your definitive guide to ai in corporate learning
TRANSCRIPT
1
YOUR DEFINITIVE GUIDE TO AI IN CORPORATE
LEARNINGUnderstand AI, Set Realistic Goals, and Achieve
High Performance in Business and Learning
WHITE PAPER
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
2
Foreword
We are living in the Information Age, an era in which knowledge is the most valuable resource for organizations
and individuals alike. Buckminster Fuller introduced the Knowledge Doubling Curve in 1982, displaying the
combined "knowledge" of mankind doubling in 500 years from the 14th century to the 19th century, and later
increasing by a factor of 16 by the year 2000. In 2017, it was estimated that all human knowledge doubles every
13 months. By 2020, it will double twice a day. 1
This amount of knowledge opens the doors to many things we wouldn’t have believed were possible 10 years
ago. Similar to the first three revolutions, the Fourth Industrial Revolution will have a major impact on the way
we live and work. The internet is now the most powerful tool on the planet, and there is little room to keep up
with the pace of the world without it anymore.
For a majority of Millennials and Gen Z, the internet has always been a part of everyday life. They are used to
keeping up with evolving technology, because it feels normal. Nonetheless, they need to take the time to learn
the newest developments just like older generations.
Technological change is occurring faster than ever before, and organizations also have to keep up in order to
remain competitive. New devices, more complex software, communication — everything is becoming faster
and therefore can at times seem more demanding to learn and implement. In order to be able to fulfill the duties
of their jobs, employees need to continuously learn new skills during their careers.
The World Economic Forum predicts that “by 2020, more than a third of the desired core skill sets of most
occupations will be comprised of skills that are not yet considered crucial to the job today.” 2
Using technology for learning, organizations can achieve better performance by upskilling the workforce. The
underlying reasons why it is necessary to adopt modern learning solutions and suggestions on how you can
implement these solutions into your own processes are explained in this white paper.
A U T H O R S
Janne Hietala started his first company at the age of 21. Selected
as The Young Entrepreneur of the Year in 2012, Janne has been
leading Valamis' commercial operations since 2008. He is respon-
sible for commercial strategy and development within the key
areas of marketing, sales, product development and services.
He completed his Executive MBA at the London Business School.
Kevin Groh is a German author and Marketing Specialist with a
background in electrical engineering. As a trained industrial clerk
Kevin quickly moved to the marketing area while studying to be
an industrial engineer. At the same time he published his first two
books in the German market. He started at Valamis in 2018 and
he is responsible for the marketing in the DACH region.
Janne Hietala CCO, Valamis [email protected]
Kevin Groh Marketing Specialist, Valamis DACH [email protected]
L A Y O U T Kimmo Pukkila
I M A G E S Kimmo Pukkila, Johanna Kokkola,
Jussi Ratilainen, Getty Images
E D I T O R S Katy Roby, Jenni Härkin
E X P E R T S Ville Tuominen, Samu Kuosmanen,
Jens Harju, Ana Gebejes, Mika Kuikka
I N D U S T R I A L R E V O L U T I O N
500 YEARS for knowledge to double
in 1400-1900
I N F O R M A T I O N A G E
13 MONTHS for knowledge to double
in 2017
P O S T - I N F O R M A T I O N A G E
12 HOURS for knowledge to double
in 2020
3
TABLE OF CONTENTS
Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .2
1 . Reskilling in Response to Rapid Technological Change . . . . . . . . . . . . . . . . . . . . . . .5
2 . Implementing Corporate Learning into Everyday Habits . . . . . . . . . . . . . . . . . . . . . 6
3 . What Is AI and How Can It Improve Learning? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
The True Nature of AI
AI and Personalization
4 . How to Build AI for Learning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
General Idea
Laying the Foundation for AI
Know What Success Looks Like
Collect High-Quality Data
Leverage Data Science
Take No Shortcuts
Real World Example: Onboarding
5 . Use AI to Achieve High Performance in Corporate Learning . . . . . . . . . . . . . . . . .12
Achieve Efficient Engagement With AI
Mine the Learning Data
Connect Team Performance Data With Learning Data
Harness the Knowledge
Real World Example: Meet Jack
Chatbots as a Part of a Learning Solution
6 . Achieve High Performance in Business With AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
Possible Steps to Take for Introducing AI in Teams and Projects
AI Helps You to Optimize Learning to Achieve Business Goals
7 . Conclusions: Benefits of AI in Corporate Learning . . . . . . . . . . . . . . . . . . . . . . . . . 20
Key Takeaways
Glossary (A-Z) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
4
5
Reskilling in Response to Rapid Technological Change1Technology is evolving every day, becoming more and more capable of doing many routine tasks more efficiently than any
human can. This will result in the replacement of entire jobs. Being replaced by technology is sometimes regarded negatively
and met by fear. However, this is not entirely accurate. According to Gartner’s prediction, “In 2020, AI becomes a positive net
job motivator, creating 2.3 million jobs while only eliminating 1.8 million jobs.” 3
Still, “Nearly 80% of IT and business executives project that the skills and knowledge their organizations will need in 10 years
will have little resemblance to those they have today.” 4
Organizations are facing a tough decision for the future: If most of the workforce becomes redundant, should they fire and
hire for new roles, or should they reskill them to meet the new business-critical needs? There are various factors to consider,
from employee retention and loyalty, to financial aspects and competitive advantages, to reputation and attracting new talent.
According to the World Economic Forum, in many cases it is cheaper to reskill workers than to hire new ones5. It is also a
more sustainable compared to firing and hiring.
In 2018, International Data Corporation (IDC) forecasted worldwide spending on cognitive and artificial intelligence systems
will reach $77.6 billion in 2022, more than three times the $24 billion forecast for 2018.6 The massive investment in AI will
result in necessary reskilling in order to keep up with the new shift in job roles brought by the new technology adoption.
The fearful outlook spurred by the thought of losing jobs has shifted to the expectation that this change will actually create
a lot of new jobs. Despite old jobs becoming obsolete, the process of automation will help give rise to new practices and
opportunities. This increases the need for the workforce to learn new skills and develop so that they can fulfill the require-
ments of the new jobs that emerge.
Gartner stated, “Data science and AI skills are in great demand, while talent remains in short supply. Job postings analyzed
with Gartner’s TalentNeuron™ show more than 236,000 data and analytics job openings in the U.S. alone, a 43% growth rate
year over year.” 3 Tasks concerned with topics like machine learning and data science, as well as creative work, will become
more popular, while routine tasks will become automated.
