Learning and Development

Applications of AI in Learning and Development (Training Needs Analysis) Part I

Who should read this blog? You should read this blog If you are a part of the Learning and Development team or an Operations or a Technology Leader wanting to understand the different possible applications of AI across your Learning and Development function.

This blog is going to focus on AI in the first step of the process that is: Training / Learning Needs Analysis.

2020 is the year where L&D is going to have a critical role to play in the Business as they will drive significant Business Outcomes. Why so? Because the digitally transformed world we live in today, has a rapidly changing skill map. The Average skill shelf life has decreased to about 5 years and the need to be agile to have new skills has proportionately increased.

Let us look at the top 10 Soft Skills required in 2020 and compare it with 2015. Significant change, wouldn’t you say? Creativity and critical thinking have moved up, emotional intelligence and Cognitive Flexibility are new skills that have been added. Cognitive Flexibility in many ways is like parallel processing, it is the mental ability to switch between thinking about two different concepts, and to think about multiple concepts simultaneously.

Source: Top 10 Skills in 2015 & 2020 (Source: World Economic Forum, 2016 https://www.weforum.org/agenda/2016/01/the-10-skills-you-need-to-thrive-in-thefourth-industrial-revolution/

The top 10 hard skills are:

  1. Cloud Computing
  2. Artificial intelligence
  3. Artificial Reasoning
  4. People Management
  5. UX Design
  6. Mobile Application
  7. Video Production
  8. Sales Leadership
  9. Translation
  10. Audio Production
  11. Natural Language Processing

How many of these would you have recognized back in 2015? Barely any.

The global training spends in 2019 was $370.3 BN an increase of ~40% in a decade. (https://trainingindustry.com/wiki/outsourcing/size-of-training-industry/). To hire and retain good talent organizations will literally have to become learning campuses and ensure they are owning lifetime learnability of their employees.

The first and most significant step in establishing a learning campus is to assess the learning needs. Also called training needs analysis, unfortunately this one step that is the most rushed and usually not completed.

Training needs analysis is conducted at three levels:


However, most organizations manage to complete the 1st 2 levels only. Individual level done manually, is a highly time consuming and intensive process and therefore is usually left incomplete or done in parts.

The data and information needed to assess at an Organizational level is Business Goals, Employee & Skills Inventory, Customer Satisfaction Data and Organization Culture. Usually easily available.


Next you look at the Operational Level which is more detailed but again available with each of the operation & functional HR heads. Data like Job description & specifications, Work performance standards and metrics.

Job Description

The 3rd Level is the Individual level which is most challenging. As you need to get into a lot of data mining across the organization. Performance appraisal of an employee, to skill assessments, Interview questionnaires, work samples all this data needs to be collected, looked, and analyzed together to tell you the complete story.

Data From Any Where

Imagine even for an organization of 100 it is a lot of data and if we are talking of 100,000 it is a mammoth task.

Artificial Intelligence can be of great help here. As its most significant accomplishment is that it can help you process large amounts of data at scale yet personalized at significantly lower costs.

A key component of Learning Needs Assessment is assessing the skill gaps. And that is where one of the oldest applications of AI Adaptive Testing can be of significant help. This technique dates back to 1970s’. Companies like Pearson have created several adaptive tests to be able to provide personalized coaching to students.

The corporate sector is a late adopter however we have platforms like IRIS from Plural Sight which combines adaptive testing theory with machine learning to create statistical models for skill levels. To stay current, Iris adapts and collects data on trending technologies in terms of which are popular, and what is getting obsolete.

Adaptive tests when combined with machine learning can maximize Learning Need Assessment measurement efficiency. The Assessment will be more accurate and precise too. Since Adaptive tests adapt to the test taker’s ability the assessment times are shorter.

Adaptive assessments work by leveraging Item Response Theory (IRT). IRT focuses on the difficulty of each individual item as in question, rather than the overall assessment. As users answer questions, the adaptive assessment infers their likely skill level with increasing probability based on the difficulty of each question answered.

Because you calculate probabilities of the employee’s capability level It becomes unnecessary, then, to serve up all assessment questions, enabling short-form assessments to replace traditional long-form tests. So, in a way based on the questions answered you already know the questions you don’t need to ask. The probability of a correct response is determined by the employee’s ability and the difficulty level of the question.

If you were to plot the probability curve it would look like this. As the ability increases the probability of answering the question goes up. For example, at Ability 5 the probability is .98 and at 0 its .5.


