Executive Guide to getting started with AI

Innovate or die! While not the official mission statement for business, there are enough leaders who abide by this mantra. The belief that without innovation, business will stagnate, and leaders will lose their position in the market is popular on both side of the equator. The concept makes sense, with technology, synonymous with innovation, we can do all sorts of things like grow the business, reduce costs, improve employee engagement and grow customer satisfaction. When you put it like that, who wouldn’t want to adopt new and exciting technology?

 When it comes to technology that can change the world and if the hype is to be believed, can solve any problem we throw at is, very little compares with Artificial Intelligence (AI).

 One approach to adopting new technology is to deploy the technology, hire smart people and then figure out what problems to solve. Reminds me of the saying " when all you have is a hammer, every problem looks like a nail". While a common approach - I would like to suggest an alternative. Before hiring the smart people, before buying and deploying the technology - focus on the problem or problems to be solved. This enables business to prioritize their problems, according to impact, and once that is done, identify the tools, technologies and people needed to make it happen over the coming months and years.

 The purpose of this article is to show business leaders how to get started with Artificial intelligence.

 Before we get started, we need to first understand:

• What is AI?
• Why the hype around AI?
• Realities of AI

What is AI?

If a computer can think and learn by itself without a human interaction, we classify that as Artificial intelligence. To be effective, AI needs the following pieces:

• Access to vast amounts of data
• Computing Power
• Expertise
• Time to process

 Within AI, there are two categories to be aware of as they solve problems differently.

 Machine Learning (ML):

ML process vast amounts of "practice" data looking for patterns and makes predictions on what happens next. ML remembers what it did previously on "practice" data and uses previous experiences to learn.

 For informational purposes, ML has 3 models to process data

• Supervised learning
• Unsupervised learning
• Reinforcement learning

Deep Learning (DL):

DL is a subset of ML and acts more like a human brain when it comes to processing data as it has a multi-layered "brain" called a neural network and is best suited for solving complex problems like image and language recognition.

Why the hype around AI?

While the idea of AI has been around in one form or another for the last century, adoption has been slow. The reason is, up until recently, the hardware and data needed to build an effective AI were not available or too expensive to be practical. Advances in computing power, GPU, cloud computing and the amount of data accessible on the internet means building an AI is now a reality.

AI is seen as cornerstone of the 4th industrial revolution and it is the first time that academic and business worlds have been so aligned about the developments of AI.

AI has the potential to greatly improve business as it delivers personalized services and products directly to their customers. While the AI movement is only getting started, here are a few examples of what AI can do in business.

•  Improving efficiencies in repetitive tasks with no human error, freeing up employees to focus on tasks that require more creativity or thought.
• Analyzing huge amounts of data looking for patterns or anomalies
• The ability to work 24/7/365
• Fraud detection for financial institutes by analyzing vast amounts of data and looking for anomalies that we as humans would miss.
• Cyber Security to prevent attacks on computers that we are not yet aware of buy understanding how these attacks work and their end goal
• Analyzing market data and predicting future trends in the stock market to make trades in milli-seconds.
• In Healthcare, AI could go through historical treatments, analyze the medications being used and how it may impact future prescriptions before causing problems and can detect cancers from images.
• Creating personalized customer experiences based on previous shopping habits and their online activities.

The items above exclude the soft benefits of what AI could address in the workforce:

• Burnout
• With more of the workforce experiencing burnout, AI could be used to take over certain tasks to prevent or detect burnout early enough to warn the employee and prescribe a solution.
• Employee engagement
• Employee engagement continues to be a challenge for business and with AI, we may be able to automate certain tasks and re-purpose the employee to tasks better suited to their skills.
• Reskilling workforce
• While some jobs will disappear with AI, others will be created, and business needs to focus on programs to re-skill their workforce for an AI enabled future. AI could monitor employee workstyles and strengths and suggest new roles or responsibilities based on previous projects and feedback.

This is what we are aware of today, once AI starts rolling out, the landscape will change, and new tasks will be added.

Realities of AI

While AI has made significant progress over the last few decades thanks to advancements in technologies, there is a lot AI cant do which makes the argument for a workforce compromised of Humans and AI to better address the shortcoming in both as we tackle problems we have never imagined before.

A few examples of AI limitations:

While AI may be able to tell us the best route to the office and how to avoid traffic jams, that same AI is unable to tell the time, the weather or calculate 2+2. This is because an AI can complete specific tasks only, known as Artificial Narrow Intelligence. It was trained to do a task and it does that task perfectly. It cannot do anything else and it can’t tell you why it is doing it.

While this is expected to change over time, the arguments over how far away we are from Artificial General Intelligence are a topic for another day, the key message here is that AI is great for certain tasks - not all.

Getting started

To begin, identify the problems at the company. The goal is to write them down, not think about solving them or the benefits of solving them. For now, write them down.

 With the list completed, next to each problem assign a ranking using the following numbers.