“AI techniques, including machine learning (ML) and natural language processing (NLP), are increasingly taking over and
automating many of these nonroutine tasks. Using AI techniques to augment back-office work will drive greater employee
productivity and increase employee (and customer) satisfaction.” 3
In order to enable employees to meet this increasing demand for new skills, personalization will be a key factor in corporate
learning, and artificial intelligence is the main tool for achieving the level of personalization necessary.
$77.6 BILLIONYEAR 2022
YEAR 2018 $24 BILLION
Global spending on cognitive and AI systems
F I G U R E 1 . IDC predicts rapid growth in investment on AI systems, which is already changing the job market with data
science and AI skills in great demand, according to Gartner.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
6
Corporate learning is nothing new for most organizations, and many global players have already realized its advantages and
have invested in improving their learning programs. Despite those growing efforts, many organizations are having a hard time
convincing employees to utilize corporate learning.
Just like the need for corporate learning has evolved, so too have the software solutions to address those needs. The core
idea of corporate learning software solutions is to organize learning content and make it more accessible for learners. With
modern Learning Management Systems (LMSs) and Learning Experience Platforms (LXPs), there are sophisticated possibilities
to have content libraries alongside a cognitive search tool that can deliver the most relevant and personalized content to the
learner. Still, learner engagement does not improve as much as organizations are hoping for.7
So how can organizations engage employees to learn? Five-day seminars twice a year are still a reality in some companies.
People don’t want to learn things they won’t need and can’t apply to their own roles. Thinking about presentation slides and
weeklong seminars makes most employees yawn.
The first step in engaging employees to learn is to make accessible learning content that fits the learners' individual needs.
Learners also need space and permission to learn, as well as limits that remove uncertainty and create a safe atmosphere. The
second step is to segment the learners and recommend them relevant content in the right context. If the employees study
during the work day — the learning has to fit their schedules, and the content needs to be accessible with mobile devices.
Implementing Corporate Learning into Everyday Habits2
7
The third step is to deliver the right content to the individual learners. The delivery can be done by recommending the right
content with the help of a recommendation engine and push notifications, and in an ideal case, the organization's learning
culture and tools in use will make learners want to come back for more.
The top management of many organizations has a significant role in communicating the importance of learning and encour-aging the whole community to become a learning-supporting
culture, especially at the beginning of the change process.
The internet has changed our behaviors as consumers. We have grown used to getting personalized recommendations, and
everything is just a few clicks away. It is clear that this behavior also transfers to the workplace. In order to engage people
to learn, the means to deliver learning need to be modernized to match and surpass society’s everyday consumption of
personalized content. Organizations must realize that in the workplace, employees are their internal customers; employees
buy in to the company mission and become advocates for the brand to attract more talent, and they are more engaged
to gain more skills that will in turn make the company better. The feeling of the content being personalized is the key to
learner engagement.
David Wilson, CEO of the Fosway Group, states that “too much focus is on delivery and too little focus is on what really
matters: the learner and what works best for them. Personalizing elements of training or learning content alone is not going
to build the relationships and engagement that will keep people coming back. There is an important shift in mindset towards
thinking of learning as an ongoing experience — not one-off events or a single transaction — but a blend of opportunities
and engagements across a tailored trajectory that engages and develops people over time.” 8
This requires companies to truly understand how, where, and when people are learning. Some individuals pull out their
smartphones and search the internet for an answer. Some people are looking to become more informed on a general topic,
and sometimes they will have a specific question in mind. In that case, a 20-page article on the subject is not ideal for the
context, but rather a small, quick chunk of information that targets the exact question is more suitable for the learner.
According to Fosway Group’s 2019 Digital Learning Realities Research, learner engagement is one of the most desirable
features needed to improve an organization’s approach to digital learning. Personalization is a key factor in this, with nearly
70 percent of the 800+ L&D professionals rating it as an important feature.9
The individual, their need for an individual learning experience, and their case-by-case provision of content pieces appear to
be an impossible task at first, but AI is a great tool to help with this.
TIME
EMPLOYEE’S PROGRESS
Detecta problem
Reach a solution
Recommend context-sensitive,
personalized learningcontent
WELL-BEINGLESSON
“HOW-TO”VIDEO
TECHNICALDOCUMENTATION
F I G U R E 2 . AI can help to detect the need of learning and recommend appropriate on-demand content.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
8
Nowadays, the term AI comes up in various contexts with increasing frequency. It is also a very useful tool in the corporate
learning arena. Before we dive deeper into how AI applies to learning, we should shortly define what AI means. There are
lots of definitions of AI to be found in various contexts. Here’s one example: The English Oxford Living Dictionary defines
Artificial Intelligence as, “The theory and development of computer systems able to perform tasks normally requiring human
intelligence, such as visual perception, speech recognition, decision-making, and translation between languages.” 10 Amazon
Web Services also mentions learning, problem-solving and pattern recognition in their definition.11
In general, most definitions focus on four main areas: human thinking, rational thinking, human acting and rational acting.
While those aspects appear to be completely separate at first, they are all part of the overall AI topic. For a computer to
think and act like a human means enabling it to be perceived as human-like in any situation. This includes the ability of the
computer to understand and use language, memorize what it hears and learns, use this information to draw new conclusions
and find answers, and adapt to changing circumstances. More sophisticated models of AI would even have robotic bodies
that are able to detect objects and manipulate them. All of those capabilities help to make the machine appear more human.
Thinking like a human would require mapping the human brain processes and translating them into a computer program. The
focus here is not on how effectively AI solves a task, but rather on what steps were taken to solve the problem and how similar
they are to a human approach. At this point, we need to move from computer science to cognitive science. The underlying
purpose of a machine, however, is to go beyond human levels of intelligence and skills. This is where the rationality factor
comes in. Aside from being lifelike, AI should be able to solve highly complex problems and perform with machine-like preci-
sion. Rational thinking is different from human thinking, since it is solely meant to find the optimal result for any problem, free
of emotions, preferences or morals. Rationality is based only on logic.
Rational behavior sometimes requires more than just logic. As soon as physical aspects like a robotic body are involved,
thinking about something may take too long to act rationally. Quick reactions must be taken without them being an inference
to prior thinking, like evading a thrown object to avoid being hit by it.
If we want to define AI, we must take all of those four areas into account and say that AI is an intelligent computer system that uses human thinking to find rational solutions, and then it acts
on those rational solutions with humanlike behavior. In the end, there is no fixed definition though, since the perception of AI is different for everyone and it depends on the desired outcome.