Adaptive learning can be used both for Hard and Soft Skills. We saw an example of hard skills with IRIS where they are using it for assessing the technical skills of their employees. For Soft skills you could create a host of psychometric tests across the following areas and run it on an adaptive learning platform:

  • Personality Tests: Personality tests are a method of assessing human personality constructs. They are based on collecting personal information about people they draw inferences about how an individual think, feels and behaves. That in turn can predict persons performance at work when it comes to interpersonal skills, management style, motivation levels, ability to handle crisis and performance under pressure
  • Cognitive Tests: Cognitive tests are all about measuring competence and intellectual capabilities. They’re able to fairly accurately predict performance at work as they measure a person’s thinking abilities such as perception, reasoning, memory, verbal, and problem-solving ability. They also test for the ability to solve problems when learning new job skills or tackling workplace issues.
  • What about soft skills or acquired skills? can you think of ways we could use adaptive assessments to measure these. Challenging but not impossible. If you could find ways to marry on the job performance data to personality and cognitive traits of an individual, you could get a good measure of progress made on soft and or acquired skills.

What are the Benefits of Adaptive Assessments?

Highly Accurate: Adaptive testing allows you to identify an employees’ true level of ability faster and more accurately than with other types of assessments. By starting with a question of average difficulty, and then asking harder questions when they get it right and easier questions when they get it wrong, each additional question homes in on a narrower and narrower range of ability. Until the test has zeroed in on an employee’s exact level of ability and you get to the specific skill gap that needs to be addressed.

Positive Employee Experience Since each examinee is challenged appropriately during an adaptive test, the overall experience is more positive than a traditional assessment. Low performers are not discouraged or intimidated, while high performers won’t get bored and even enjoy receiving more difficult items. A better test-taking experience encourages employees to try harder than they might with a conventional test. Low performers end up feeling more comfortable and less judged and high performers will feel challenged and more motivated.

Innovative and Leading Edge: Adaptive testing leverages artificial intelligence, as candidate responses to each test question inform the next item that appears in the test.  This dynamic and interactive back and forth between the candidate and the test quickly homes in on the candidate’s true level of ability, more so than any other type of assessment available today. They can use various forms of audio, visual, and video content in their adaptive tests.

Adaptive learning represents a paradigm shift—from the conventional model (an instructor-centric, passive learning experience) to an intelligent one (a learner-centric, interactive, active learning experience). In the adaptive model, each employee is paired with a virtual “coach.” It’s a concept that can be scaled to millions of employees at a fraction of the cost of human coaches.

Do you want to put an adaptive learning platform to work? Do share with us your thoughts in the comment section below.

Featured Image

Why every function needs to be trained in AI & Robotics?

1. Why AI?

Today the industry is at the inflection point where it has made the exponential shift from basic analytics to leveraging the power of deep learning, machine learning and AI. Mckinsey collated and analysed Artificial Intelligence Use Cases across industries and functions. They were trying to study what incremental value can application of advanced AI deep neural network techniques provide to the business. The answer was that in more than 2/3rd of the use cases artificial intelligence can improve performance beyond that provided by traditional analytics techniques.

Screenshot 2020 08 17 At 8.03.54 Pm

This has unleashed tremendous potential that can be mined. The incremental value that AI can provide in travel is 128%. Retail 87% and Banking 50%.

Almost 99% of all successful travel portals have AI based recommendation engines. This is very similar to Netflix or Amazon recommendations. These engines recommend to you based on your previous choices or history.


Cyber Bank Robberies or Fraud contributes to $1 Trillion in cybercrime losses. A rule-based method of fraud detection will typically take about 40+ days for a Bank to assess a fraud. That is where Machine learning and AI play a huge role. Fraud detection problems are typically constructed and analyzed as classification problems in machine learning. Based on supervised learning these models become intelligent enough two classify a claim as either an authentic & legitimate claim or a fraudulent one and are super-fast.

By learning from past experience, the algorithms become smart and can start identifying potential fraudulent transactions which have never occurred before. We must remember that hackers and fraudsters excel at find completely novel ways to beat the system every time. The fact that a Machine Learning algorithm can also beat the supervised learning by becoming smarter over time as it starts identifying hidden patterns, is a big win. Another very practical business application of AI is in claim assessments. Insurance companies like Lemonade are using AI and chatbot to give their customers a completely hassle free and personalized claim assessment service which can take as low as 3 min.


If you are filing a claim you can open their app to record a video of you talking about what happened, machine learning will convert speech to text and then match what you said with existing claims in the database to do an authenticity check. If all is green your claim will be approved within minutes, and money transferred to your account.

