Experts Guide to Getting

Started with Industrial AI

The potential impact of Industrial AI cannot be
understated.

According to Accenture, by leveraging AI, the

manufacturing sector stands to increase profits
by 39% by 2035. For the oil and gas sector, the
World Economic Forum estimates that AI can
create $1 Trillion of value for the industry. However,
in the race to capture this opportunity, industrial
companies have instead experienced a high failure
rate when it comes to extracting value from their
AI investments. According to Gartner, by 2022,
85% of AI projects will fail, leaving benefits such as
enhanced operational efficiency, improved quality,
reduced downtime, and lower carbon emissions out
of reach for many in the industrial sector.

“Enabling operational agility with

a digitally connected workforce
provides a competitive advantage and
is now more critical than ever,”
said Matthew Littlefield,
Principal Analyst at LNS Research.

Start your AI journey on the right footing

by using this Expert’s Guide to Getting
Started with AI.
Assess Your Data Readiness
AI is extremely data-dependant. It takes operational data, Example:
analyzes, and learns from it to drive actionable insights.
Simply, data is instrumental in your AI success.
AI Data Requirements
When determining if your data is ready for AI, assess AI use case: Optimizing
the boiler fleet’s natural
the following:
gas consumption based
on thermal efficiency
• Relevant data: One myth about AI and Machine
Learning (ML) is that large quantities of data is all you Data: 18 months of time-
need. It is a misconception that ML can extract insights series data that included
from any data. Instead, it requires relevant data on the features that represented:
operational challenges that you seek to extract insights
for. For example, if your goal is to improve productivity, – The performance of
the data needs to represent everyday operations. While each asset to predict
fuel usage and cost
it is true that you need a large volume of data, if the
data is not relevant for that particular use case, then the – The variables that
entire exercise is futile. affected steam
generation (e.g., steam
• Large volumes: AI algorithms depend on large volumes flow)
of data, including historical and real-time data, to
identify accurate patterns and behaviours. The data – The environmental
variables that affected
should be representative of the process or asset for you
thermal efficiency (e.g.,
to gain meaningful insights and the desired result. ambient temperature)

• Data is multifaceted: It is crucial to ensure that the data

reflects your typical operations so that ML can observe
meaningful patterns from it and detect anomalies. For “AI can only work with
example, this includes ensuring that you have data what you feed them, and
that represents failures, which is required for predictive again it’s technology; it’s
and preventative maintenance. not magic. So, having an
excellent data set that fits
the purpose, the problem
you’re trying to solve is
essential.”

For successful AI implementation, data needs to be: - Lindsay Dempsey,

Senior Lead Optimization
Engineer, Enmax Energy
Relevant Large volume Multifaceted

3
Experts Guide to Getting Started with Industrial AI

Identify the Right AI Use Case

Identifying the right use case sets the stage for the Examples
success of your AI rollout. It helps you demonstrate
viability, obtain buy-in, and creates favourable conditions
of common Industrial
for scaling. Without a clearly defined purpose and KPIs AI use cases
along with the timelines, the initiative is unlikely to reap – Optimizing Assets and
benefits and will jeopardize any future trust or investment Processes to Improve
in AI. Product Yield

What’s more, AI is not suitable for everything. AI works – Reducing Water and
best for processes and assets that are repeatable Energy Consumption to
Improve Sustainability
and have historical data, where models can be
developed and validated. For example, AI can be used – Optimizing Energy to
to predict failures in a process that keeps happening Cut Rising Costs
repeatedly. But would have been unlikely to accurately
predict the sales surge of toilet paper that occurred at the – Continuous Process
Optimization to
beginning of the COVID-19 pandemic, due to the
Improve Product
unavailability of historical data. Quality

A successful use case comprises of the following features: – Reducing Product

Variability to Improve
• Demonstrates Clear Value Yield
The use case must be able to show value to all
stakeholders involved. You can do this by choosing a – Predictive Maintenance
process that incurs high operational expenditure as to Cut Unplanned
Downtime
it will provide ample opportunities to cut costs, build
efficiencies, reduce downtime, or improve quality. – Optimizing Process to
Lower Raw Material
• Has Enough of the Right Data to Work with Wastage
While selecting a process or equipment to apply AI,
it is best to go for a function that already measures – Optimizing Cycle Time
and collects a lot of data. This is because you need to Improve Batch
Consistency
the historical data to train the models before they are
operationalized. It is important to note here that data – Predicting Asset Failure
quality is equally important, not just the quantity. to Extend the Life of Oil
and Gas & Geothermal
Wells

4
• Augments Your Existing Processes
AI works best when the intention is to augment an
existing process or asset – and not used to rip and
replace one. AI uncovers data-driven insights that
incrementally and continuously optimize a process,
improve energy efficiency, and increase OEE. In addition,
when AI is used to augment an existing process, the
institutionalized knowledge that is leveraged also means
that the operational risks are reduced as well as any
disruption to your workforce or revenues.

• Define KPIs to Secure Management Buy-In

In order to secure management buy-in, it is crucial
to demonstrate tangible results through the selected
use case. To manage expectations properly, start by
defining a goal or a measurable outcome that the use
case will contribute towards. By clearly articulating a
problem, and having the AI predictions to mitigate these
problems, you can measure the ROI and justify your AI
initiative.