Lauri Järvilehto, founder of the Finnish Academy of Philosophy and co-founder of the learning game studio Lightneer, warns
us to be cautious of misinterpreting AI: “The biggest challenge by far is the overinflated hype. We shouldn't even talk about
Artificial Intelligence — no such thing exists. Machine learning, or algorithm-driven statistical big data analyses, are far less
developed in reality than all the hype suggests. To this end, it is especially critical to educate the leadership in companies to
understand what machine learning really is.”
Järvilehto continues: “Having said that, machine learning and complex algorithmics will be a more prominent part of running
business and market analyses in any companies moving forward. Such tools can help business leaders manage large data
sets in a convenient way and provide new and more salient insight.” 12
The True Nature of AI
The term that best describes the “true nature” of AI is machine learning. There are many different ways to approach machine
learning. Most techniques are either supervised or unsupervised, depending on labeled or unlabeled data. One example of a
learning technique is deep learning. Deep learning uses a cascade of multiple layers of nonlinear processing units for feature
extraction and transformation. The “deep” refers to the number of layers through which the data is transformed.13 Another
common technique is reinforcement learning, which uses responses. If an action caused a positive response or a negative
response, this will result in the machine doing it more or less often.
What Is AI and How Can It Improve Learning?3
9
AI and Personalization
AI is very often used in combination with personalization. This is because personalization is becoming more and more inte-
grated into our lives, and engages us to consume content. Music, videos, social media, and online shopping are all personal-
ized and delivered through your personal devices. Good examples of this can be found everywhere: Google search, Netflix
recommendations, Amazon offers, and devices like Google Assistant, Amazon Alexa, and Siri are all facilitating AI to enhance
the quality of information they provide to the user.
All these services make informative suggestions according to our preferences and previous experiences, and also based on
the behavior of similar groups of people. This leads to increased efficiency and time saved by providing the right information
needed, clearing out irrelevant data.14
The recommendation algorithms for typical commercial applications are based on classification, ranking and user ratings.
They only work if the content has been rated, so the machine can use this data to understand the value of the content for
certain audiences.
The success of AI in the commercial area demonstrates the power of personalization. It increases people’s engagement,
because everybody wants to be treated as an individual. If a person feels that something is tailored to them, it makes them
feel special, and they will be more likely to be engaged and find it beneficial15. Leveraging this concept in other areas is
the next logical step. If this works exceptionally well in business with consumers, why wouldn’t it also work well internally
in organizations?
Machine learning
INPUT FEATURE EXTRACTION CLASSIFICATION OUTPUT
Car
Not Car
INPUT FEATURE EXTRACTION + CLASSIFICATION OUTPUT
Car
Not Car
Deep learning
Reinforcement learning
NEGATIVE RESPONSE POSITIVE RESPONSE POSITIVE RESPONSE
Ouch!
F I G U R E 3 . Common types of AI: Machine learning, deep learning and reinforcement learning.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
10
How to Build AI for Learning4
General Idea
Simply put, AI is nothing more than an advanced way of processing information, a program designed to calculate,
predict and take actions to reach certain goals. Without information, it cannot operate.
The more high-quality data you have, the more precise the results will be. Big players in the market like Google and
Amazon are collecting all kinds of data (Big Data) from millions of users: behavior, preferences, activity, education,
purchases, website visits and much more. They use web browsers and apps and various other sources to gather
that information. This grants a multitude of insights into the way people behave but also insight into their own
businesses and how personalization influences desired behavioral outcomes.
The same thing can be achieved for learning activities in organizations by utilizing Experience API (xAPI). It can
track different types of learning activities and store this information in the Learning Record Store (LRS). xAPI can
collect data from other internal digital environments like a corporate intranet. Also, documents like resumes and
CVs or surveys that contain information about skills, trainings attended, and certifications can provide data on
pre-existing know-how.
The goal is to gain aggregated data about the learners that can be valuable for many purposes. AI can then find
patterns and correlations in learning behavior, and trends that are connected to certain industries, job roles, coun-
tries, regions, backgrounds and so on. Such findings often lead to the elimination of potential skill gaps, process
optimization, and other advantages that can improve your business.
1 1
Laying the Foundation for AI
With most things in life, you can’t jump from zero to fully functioning AI within a couple of minutes. There may be providers
that are offering you shortcuts, but there is no easy and quick way to perfect AI.
Every organization is different, and each one has to set custom goals for AI. Any out-of-the-box solution will struggle to
deliver the specific results your organization needs.
Know What Success Looks Like
The first step you can’t skip is to think about the exact goals you want to achieve, because they influence
everything you do, including how the data is prepared. It needs to be clear what you are trying to achieve as
an outcome. When you define your goals, your outcomes can be measured and made visible. Is AI the solution
you need to achieve your goals?
Collect High-Quality Data
Once you have set the goals, the second step should be to implement data engineering to collect quality
data as groundwork for any AI.16 Then it needs to be put in a usable form. For example, the easiest commonly
agreed-upon source format for learning purposes is xAPI.
Leverage Data Science
The third step is to use data science. This means utilizing feature engineering, analyzing data, and creating
models and visualizations based on the collected data. With data science, you can create models based on
your data. If the models are not good enough, you need to take a step back and collect better data and fine-
tune them. Only if the models are of high quality can you start with the last step: building the AI applications
where the data is processed, where AI will start to learn how to react. This is where the models are put into
use and can be adapted and improved.
Take No Shortcuts
It is important to note that many organizations believe that AI is going to change the world and it could make them immedi-
ately more successful, but they don’t always realize the amount of work it takes to prepare the data and create the models.
The two essential statements this paper intends to highlight in particular are concerning this point: There are no shortcuts
and you need to be careful to define the right goals.
The best way to start is to choose an area of the business where it is cheap to ramp up the data quality and do small-scale
experiments. Working this way you can provide evidence and create a sense of effort and value within the company as well
as get some practical experience.
R E A L W O R L D E X A M P L E : O N B O A R D I N G
A new employee, Rose, starts in your company. A traditional approach to onboarding could
be offering standardized learning material for every new employee. Because you have been
collecting data about previous onboarding experiences, you have gained more information
in order to offer her a more personalized onboarding experience. With data about learning
patterns and trends, you have been able to improve the quality of the learning materials
over time to make the onboarding process optimal. By tracking Rose’s learning habits, you
will be able to offer her learning materials in her preferred way. Microlearning, mobile apps,
and chatbots can help you to further personalize the learning experience and reduce the
time-to-competence, and thus accelerate productivity within.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
12
Use AI to Achieve High Performance in Corporate Learning5Personalized learning has a large impact on learning outcomes. Employees are motivated to learn things that they are inter-
ested in, in their preferred way. Personalization allows learning to better support the work of individuals, and makes it possible
to achieve organizational goals. What’s the best way to engage the employees to learn, and more importantly, to learn the
right things? The first factor is the right timing. Organizations should be able to offer learning when it is most needed. The
second thing is providing relevant content for the learner.