Like this across every industry you will find use cases for AI & Machine learning. No wonder in 2019, global private AI investment was over $70 billion, with startup investment $37 billion, M&A $34 billion, IPOs $5 billion, and minority stake $2 billion.

2. Why Robotics & Automation?

Ey Automation Will Impact Business Functions Differently.png.rendition.3840.2560

This study was conducted by E&Y in 2018 it concluded that functions like Finance, Administration Customer Service are very high on the automatability scale at around 80%. Human Resources is at 29% and Marketing and R&D at 24% and 22%.

Globally it is estimated by World Economic Forum that automation will displace 75 million jobs but generate 133 million new ones worldwide by 2022. A mammoth change in the skill sets.

3. What is the success rate for organizations adopting AI & Machine Learning?


According to Mckinsey’s Automation Survey, only 55 percent of institutions believe their automation program has been successful to date. And about 50% said that the program has been much harder to implement than they expected.

4. What does it take to run a successful AI & Robotics program in a company?

The biggest driver one should focus on is involving the employees and engaging them upfront. The entire success of your AI & Robotics program can be dependent on this.

33% of the employees fear that they will
lose their job to automation.

Source: https://www.siliconrepublic.com/careers/automation-employees-future-of-work.

There is a lot of un-certainty introduced in the work environment as soon as AI is brought in. Workers feel scared for their employability as well as they are unsure of whether they will be able to adapt to the new digitized high-tech way of working.


As a company, you have to recognise that automation poses more challenges to the workforce because of the need to upgrade skills and thus you have to shift the culture to support continual adjustments to the way people do their work. And how do you this:

Provide employees proper training of what is Automation & AI. What they can expect, and what role they can play. The question that comes up here is that what level of training should be provided, and to whom. A good analogy to draw is the six-sigma program which has 5 levels, each designed to serve a specific business need:

Belt Levels Of Lean Six Sigma Goleansixsigma.com

For the Automation and AI program as well, a business should look at how they want to segment their employees across different levels and have a very targeted training objective for each. We recommend the following model:


Executive Sponsors

Provide high level contextual understanding of AI Machine learning with emphasis on the key pitfalls to the success of the program & the role they need to play. They should be taken through a couple of case studies from industry similar to theirs. The importance of solutioning and Data Collection, labelling and mining should be highlighted

The Program Leads


She has transformation and reengineering background along with Project management and a solid understanding of AI & Robotics. She needs to be trained continually on the latest and most upcoming advances in AI machine learning and deep learning algorithms.

Project Managers


Project Managers will need to be trained on AI & Robotics concepts focusing on the solutioning. They need to have an in-depth understanding of the process and business they are supporting. They should be trained on the end to end process as well. Refresher trainings on continuous improvement and reengineering should also be organized.

Project Team Members

Project Team members will be employees from the function and the process. They are process SMEs. They need to be given an overview level training on AI & Robotics. Since they will be participating in data collection, solutioning, data analysis & testing they need to understand the development lifecycle and should be taken through both agile and waterfall methods. They should be grounded on basics of data analytics and statistics.

Process Team Members


These are people who are not directly participating in the project but are getting impacted by it. They should go through a basic level course on AI & Robotics which is more focused around the solution being built for them as a case study. Training on the change that’s about to come and how their roles will transform with what will get eliminated and what will get added should be provided. Wherever reskilling requirements have been identified the training should be conducted. How to work with Robots is also an important subject to be covered. Provide hands-on experience and live demos early in the process, clearly explaining constraints. This will involve them upfront and thus they will be more accepting of the change.

5. Transform a culture of Fear into a culture of Collaboration and Innovation

Treating employees as problem solvers can be culturally very transformational. Delegating authority over the bots to those employees (versus running the bots centrally) can also be a way to ensure continuous improvement and employee participation and instead of fearing a culture of collaboration and innovation will be created.

Amazon is a great example. It announced last year that it will spend $700 million to train about 100,000 workers in the US by 2025, in Robotics and AI.


At Amazon Fulfilment Centres It’s not humans vs. robots, it’s humans + robots. Amazon runs 175 fulfilment centres worldwide. In 26 of them, robots and people work together to pick, sort, transport, and stow packages. Carrying and transporting inventory across building for example is taken over by BOTs and understanding how best to unpack honey or maple syrup is something left to humans. Thus, establishing a good Human and Robot collaboration is the way to go!

Abraham Lincoln said “Give me six hours to chop down a tree and I will spend the first four sharpening the axe”. Good preparation means more than half the battle is won. Prepare well, invest big on your biggest assets – your employees and reap the fruits of Automation and AI.