How many ways can a project be built?

Millions
Zoning X Sequencing X Resources X Equipment =
of ways

These options would take a human

being over 13.5 years to explore if
they worked full-time with perfect
efficiency.

ALICE explores these millions of

ways to build in minutes and
identifies the optimal paths, with
trade-offs between time and cost.
ALICE does the busy work for you

Super

Sub Owner

Scheduler

Architect Engineer
PM

Planning Scheduling Management

Rule set Constraint resolution Delay mitigation
Scheduling Sequence variation Project acceleration
parameters Optimization and exploration Design changes
Communicate your
schedule visually
ALICE helps evolve the way you build
Customers are seeing an average of:

16.2% 14.7%
project duration reduction labor cost reduction

“The integration of our building expertise and ALICE enables

us to evaluate hundreds of potential scenarios to determine
the best possible plan for our customer”

- Ricardo Khan, Sr. Director of Innovation

ALICE shaved 84 days from one project’s original 540 day schedule.
Getting the ROI
out of AI
Robert K. Otani
P.E., LEED AP
Chief Technology Officer
Thornton Tomasetti
Andy Leek
AIA, LEED AP BD+C
Vice President - Technology & Innovation
PARIC Corporation
Iro Armeni
Ph.D. Candidate
Stanford University
 AI and Design
Leveraging Domain Knowledge Via
Data Science

Rob Otani, PE
Chief Technology Officer
Thornton Tomasetti Inc.
TT Intro Our Locations
5 Continents
Albuquerque, New Mexico
1200 Engineers, Projects in
Austin, Texas
Boston, Massachusetts
Chicago, Illinois
Dallas, Texas
Architects and
Professionals
36 Offices
54
Countries
Dalgety Bay, Scotland*
Glasgow, Scotland
London, UK (Head Office)
Denver, Colorado
London, UK (PLC City Office)
Ft. Lauderdale, Florida
Moscow, Russia
Irvine, California
Beijing, China
Kansas City, Missouri
Hong Kong, China
Los Angeles, California
Shanghai, China
Marina del Rey, California
Mumbai, India
Mountain Lakes, New Jersey
UAE
Mountain View, California
Ho Chi Minh City, Vietnam
New York, New York (2)
Christchurch, New Zealand
Newark, New Jersey
Philadelphia, Pennsylvania
Portland, Maine
San Diego, California
San Francisco, California
Seattle, Washington Weidlinger locations
Washington, D.C. Shared Locations
West Hartford, Connecticut Thornton Tomasetti Locations
São Paulo, Brazil *Additional offices in Bristol, U.K.
Practices We i d l i n g e r
Fo r e n s i c s
Tr a n s p o r t at i o n

We i d l i n g e r
Sustainability Applied Science

Fa ç a d e CORE Re n e w a l
Engineerin studio
g

We i d l i n g e r
P r o t e c t i ve D e s i g n Proper ty
Loss
Consulting
Constr uction Str uctural
Engineering Engineering
What is AI?

Geospacial World: “What is Artificial Intelligence,

Machine Learning and Deep Learning” By Meenal
Dhande
Why use AI?
 In designing structures, there are many mundane and repetitive tasks that
engineers are “overqualified” for. AI can automate those tasks. This allows
engineers to spend their time being creative and solving problems.
 AI can also automate QA/QC since it can find “outlier” patterns.
 AI can process millions of data points and therefore can understand patterns
in engineering decision.
 AI can assist large firms by capturing institutional knowledge from the
“experts” in firm and allow that knowledge to be at the fingertips of the
younger engineers.
AI at Thornton Tomasetti
2015 Thornton Tomasetti
published this research paper
outlining the use of Machine
Learning for the design of
buildings at the 2015 International
Shell and Spatial Structures
symposium.
AI at Thornton Tomasetti
 AI and Automation

ASTERISK
“Asterisk” is a software application developed in CORE
studio | Thornton Tomasetti as an ongoing research
initiative to leverage AI/Machine Learning.
 Asterisk
“Asterisk” is an application developed at Thornton Tomasetti that
performs a structural design of a building in seconds.

The application uses Thornton Tomasetti past building design data and
machine learning algorithms to “choose” structural member sizing for
floor framing, columns, slabs, foundations, and core walls for both steel
and concrete buildings. Only a building massing is necessary.

This speed of design and modeling is not possible otherwise. A single

iteration would take a team of engineers a minimum of a week to
provide the same level of design.
User Journey

 Their Process

Switch Attribute Visualization

User Journey

 Their Process

Properties Panel
Comparison Page
User Journey

 How Asterisk can Help 02

Compare Iterations
 Reports
---

 How Asterisk can Help 01

Iteration Cut Sheet Massing Comparisons Project Comparisons

 AI and Automation

DAMAGE DETECTOR
“Damage Detector” is an application that processes
building structural damage to be found and documented
by using image recognition AI.
Damage Detector
What damage can AI detect?