Companies often offer loads of learning content, but the needed information is hard to find. Julie Hiipakka, leader of learning
research for Bersin, Deloitte Consulting LLP, describes the advantage of using AI in corporations:
“There’s no shortage of content available for those who are continuously looking to learn, but organizations have worried
about what information is relevant and helpful — some are still building human curated content libraries instead of fully
opening up everything available for their workers. AI enabled search and curation can make open searching become less of
a fishing expedition. This technology can help those seeking answers to their questions get their answers faster. As workers
spend less time seeking answers, they can apply them, and test what’s working. As these systems continuously learn from
experience, they’ll get even better at serving up what works.” 17
Achieve Efficient Engagement With AI
The well-known recommendation systems like Spotify and Netflix are built to keep people consuming more content. Based
on our experience, many companies think that having more content will solve engagement issues, but this is not necessarily
the case.
Even though you want your employees to learn, you may not want your employees to spend more time on learning. More
likely, you want to ensure that they learn efficiently and then keep working on their tasks. So, unlike for example, Amazon, a
large corporation, driving engagement is not the goal to optimize.
What is really needed is a way to include a contextual compo-nent of your organization, to create value for the learner by clarifying how this content is relevant to them specifi-
cally. If a person understands why they need to learn some-thing, they will be a lot more engaged in doing so.
Often the most efficient way to go is targeted recommendations that drive the learners to the content that helps them solve
current problems within the flow of work. To achieve this, a recommendation algorithm that helps employees to choose the
next best step for them needs to be built. This algorithm is specifically for your organization, and as your organization keeps
evolving, in order to keep the recommendations valid, you need to adjust your algorithms and environment accordingly.
Integration with communication tools like Slack enables the AI solution to notice when a person is asking the team to try to
solve a problem. Recommending appropriate content at this moment has a very high success rate. Conversation patterns
can be used as indicators for better content recommendations, and also help support social, peer-to-peer learning. Gaining
the right context, understanding the relevant content for AI to leverage will cause engagement to increase substantially.
Mine the Learning Data
When using AI solutions, data is the key, and a logical starting point. When it comes to learning, you can collect data, for
example, about the amount of time spent learning, the materials used, and the completion rates. You can track learning
activities, performance metrics and organizational data to create a solid foundation for learning personalization. You can
use natural language processing (NLP) in social collaboration tools like Slack to explore your employees’ learning history.
The history may contain hints that show what content is working for successful learners. It might also tell you what content
contributes to people disengaging with the platform.
13
Once the data is collected, the exploration, visualization and reviewing can begin. The idea is to find patterns and correlations
within your data that have not been obvious before. You can enrich the data with additional details and information to possibly
uncover more connections. For this, unsupervised machine learning algorithms are used.
Classification algorithms can recognize patterns and correlations in an unorganized mass of data, and determine what factors
lead to particular outcomes, and how to influence this process. Predictive models18 can help to identify the most significant
features, content and behavior that drive learning outcomes.
To further understand the behavior of the learners and what actions lead to increasing or decreasing engagement, you can do
additional surveys and further analysis. Adding qualitative data into the mix will help you to better understand the underlying
behaviors and root causes. The better you understand what happened, the better you can counteract in the future.
Based on the collected data, you can come up with a hypothesis. For example, “Learning has an effect on the likelihood of
getting a promotion.” Based on this hypothesis, you can build a supervised machine-learning model. There are three main
types of models available that use the data input to calculate certain outcomes:
These are the base models, and an organization needs to create a complex mix of them to reach its goals. Understanding of
the organization-specific corporate learning goals and challenges is the starting point for analytics model development —
focusing on the right issues will speed up insight generation.
C L A S S I F I C A T I O N A L G O R I T H M
Results in a probability, for
example if a learner has a degree .
R E G R E S S I O N A L G O R I T H M
Produces a specified number,
like a prediction of the required
learning hours to pass the exam .
R A N K I N G A L G O R I T H M
Refines a ranked list, like
the best options for further
learning materials .
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
14
Connect Team Performance Data With Learning Data
Learning analytics alone is not sufficient enough to calculate the decisive factors for business success. A good example is agile
development teams. You can see the difference in their performance, depending on analytical and data tools that combine
learning interactions and interactions in the real world. Evidence has shown that combining the knowledge and actions of
daily work leads to better decisions, so we need to find a way to make learning a part of the job and the daily work processes.19
Data from the organization's operational systems (ERPs, system logs, status reports, HR systems, etc.) can be used to identify
the best-performing individuals or teams. We can analyze the defining characteristics that these people hold. Furthermore, we
can use this information to guide others to improve these skills, behaviors or knowledge to become like the best performers
— i.e. the best teams provide patterns for optimal outcomes. Productivity, quality and customer satisfaction, in addition to
other metrics, can show you which learning content leads teams in the wrong direction so you can eliminate or alter it. This
enables you to see which learning content leads teams in the wrong direction so you can eliminate or alter it. It displays the
root of success and failure.
This means machine-learning modeling can have benefits like:
• Improved comparability and measurability of development work
• Transparency of development work
• Possibility of early corrective actions
• Improvement of development processes by identifying and eliminating root causes leading to weaker results
• More efficient use of resources
• Improvement of personnel know-how
15
Harness the Knowledge
AI adds value to organizations that need support with knowledge management. As the amount of knowledge grows expo-
nentially, NLP and AI are needed to make it possible to harness the knowledge. In order to improve the results of NLP, IBM
Watson technology can be taught by your company experts. The experts need to help improve NLP to further this capability
in order to find the right data inside and outside of corporations. NLP can process lots of videos and documents, and when
combined with machine learning can find connections and correlations between content pieces. It can do the same for
your workforce: locate skills and expertise more effectively and help you to connect the right people to the right tasks. It is
a helpful tool in creating a detailed segmentation both for your content and your employees’ skills. We call this Intelligent
Knowledge Discovery (IKD).
There is also the possibility to measure the working context of the user, measuring what tasks a person has and predicting
possible learning requests based on that. For example, if you are looking for an answer to a question and the search result
consists of five hour long videos, you most likely won’t watch them all. NLP combined with visual recognition locates the
exact moment in the video where you can find your answer. Your search results will be accurate and contextual, and you
will save time.