• Small or 1
• Medium or 4
• Large or 10

The numbers mean that at a glance it is easy to identify higher priority problems versus the rest without having to wonder about the difference between a 3 and a 4. with the problem identified, assign a score to the problem in one of the following categories

The impact of solving this problem for

• Customers
• Partners
• Employees
• Revenue (saved or earned)

The time needed to solve this problem

• Short amount of time
• Longer amount of time

Difficulty to solve the problem

• Is this an easy problem to solve
• Is this a complicated problem to solve

Skills needed to solve the problem

• Do you have them
• Do you need to hire / find them?

An example here might be reducing the number of helpdesk calls.

• Impact Score: 4 (Customer), 1 (partner), 4 (employees) & 4 (revenue)
• Time Score: 4
• Difficulty Score: 4
• Skills Score: 1
• Total Score: 22

Now compare that to the other problems on your list to see how long or difficult this one is compared to the others. Problems hardly ever occur in isolation, meaning solving one problem may not yield the returns we want. The next step is to look at the problems listed and try to connect them to others, creating a workflow of events that need to happen before the problem is solved. Think about it this way, if this problem is solved, will it improve the experience enough to be noticed or will it surface another problem that needs to be addressed? 

 Step back and review the problems and workflows listed, while it's easy to focus on the problems that will have the most impact, don’t. instead look at the easiest problem that will take the least amount of time and resources to complete. The point is the team is doing something new, they have probably not worked together, and things are going to go wrong the first few times. Best to sort out these issues on something small versus a project that has the company watching and talking. If picking the easiest problem is not an option, try picking the easiest problem within one of the workflows. This way the team, the tech and the process are tested on an easier problem while building momentum to tackle a larger one in time.

One thing to remember when introducing change, is the culture of the company. When tackling something new, something outside of a company’s comfort zone, is the pressure to get it right the first time. When using AI to tackle problems, this may not work as expected. The goal is to take calculated risks and to not be afraid of failure. The team are going to be doing something new and failure is part of the process. Think back to the expression " if you are not failing, you are not trying hard enough"

To succeed in the digital age, this fear of failure needs to be squashed and a culture, albeit on a smaller level, to experiment with new and disruptive technologies and thinking - introduced. Without it, AI will never be the disruptive force it can be, because the company will not be pushing the boundaries. Instead it will be following where others have gone before and be a follower of best practices which is where disruption and innovation go to die. To address this, get support of the leadership team, they need to be the key stakeholders to make this project succeed. This approach increases the chances of success as employees understand the importance and see where the project fits in the company vision. Things will get difficult and when it does, it is important for the team to understand that failure is part of the process and they will not be punished. 

Projects around AI are scary to most as employees believe AI will replace them. If that is the belief or rumor employees are hearing, how motivated would they be to see this project succeed? It would be in their best interests to sabotage this project and see it fail. Address the rumors around automation leading to job replacements and people being laid off. Share where automation will be used, how it will be used and offer rough guidelines on when this will take place. Explain why this is important to both the company and employees, talk about improving employee engagement and creating a more productive environment where their skills are better utilized, and they are not running on autopilot.

The HR team should be thinking about reskilling projects and how AI can be used to free up employees to attend the training. If this is not something being thought about, it should be. Are employees a part of the conversation to identify the skills or training needed? They should be, so they can see the challenges being addressed and why automation will be part of the new normal as well as what happens to the company (and employees) if they don’t take this step

Technology - what about the Tech?

With the problem highlighted, the stakeholders identified and employees excited, now we think about the technology & skills needed to make it happen. The problem being addressed could be anything and to make a recommendation based on an unknown problem is not advised, so instead, here are a few questions to ask to when deciding on the tech needed:

• Should you build your own infrastructure or use one of the existing platforms that are up and running already?
• Amazon Web Services or Azure are the two most popular
• What are the benefits of building your own versus existing platform?
• Do you have the skills to manage the infrastructure needed?
• Do you have the skills, or do you need to hire them to build your AI?
• Do you have the data needed to run your algorithms?
• Is the data ready, has it been cleaned?
• Regarding your AI, will you use Machine or Deep Learning?
• Is RPA an easier fit for the problem?
• Is this an once off project or could this be a series of projects you should plan for in long term?
• How will you determine the ROI and over what period?
• What methodologies should you use to manage the team?
• Waterfall or Agile?
• Are there any regulations that you should be aware of regarding compliance or privacy laws?

Next Steps

If you don’t have the skills or time to build this, then I would suggest hiring a company with experience and leverage one of the existing platforms to get up and running. After the project completes, it is important to review what worked well, what didn’t and what could be improved on to address the next problem. As part of this review, include the realized benefits of addressing the problem to improve your estimations before you tackle the next problem.

You will not have all the answers to the questions, and you will fail with some tasks. That is part of the process and everyone will stumble before the find their stride. The important thing to remember, is that this is unchartered waters and the journey is best taken with those you trust to help you when things get rough.

This guide is meant to get you started with your AI implementations and share an overview of what AI can do so that you are better informed how it can address the problems you are dealing with.

Best of luck and if you need assistance - dont be a stranger, ask.