Anything a human inspector can find by visual inspection!

 Cracks, spalls, exposed rebar, (concrete/masonry)
 Moisture ingression, mold, (stucco)
 Corrosion, cracking, buckling (steel)
 Rot (wood) etc. etc.
Damage Detector
How does it work?

• Supervised learning of
deep conv. neural nets on
many labeled images
• Leverages pre-trained
models for faster training
Damage Detector

How does it work?

• Using just images, types

of damage can be found
using AI
Damage Detector
How does it work?

• We will be able to
inspect structures
faster, more frequently,
and in areas where
people may be
otherwise at risk
Damage Detector
Future Plans

• Integrate with 3D photogrammetry (via

drone-captured imagery) to build a 3D
digital twin of inspected structure
including damage zones

Image courtesy: geocue.com

In Summary
 20-30% of current engineering workflows can be automated using AI
 AI can increase productivity by automating engineering tasks
 AI can learn to assist checking our work to find errors
 AI can find trends and learn from large data (engineering/financial) sets
quickly that can assist engineers to manage projects more efficiently
and make key decisions faster
 AI can process domain knowledge of engineering firms and enable
knowledge capture from generations of engineers
ROI out of AI?
Where to start? When to start?

Andy Leek
VP Technology & Innovation
PARIC Corp.
 ENR Top 200 Nationally
Established in 1979
$635 MM in project revenue for 2017
Services
General Contractor, Construction Manager, Design/Build, Self-Perform

St. Louis Based

+ Kansas City, Des Moines, Cincinnati
Where / When to Start?
BIM Controls, Prefab, Metrics, Reports Insights
What is AI? www.delltechnologies.com

Definition: Artificial intelligence (AI)

also machine intelligence (MI) is
intelligence demonstrated by machines,
in contrast to the natural intelligence (NI)
displayed by humans and other animals.
(Wikipedia)
We have options…
Design and Sequence
We have options…
Design and Budgeting
We have options…
Design and Coordination

 Generative design leveraging BIM and point cloud data to route

systems while avoiding collisions
We have options…
Scheduling, Production - Integration too!
Estimating / Budgeting
 BIM-based QTO and WBS alignment
 Design Team Agnostic
 Machine Learning
 Version Tracking
 Change Management
 Common blindspots
 Audit Trail
 Historical Cost Analysis
 Trends
 Projections
Decisions driven by DATA
Recap / Survey Data
Find and Connect stuff…
Changing the industry’s
AI mindset

Iro Armeni | Stanford University

 Industry is open to technological changes

 There have not been productivity increases in 40 years

• The AEC industry has not had the same gains as other industries

 It is becoming increasingly mature and open to technological

advancements

 The Artificial Intelligence trend is not an exception

© Iro Armeni 2018 40

 The ‘AI solver’ phenomenon

 There is not much understanding/education on what AI can do and

how it works

 Usually referred to as the ‘AI solver’ and treated as a black box

 AI can be a very powerful tool if treated/used correctly

© Iro Armeni 2018 41

 Current issues in AI adoption

 Witch-hunt: do not know what they want to do (task, end goal, etc.)

 Tendency to hoard data

• Start thinking about how to really unlock the potential of AI

 No holistic understanding of a topic within and across companies

• Fragmentation on building projects
• No processes/benchmarks - hard to determine where to apply AI
to change the value proposition or efficiency of a process

© Iro Armeni 2018 42

Why is tech not a bigger part of the industry?

 Unknown risk factor (need for safety/managing mechanisms)

 Data privacy

 Data sharing

 Fragmented approaches (no larger frameworks, goals and

structures)
• Fragmented resource allocation within companies
• Across many projects and buildings
• Within a building as well (there is not one company)
• How is innovation supposed to happen? 43
 AI-fy existing processes or radicalize them?

 Just because a process exists it does not mean it is the most

efficient

 Why digitize it and not find new ways to address current issues?

 We don’t ask the questions and not in the right way

© Iro Armeni 2018 44

 What to do next?

Think of things you would like to do that you:

 cannot do well

 cannot do at all

 need to do a lot more and it is not feasible due to expertise or

budget

© Iro Armeni 2018 45

 Will AI address questions

or
Will AI change them?

© Iro Armeni 2018 46

 Contact Us
Robert Otani
ROTANI@THORNTONTOMASETTI.COM
Andy Leek
AGLEEK@PARIC.COM
Iro Armeni
IARMENI@STANFORD.EDU
Get started with
a low risk trial
Try ALICE on a single
project and see how AI
can make your job easier
while saving you time and
money.
 Thank you!

Visit us at
www.alicetechnologies.com

Or contact us for more info

info@alicetechnologies.com