Intelligent Knowledge Discovery helps you to build a living, breathing network of experts and knowledge. It can be used to
retain the knowledge of retiring employees, to build an internal knowledge base from videos, to knock down knowledge silos,
or to reach learning content via integrations with, for example, Youtube, edX or Google Drive.
This concept, however, starts again with the data. So what is the right strategy for building AI?
Vast amounts of information Processing with AI Contextual search resultsMULTIPLE SOURCES IBM WATSON LEARNING PLATFORM
F I G U R E 4 . Using AI, IKD can help uncover trends, correlations and similarities in information.
R E A L W O R L D E X A M P L E : M E E T J A C K
Jack is a long-time employee in the organization, and is about to retire. Jack has a lot of
siloed knowledge that he has never written down. Since Jack does not have time to prepare
courses or presentations, he has decided to share his knowledge by answering the questions
the younger employees want to ask him. The younger employees vote for the best questions,
and every week Jack turns his webcam on and broadcasts a webinar, answering the questions
and discussing the topics. After the webinar, Intelligent Knowledge Discovery processes the
video, compares it to other content, sets the theme and tags, then indexes it, and makes
it readily available for anyone within the organization to search. Jack’s knowledge is now
saved and easy to access for future employees of the organization.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
16
Chatbots as a Part of a Learning Solution
People are getting used to AI assistants on various websites and even at home with devices like the Google Assistant or
Amazon Echo. According to Verto Analytics20, as of May 2017 more than 90 million U.S. adults (total monthly unique users)
used AI-powered personal assistants on their smartphones, and they interacted with their assistants an average of 38 times
per month. According to another study by Radio Centre21, in the UK people are interacting with voice assistants (on different
devices) more than 167.75 times per month, which is about 5.5 interactions per day. Given these numbers, it is only a matter
of time before AI assistants are expected in organizations as well.
One way to introduce AI to your organization is with an AI-powered chatbot as a learning assistant. It's easy to integrate
with learning environments and communication channels. An easily accessible learning assistant can communicate with the
learners and provide answers to their questions. This way, learners will better orient themselves in a learning environment
and receive personalized content recommendations.
Businesses can profit from chatbots by having improved onboarding, employee training, compliance training and much
more. The advantages of a chatbot are not only for internal purposes — they can also be leveraged in training for channel
networks, partner networks, resellers, and customers. They can be integrated with the intranet or communication channels
like Slack and inform employees about new learning materials, guidelines, and compliance regulation materials, as well as
help them with existing questions.
“Imagine lifelong learning companions powered by AI that can accompany and support individual learners throughout their studies...”
ROSE LUCKIN AND WAYNE HOLMES, PEARSON22
IBM Watson is constantly improving its AI by checking the success rates of recommendations. This way, chatting with a
bot will become increasingly convenient, and soon it will feel as natural as talking to a human colleague. Considering the
fast progress of chatbots, AI will become an important part of daily work in no time. Eventually, a chatbot will act like a
personal tutor.
17
Achieve High Performance in Business with AI6Which teams at the company are doing the most impactful work? Why is it the most impactful? Once you have
those answers, you will be able to use AI and predictive analytics to find out the causes and the probable effects
of certain actions.
A corporation’s processes aren’t always linear. For example, in patient safety in medical care, the optimal process
that can be statistically captured varies. Any action or previous piece of knowledge can influence the whole
process of patient care, and it can also influence the outcome. In this case, the goal is to reduce the number of
possible variations by using data to save costs. This can be achieved by making sure everyone knows the most
efficient procedures.
With teams, especially agile teams, e.g. in development and IT, the complexity is a lot higher. Predictive and
prescriptive algorithms work well in those cases because there is a lot of data available from the start, but the
quality is not always given (versioning, working times, etc.).
How can we use AI to utilize all the existing data and use the insights we can gain from it to improve the efficiency
of teams, thereby improving business performance as a whole?
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
18
Possible Steps to Take for Introducing AI in Teams and Projects
There are multiple ways to introduce AI in an organization, and finding the best model depends on data availability, burning
business questions that need insights, or resource availability, e.g. maybe just having a data scientist available and getting
them to work on a proof of concept. Data Science and Artificial Intelligence solutions usually require iterative development
and new ideas, and use cases come up as you progress on your model. The key thing is to start with something and get the
organization used to AI projects. Here are a few possible steps that can be taken:
Gather historical data (successful projects, failed projects, number of
releases, customer satisfaction and so on). This data is used for Business
Intelligence tools that predict future developments.
Example: A large bank starts using agile teams to generate innovation. They build a predictive system to
help them improve projects while they continue to run operations instead of looking back retrospectively
on lessons learned. While doing this, they are able to apply learned insights from completed projects in
other early-stage projects to increase their success rates from the start. This predictive system uses various
metrics as a basis. For example, an app development team could use the following metrics: App downtime,
number of bugs, bug removal efficiency, change frequency, etc.
Measure social patterns. Visualizing communication within a team compared
to the learning paths of the team members can reveal correlations. With
additional data, this allows you to recognize how learning influences actions
and decisions in the team.
Use AI to influence results in set-based design and feature engineering. In
feature engineering, test data is used to verify the accuracy of the trained
module and find weaknesses. It starts with the exploration of hundreds of
different metrics, matching them, highlighting problems, and detecting data
shortages. Match correlations, develop one module from them, and adapt
it, if necessary. Then you can either modify the training data or the module
to avoid any mistakes in the future.
Use machine learning or deep learning to turn predictive analytics into
prescriptive analytics. The difference is that the feature extraction from input
can be done by humans or by the system.
Executing these four steps successfully not only enables organizations to measure the way teams are working on projects,
but also allows you to understand the actions successful teams have taken. With this kind of information, it is possible to
predict the success of a project and support less effective teams with the right kind of training. The results can lead to a more
efficient way of completing projects, better teamwork, and in the end, a better ROI.
1
2
3
4
19
AI Helps You Optimize Learning to Achieve Business Goals
Measuring how learning activities affect business outcomes is the most significant problem in Learning and Development
(L&D). According to LinkedIn's first annual Workplace Learning Report: Business impact is the No. 1 measure desired by CEOs
in L&D and ROI is the No. 2 measure. And yet, only 8 percent of CEOs in L&D currently see the business impact, and only 4
percent see ROI of L&D.23
AI-powered technologies reveal hidden connections between learning activities and employee performance. As a result, you
gain transparency and insight into how this transforms business performance metrics, to envision what the learning impact
is. Over time, AI will understand more and more correlations in data and open up further connections to help organizations
understand how learning can be leveraged to affect their business. Understanding the learning impact will help companies
calculate the ROI of L&D. It will also help to continuously improve the base goals you set for your AI.
