizard psychologist travels the universe talking to strangers about absolutely nothing. TO CALL THE GECKO: follow me on https://www.twitch.tv/lyleforever to get a notification for when I am taking calls. I am usually live Mondays, Wednesdays, and Fridays but lately a lot of other times too. I am a gecko.