• Assess its Scalability to Other Processes

The golden rule of an effective AI adoption is to start
small, gain success, get buy-in, and then scale it to
other processes. Hence, the use case you select should
have the potential to extend the framework to other
processes and functions across the organization. This
way, the learnings of one use case can be applied to
another function.

“When applied to the right types of problems and business

issues, AI can demonstrate great value. By addressing
those problems, CIOs can inject confidence in senior
management for AI technologies, and secure funding for
future initiatives.” - Gartner

5
Experts Guide to Getting Started with Industrial AI

Address The People Factor

Successful AI rollouts require the buy-in of all stakeholders
and not just the data scientists or engineers championing
the initiative. This means that the entire workforce needs
to embrace this change from the operations floor to mid-
level managers and top management. Failure to do this
has been a key contributor to why 85% of AI projects have
failed, according to Gartner.

When investing in technology, such as Industrial AI,

success will not come just by investing in the infrastructure
and hiring the technical expertise. Instead, the asset or
process Subject Matter Experts must be at the forefront
of the deployment if AI is to succeed beyond the pilot
stage. The domain experts (i.e., the industrial engineers
or operations teams) are in the best position to identify
where AI will provide meaningful impact and then they can
evaluate how they scale it across the business. Where AI
projects can fail is when organizations look for short-term
benefits and do not commit the organizational effort and
reskilling learning curve required.

“The most important thing is that you need to put people at the center of your
AI-powered transformation. In the end, it is the people that will make decisions
based on that data and take action. Data without insights is irrelevant, and
insights without somebody acting on them are useless, which is why we believe
people are at the center of driving transformations with AI at scale, as you create
a high-performance culture and become a data-driven company.”

- Sachin Lulla, Global Digital Strategy and

Transformation Leader, Ernst and Young.

6
Select an AI Platform Built for
Industrial Environments
The AI market is growing at lightning speed. However, AI
platforms come in different shapes and sizes that may not “Just like many other
suit the industrial environment or the right power users. technologies in the
For example, some AI platforms are built for cross-vertical manufacturing sector,
purposes, such as health, finance, or retail. While these only AI platforms that
are built specifically for
platforms may provide a horizontal offering, industrial
Industrial environments
companies need the depth of industrial-specific expertise that will meet our
can manage the volume, variety and velocity of industrial data, requirements. Industrial
and the variability that is typical in industrial environments. AI platforms are a must
for manufacturers who
What’s more, some AI platforms are built for data scientists. want to achieve the
However, the demand for data scientists globally far impacts they seek.”
outstrips supply, and not all industrial companies may want - Kazuyoshi Watanabe,
to invest in teams of data scientists to fuel their AI strategy. Manufacturing
By taking this approach, industrials also risk leaving the SMEs Technology Division,
Digital Engineering
on the sidelines during the digital transformation of their
Group,
operations. Yamaha Motor
Company
Industrial companies should consider an Industrial AI platform
that is built to support industrial operational environments
and empowers industrial operators. In particular, industrial
companies should consider the following capabilities:

Industrial AI No code platforms

built for industrial processes and removes the need for
assets and volume and variety of coding expertise
data sources

Pre-built Industrial AI Solutions Cloud-based infrastructure

to accelerate the industrial to support scaling AI
sector’s AI RoI and across multiple processes,
time-to-impact operations, and facilities

7
Experts Guide to Getting Started with Industrial AI

About Canvass AI
Canvass AI provides a no-code AI platform that puts the benefits of AI in the hands of
industrial engineers. With Canvass, industrial engineers extract value from their data in
days not months – without requiring coding expertise. Canvass AI is purpose built to
address industrial operational challenges across their processes and assets, enabling
them to reduce carbon emissions, lower costs, and optimize their energy consumption.

Project
Data Ingestion
Operationalization
Use Case Building Deploy, Monitor &
Data Collection Systems & Evaluation Operationalize
OSIsoft PI Dashboard
Integrations
Data Sources Use Case Templates Augment expertise and
Sensors control your operations
Control Systems with AI-driven
Lab & Maintenance predictions.
Etc.
• Power BI
3rd Party Visualization • Pi Vision
& Curation Tools • Custom Dashboards
OSIsoft/Aveva Data Analysis & Train & Evaluate
Azure Transformation Models

UNDERSTAND GAIN INSIGHTS INTO CREATE SCALE AI ACROSS

YOUR DATA - YOUR OPERATIONS - IMPACT - YOUR ORGANIZATION

Within one platform, Canvass AI provides:

• Strong ROI: Identify and select AI use • Pre-built templates for industrial
cases with strong ROI solutions: to fast track how users
build and train AI models for specific
• Powerful data visualizations: to outcomes without needing coding
help users immediately discover key experience
insights into their processes and
assets • Explainable AI capabilities: for
deeper model evaluation and
• Real-time predictions: deploy, explainability so that users can build
operationalize, and monitor AI models the best performing models faster
direct into operations

8
Across the globe, Fortune 5000 industrial companies in the manufacturing and oil and
gas industries are implementing Canvass Al to improve key business metrics, such as
energy costs, production yield and greenhouse gas emissions, 12x faster than other
platforms in the market without the need for deep technical skills or expensive and
lengthy consulting projects.

For more information about Canvass and to get started on implementing AI into your
operations, please contact us: Canvass AI | info@canvass.io | www.canvass.io