"For analytics efforts to yield impact, we need to analyze our learning efforts in connection with performance metrics (the behaviors we are trying to initiate in our learners) and - ideally - with business
metrics (how those behaviors affect the organization’s bottom line).”
A.D.DETRICK, S. VIPOND24
Dmitry Kudinov, Chief Technology Officer at Valamis, stresses the fact that in order to truly benefit from AI, L&D needs to
understand it first: “Improved outcomes of a learning process are the benefits that could be realized by properly applying
AI to corporate learning. Those benefits could be achieved through the personalization of learning, better adaptive learning
activities as well as deeper classification of learning materials, which is achievable by using AI. At the same time, I see the most
difficult challenge faced by L&D is the understanding of what AI is in general and what it can do for corporate learning. There
is so much hype around AI solving everything, and there are so few examples of practical benefits achieved in corporations,
so L&D either rushes to get a new system that claims to be “AI-driven” just to realize that they don’t know what to do with it,
or they decide to fall back on the existing systems that are somehow working.” 25
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
20
Conclusions: Benefits of AI in Corporate Learning 7In the beginning, we discussed the challenges the modern learner faces, and offered AI as one solution for companies to
meet their needs. We have explained the habits and expectations of millennials and how work will move towards continuous
learning. Additionally, we defined the term “AI” and its most common uses in organizations, focusing on the three main areas:
learning, business, and knowledge.
We learned that the starting point is always to collect and prepare data to bring it into a usable format. That alone is a benefit
in itself, because usable data is valuable for any organization. The most important factor is defining the right goals and building
every algorithm with these goals in mind.
We also learned that adaptive learning and recommendations can be made by including a conversational AI in the form of a
chatbot to solve one of the biggest problems of a classic LMS: How do I get people to use it, or increase engagement? This
also led to the additional question, of whether or not a lot of engagement is desirable for your organization.
Because AI can understand the personal needs of the user, what habits are common, what the forgetting curves are like, and when to act on those indicators, AI can curate learning
in a personalized and engaging way. It can even predict when the best time to recommend learning content is.
The working contextual understanding is key here, meaning that AI can segment people by various factors and rules like
behavior, attributes, search results, previous knowledge, job roles, colleagues, etc.
Using AI with those functionalities across all areas will allow a company to understand what makes people engaged, what
doesn’t, and how these various factors are influencing actual business outcomes. With this information, it is possible to learn
what steps are needed to improve these numbers and further your initial goals.
Nowadays, technologies like deep learning and reinforcement learning are shifting from concepts into actual operational
algorithms and platforms. They will become more and more effective for organizations to improve their processes.
Corporate learning has a background in education of course, but with chatbots and AI it moves more into the area of behavioral science:
How can we influence learning behaviors and engagement? It is also a good way to influence organizational change; you can introduce learning that supports cultural changes and other developments.
Technology’s presence in the workplace, combined with an increased demand for employee and employer learning and
development, and a lack of time to learn, results in a digital workplace where learning and technology are blending together.
That means you need a learning platform that integrates with all of your organization’s systems. This way, you and your
employees can benefit from permanently improved data from all sources. Improved data and this blended approach lead to
a better learning culture and results.
Digital learning will become more and more important as time passes, and organizations need to start adapting their processes
according to this trend. The ability to provide learning in the flow of work26 will determine future success and ensure compet-
itiveness in the upcoming years.
21
Key Takeaways
DEFINE YOUR GOALS & START SMALL
What do you want to achieve? Gather the challenges
you see and the disadvantages you are experiencing
in your current situation. Identify your 'If I knew XYZ,'
that would help you to reach your strategic goals
or transform the way you work. Choose your most
important goal and break it down into small steps that
are necessary to get there. Implementing complex
processes like what is required to introduce AI into your
environment is done best at a step-by-step pace. With
growing experience, you can more easily add additional
features later on.
IDENTIFY DATA REQUIREMENTS AND
COLLECT THE DATA
Take the necessary time and effort to organize your
data. This will help in effectively implementing the early
stages of AI. Make a list of all the types of data you are
collecting and in what form they are stored. Collect
data from different areas of your business. The more
data you are able to collect, the better AI can help you.
GET HELP
Consult a data scientist or an expert to see what kind of
software or solution may help you to reach your goals.
Choose an area of your business and start cleaning and
organizing your data with data engineering. This can
be any department you want. Doing experiments on
your own may save costs at the beginning, but fixing
errors is more expensive the later you do it. Planning
your project with an experienced strategic partner can
save costs and time.
START WITH A PILOT —
ITERATE TO IMPROVE
Building a model that works towards your set goals
requires multiple iterations. Usually you test several
models with the set parameters, then fine-tune these
parameters or even go back to modify the data prepa-
ration for your model of choice. Creating a satisfactory
solution with one single model in one run is very rare.
TEST YOUR MODELSThe quality of your AI depends on how well you teach
it. Test your models, define your groups, and make sure
that the outcome meets your expectations. Reflect
the results back to your initial business question, 'If I
knew...' Does the model answer this and help you to
achieve your goals? If not, do it again. Rushing the
early stages will result in suboptimal efficiency. When
you are absolutely sure that your models work, you
can continue optimizing them and start to build the
first machine-learning features with them.
PEOPLE ARE KEY
No matter what goal you are aiming for, it all comes
down to your employees. Including them in the process
as early as possible will help to ensure acceptance and
engagement. It will also help you to get a feeling for
the actual needs in your business.
REMEMBER, THERE ARE NO SHORTCUTS
There are no shortcuts to creating operational AI. If
you want it to achieve your individual goals, an out-of-
the-box solution will not be the perfect match for you.
In order to see the numerical impact of learning on
your business goals, there are many steps to take on
the way. It requires time, effort and experts in order
to reach that goal. Make a plan, consider challenges,
and expect delays — the effort will pay off in the end.
GET STARTED NOW
AI technologies are coming with increasing pace.
Soon there won’t be a way around them anymore. The
sooner you start, the easier it will be in the long run.
1
2
3
4
5
6
7
8
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
22
Glossary (A-Z) ADAPTIVE LEARNING
The scientific term for personalized learning. It describes the
learning method that uses algorithms to interact with learners
and resources in order to provide custom learning experiences.
ARTIFICIAL INTELLIGENCE (AI)
A scientific field concerning the development of computer
programs that can learn and eventually act without human
interference; eventually, machines that mimic "cognitive"
functions that humans associate with other human minds,
such as "learning" and "problem-solving".
BEHAVIOR SEGMENTATION
By analyzing learners and their daily actions, it is possible
to cluster them into different behavioral groups that have
individual requirements.
COGNITIVE SEARCH
Powered by AI within the system that is searched. An intel-
ligent search that understands the context and intent of the
person, delivering the most relevant results for each situation.
DATA SCIENCE
Covers multiple fields of science that use scientific methods,
processes, algorithms and systems to extract knowledge and
insights from data in various forms. This can happen in either
a structured or unstructured manner. The main focus of data
science is knowing how to extract meaning from data and
how to interpret data.
DEEP LEARNING
Refers to the complexity of a mathematical model, and that
the increased computing power of modern computers has
allowed researchers to increase this complexity to reach
levels that appear not only quantitatively but also qualitatively
different from before. Computers learn complicated concepts
by designing them out of simpler ones. When creating a graph
of how those concepts are built on top of each other, it will
be quite deep — therefore, deep learning.
EXPERIENCE API/XAPI (TIN CAN API)
An e-learning software specification that records and tracks all
types of learning experiences, also allowing learning systems
to communicate with the learning content.
FEATURE ENGINEERING
The process of using specialized knowledge to create features
that allow machine-learning algorithms to work as intended.
A feature is a specific characteristic or property of a phenom-
enon that is observed.
FORGETTING CURVE
German psychologist Hermann Ebbinghaus published a
hypothesis in 1885 that resulted in the creation of the forget-
ting curve. This is a graphic depiction of the decline of human
memory retention over time. Simply put: It displays how
fast we forget things and how this changes with repetition.
FOURTH INDUSTRIAL REVOLUTION
After the steam engine, electric power and digitalization,
we are now entering the fourth revolution in industry:
connectivity and technological breakthroughs. Also known
as Industry 4.0, some of the most famous topics are AI,
robotics and nanotechnology.
IBM WATSON TECHNOLOGY
Initially, a computer system developed for answering
questions posed in natural language, the Watson tech-
nology by IBM is now the basis for many AI technologies
combining various programs that can do more than just
answer questions.
KNOWLEDGE DOUBLING CURVE
Created by Buckminster Fuller in 1982, depicting his obser-
vations and predictions concerning the multiplication of
human knowledge over time and how this would accelerate
in the future.
LEARNING EXPERIENCE PLATFORM (LXP/LEP)
A digital platform for personalizing existing learning content
by utilizing known factors from a learning record store and
providing individual content for each user, thereby increasing
engagement and success.
LEARNING IMPACT
Describes the positive effect of learning in an organization.
This includes employee know-how and skill improvement,
overall better performance, and eventually better business
outcomes. It also describes the learning retention or how
well learners can understand and utilize new information.
LEARNING MANAGEMENT SYSTEM (LMS)
Software for administration, documentation, tracking,
reporting and delivery of educational content, courses and
programs.
LEARNING RECORD STORE (LRS)
A database where all the xAPI statements and previous infor-
mation about all users is saved and stored. It is the database
for all AI functionalities and personalization.
LIFELONG LEARNING
Describes the ongoing pursuit of knowledge for either
personal or professional purposes. The key here is that it is
continuous, voluntary and self-motivated, not enforced by
external factors.
MACHINE LEARNING
A subset of artificial intelligence. It is the scientific study of
algorithms and statistical models computers use to perform
specific tasks without the help of patterns, explicit instructions
or external interference. There are vast amounts of tools with
the purpose of understanding data. Those tools can be clas-
sified as either supervised learning or unsupervised learning.
Machine Learning enables computers to tackle real-world,
knowledge-intensive problems by providing answers that
appear subjective.
23
MICROLEARNING
Based on the idea that the attention span of people is short
and their time for learning is very limited. In order to ensure
maximum efficiency, the learning content is broken down
into small pieces and delivered individually. This way, people
can learn at the best time and only the topics they need in
each specific moment, in the right context.
MULTICLASS CLASSIFICATION
Using algorithms that can recognize patterns and correla-
tions in an unorganized mass of data and determine what
factors lead to what outcomes and how to influence this
process.
NATURAL LANGUAGE PROCESSING (NLP)
The scientific study of human-computer interactions based
on languages. Includes speech recognition, NL-understanding
and NL-generation in order for a computer to process large
amounts of natural language data.
An easy example: A person types “Can I play soccer today?”
NLU will catch the meaning by analyzing grammar and
context and turn it into an intent. NLP will convert the text
into structured data. NLG will generate a text that is based
on this structured data. The intent would be knowing if one
can play soccer on a particular day, so the computer would
check the weather and respond, “It is raining today, so it is
not advised to play soccer outside.”
PERSONALIZED LEARNING
Learning is a highly individual discipline where every person
has their own unique way of doing it most effectively. In
order to achieve maximized success, it is necessary to enable
learners to get what they need, when they need it, and how
they need it. The feeling of authenticity and value within the
individual is the key to personalization.
PREDICTIVE ANALYTICS
A type of analytics concerned with the future. It learns from
past developments and predicts possible and likely future
outcomes. This is supposed to enable preventive actions
based on these insights.
PREDICTIVE BEHAVIOR MODEL
By analyzing user behavior over time and including other
factors like results, habits and success rates, it is possible to
create a predictive model capable of detecting early signs
for future problems.
PRESCRIPTIVE ANALYTICS
Goes beyond predictive analytics. The goal of this type of
analytics is to provide advice on possible solutions for the
future. Companies can assess those solutions and their
probable outcomes.
REINFORCEMENT LEARNING
Systems that improve their performance in a given task with
more and more experience or data. They will execute the
task and adjust their behavior based on positive or negative
feedback combined with a lot of repetition.
RULE-BASED DECISION MAKING
A technique providing set rules for a computer to base its
decisions on. These can be precise rules or more general rules,
depending on the tasks the computer is supposed to perform.
SCORM
The most commonly used standard for courses and learning
content in organizations to date. However, it is neither track-
able nor can it be measured in detail. For those and other
reasons, it is now often replaced by the more modern xAPI
standard.
SEQUENCE-AWARE ALGORITHMS
Not every piece of content is meant to be consumed on its
own. Many topics are only useful if they are consumed in
a specific order. This order is not necessarily given by the
designers, but by the common way people are using it. This
algorithm analyzes the order in which users are looking at
content to recommend often-viewed pieces automatically
to users.
SET-BASED DESIGN
A development practice that keeps requirements and design
options flexible for as long as possible. Instead of choosing
a single-point solution upfront, it identifies and explores
multiple options, eliminating poorer choices over time.
TIME-AWARE ALGORITHMS
This type of algorithm analyzes the time of day, the day of
the week, and other time-based factors that may influence
the consumption of content for each user. Over time it will
learn patterns and know at what time of day on what days
of the week a certain segment of users is the most active.
TIME-TO-COMPETENCE
This is the time period from the first contact of an employee
with a new professional topic to the moment they can actively
use this knowledge in a productive and competent manner.
UNSUPERVISED MACHINE LEARNING
Using statistics to make sense of raw data without requiring
human input. The most commonly used technique is the
cluster analysis that groups the data that has not been
labeled, classified or categorized. It creates groups based
on commonalities in the data.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
24
Sources
1 Fuller, Buckminster; Richey, Dr. Michael. 2015. The Boeing Company. https://www.nap.edu/read/25038/chapter/5#80
2 World Economic Forum. http://reports.weforum.org/future-of-jobs-2016/skills-stability/
3 Poitevin, Helen; Sau, Moutusi; Sicular, Svetlana; Hunter, Eric; Howson; Cindy; Maita, Kanae. 2018. Predicts 2019: AI and the Future of Work. Gartner.
4 Coleman, Mark; Poitevin, Helen. 2017; refreshed 2019. Rethinking Human Capital Management Technologies to Support Emerging Talent Operating Principles. Gartner.
5 World Economic Forum. 2019. Towards a Reskilling Revolution. http://www3.weforum.org/docs/WEF_Towards_a_Reskilling_Revolution.pdf
6 IDC Press Release. Worldwide Spending on Cognitive and Artificial Intelligence Systems Forecast to Reach $77.6 Billion in 2022, According to New IDC Spending Guide, September 2018 https://www.idc.com/getdoc.jsp?containerId=prUS43662418
7 Sivalingam, Nelson. 2018. The LMS is broken, so what’s next? Training Journal. https://www.trainingjournal.com/articles/opinion/lms-broken-so-what%E2%80%99s-next
8 Wilson, David, Fosway Group. 2019. Path to learning personalization. https://www.valamis.com/blog/the-path-to-learning-personalization
9 Digital Learning Realities Research 2019. Fosway Group. http://www.fosway.com/research/next-gen-learning/digital-learning-realities-2019/
10 Definition of Artificial Intelligence. 2019. English Oxford Living Dictionary https://en.oxforddictionaries.com/definition/artificial_intelligence
11 What is Artificial Intelligence?. 2019. Amazon Web Services. https://aws.amazon.com/machine-learning/what-is-ai/
12 2018. Interview with Lauri Järvilehto, Lightneer; Valamis Group.
13 Deep Learning. 2018. Wikipedia. https://en.wikipedia.org/wiki/Deep_learning
14 van der Goes, Maurits. Takeaways from Netflix’s Personalization Workshop 2018. rtl Tech Blog, a Medium Company. https://medium.com/rtl-tech/my-takeaways-from-netflixs-personalization-workshop-2018-f564a19437b6
15 Bright, Laura Frances. 1977. Consumer control and customization in online environments : an investigation into the psychology of consumer choice and its impact on media enjoyment, attitude, and behavioral intention; The University of Texas at Austin. https://repositories.lib.utexas.edu/handle/2152/18054
16 Ransboham, Sam. 2016. The Smart Way to Deal With Messy Data. MIT Sloan Management Review. https://sloanreview.mit.edu/article/the-smart-way-to-deal-with-messy-data/
17 2018. Interview with Julie Hiipakka, Bersin, Deloitte Consulting LLP.
18 Gandhi, Rohith. 2018. Naive Bayes Classifier; Towards Data Science. https://towardsdatascience.com/naive-bayes-classifier-81d512f50a7c
19 2011. Learning while working: how skills development can be supported through workplace learning. European Centre for the Development of Vocational Training. http://www.cedefop.europa.eu/en/news-and-press/news/learning-while-working-how-skills-development-can-be-supported-through-workplace
20 Hwong, Connie. Verto Index: Assistant Apps; Verto Analytics. http://www.vertoanalytics.com/verto-index-personal-assistant-apps/
25
21 2017. Voice assistant trends [Infographic]. IFTTT. https://ifttt.com/blog/2017/07/voice-assistant-trends-infographic
22 Luckin, Rose; Holmes, Wayne; Griffiths, Mark; Forcier, Laurie. Intelligence Unleashed: An argument for AI in Education; Pearson. https://www.pearson.com/corporate/about-pearson/innovation/smarter-digital-tools/intelligence-unleashed.html
23 2017 Workplace Learning Report. LinkedIn. https://learning.linkedin.com/blog/learning-thought-leadership/introducing-the-2017-workplace-learning-report--top-trends---cha
24 Detrick, A.D.; Vipond, Sharon. 2016. Learning Analytics: A Practical Pathway to Success https://www.elearningguild.com/insights/204/learning-analytics-a-practical-pathway-to-success/
25 2018. Interview with Dmitry Kudinov. Valamis Group.
26 Bersin, Josh. 2018. A New Paradigm For Corporate Training: Learning In The Flow of Work https://joshbersin.com/2018/06/a-new-paradigm-for-corporate-training-learning-in-the-flow-of-work/
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives
4.0 International License.
The opinions expressed in this document represent the current view of the authors on the issues discussed as
of the date of publication. Because of chancing market conditions, authors cannot guarantee the accuracy of
any information presented after the date of publication. This white paper is for informational purposes only.
YOUR DEFINIT IVE GUIDE TO AI IN CORPOR ATE LEARNING
26
If you have any questions, or would like to know more about our AI technologies or
learning solutions, don’t hesitate to contact us:
www.valamis.com
Valamis Benelux: Chris Göbel
+31 6 15065227
Barbara Strozzilaan 201
1083 HN Amsterdam
The Netherlands
Valamis Nordics: Jari Järvelä
+358 50 564 3101
Koskikatu 7 A, 4. Floor
80100 Joensuu
Finland
Valamis UK: Tom Ridley
+44 (0) 7525 049 778
Level39, One Canada Square
Canary Wharf, London E14 5AB
United Kingdom
Valamis USA: Mika Kuikka
+1 (617) 959 7438
175 Federal St Suite 860
Boston, MA 02109
United States
Valamis Germany: Holger Bräunlich
+49 (0) 611 532 88 680
Klingholzstraße 7
65189 Wiesbaden
Germany