Tel: (01) 901 1310

menu

AI For Business: Myths And Realities

In short, while there is a lot of hype around AI, the steps towards being successful with it in your own business should not start with the hype. This article covers common myths and realities about AI in 2021 & how you & your business should approach it.

Share Post:

Share on linkedin
Share on twitter
Share on facebook
Share on email
AI

Source: Forbes.com /Nisha Talagala

AI or Artificial Intelligence perhaps be a word that you feel may  not apply to your Business – this article breaks down the myths and realities of AI in a very clear and understandable way. 

The global Artificial Intelligence market is projected to grow to $169 Billion by 2025. Augmented Intelligence, the practice of using AI and human skills together, has even loftier projections, ranging in the single digit trillions as specified by Gartner. Yet, AI remains a mystery for many businesses. The news feeds us a steady diet of amazing AI accomplishments. The AI tools landscape is growing rapidly, with hundreds of new tools available every year. However, for a business with no internal AI practice, there is little information about how to get started and how to generate the first dollar of Return on Investment (ROI) from an AI project. Between the hype and the frustration, what can a business leader do?

This article covers common myths and realities about AI in 2021. By noting these, businesses can approach their entry to AI with an optimistic view driven by the technical prowess, but a realistic approach that increases the chances of success.

Real vs. Relevant

Many (even most) of the amazing AI innovations found in the news are real. But being real is not the same as being relevant (to you!). Some examples:

  • AI beat humans at chess many years ago. This speaks in general to our ability to create AI technologies that can reason and strategize. Beyond this, unless you are in the business of internationally ranked competitive chess, this is probably not relevant.
  • AI has been shown to be extremely effective at detecting diabetic retinopathy from eye scans. If you are in the healthcare field, this is certainly interesting. However, given the regulatory requirements of new technologies, unless you are a researcher, it is unlikely to be imminently relevant to you.
  • AI can be used to make ever more intelligent and engaging chatbots. These capabilities are frequently accessible via APIs. This is relevant to anyone with a business website, shopping site etc.

Your Reality

Half the battle of being able to get value from AI is to understand your problem and how to define your problem in a way that an AI can be effectively applied to. To help illustrate this – here are some examples:

  • I want to improve sales: This is too vague for AI to be effectively applied.
  • I want to improve customer retention: This is better in that a particular approach to improving sales has been identified.
  • I want to identify all customers who are likely to leave in the next 3 months. This is good. You are now starting to narrow down on exactly what you would like the AI to do.
  • I have 10 pieces of information about each customer and 2 years of historical data on past customers. I want to use it to predict whether an existing customer will leave in the next 3 months. This is great since the problem now specifies what the AI will need to do and what information it will use to learn from.

Once you have found a problem whose solution can benefit your business, you still need to navigate through the myriad of options, some of which have a lot of noise and hype associated with them. Below are some of the common myths and realities.

I need to use the most advanced AI tech

Verdict: Myth

You should expect that your AI will need to iterate and will improve with each iteration. As such, getting the first one working as quickly as possible is a great step towards success. No matter how much effort you put into your first AI, it is unlikely to be your last one.

Finding as simple (and inexpensive) an approach as possible to get to a first result will give you a great experience in how your AI interacts with your problem. Time will only increase the variables. Your business may change, AI will change, your team may get frustrated, etc. Get the first return as soon as you can to build confidence and keep going.

I need to hire a PhD Data Scientist

Verdict: Possible Myth

For many AI roles, a PhD data scientist is not required. The role of a data scientist is to help bridge the gap between the data and suitable AI techniques that can solve the business problem. Depending on what the problem is, a PhD may well come in handy. Other skills, however, can also make the difference between success and failure. Can the data scientist understand and work within the practical constraints of the business and environment? Are they able to collaborate well with the engineers, product managers and other team members that will also help generate ROI from the AI innovation? A combination of these skills is critical.

I must build a custom AI algorithm

Verdict: Myth

As AI becomes more pervasive, there are a range of offerings from fully functional APIs to no-code and auto-ML tools. If your problem is generic, you may be able to buy an AI solution as an API rather than build your own. Good examples of problems generic enough to be solved via APIs include Speech to Text translation, Language Translation, OCR document readers, Chatbots, etc. A good indicator that your problem is custom is that you have a unique dataset that is private to you. Even in this case, you do not need to build an algorithm from scratch. No-code and low-code tools can help auto-analyze the data and select a good candidate AI solution using Auto-ML.

Building the AI algorithm is the hardest part

Verdict: Myth

As more AIs find their way out of the lab and into production, and MLOps becomes a mainstay of enterprise AI solutions, we are starting to appreciate that the first satisfactory AI prototype, while critical, is just the beginning of the journey. Putting the AI in production, managing and monitoring it in production, and improving it via iteration is what ultimately leads to business success.

It takes a team to build a successful AI

Verdict: Reality

A successful AI lifecycle (see figure below) includes understanding how AI can be used to solve the business problem, finding the right and relevant data, experimenting with AI solutions, putting the chosen solution into production, connecting it to the business, managing the solution, and continuous improvement. To successfully execute this lifecycle, you will require a team with skills that range from product management, to data science, engineering, and operations.

The lifecycle of a full AI, from problem to data, to model development and testing, to deployment, business connection, monitoring and continuous improvement.

Hopefully the above tidbits have convinced you that being successful with AI is a journey and a practice and not a one time action. The best way to ensure not just your first success but a series of AI successes, is to instill best practices within your organization, particularly when it comes to AI and Data.

  • Ensure a practice of Data Literacy: AI thrives on data. The more you can collect, protect, organize and manage access to your data, the more likely that, when the time comes, the information you need for a new AI is available to your data scientists and AI teams.
  • Ensure a practice of AI Literacy: As AI becomes more pervasive, it is not just data scientists that need to understand AI. As you evaluate build vs. buy options, put AIs into production, and train your customer service and support teams to manage AI features, people across your organization will need a basic understanding of what AI is. They will need to know what its strengths and limitations are, how to interact with it, and how it applies to your business. Creating an organization wide AI literacy practice will prepare your workforce for business success with AI.

In short, while there is a lot of hype around AI, the steps towards being successful with it in your own business should not start with the hype. It should start with the problem you want to solve, a measurable goal and success criteria, the data, the simplest approach possible to get started, and a team that will learn along the journey.

Quick Contact - Or Call: 01 9011310

Sign Up To Our Latest Transmission

To be sent automatic updates to our latest news articles, please fill in your details.

More News

Leadership

Do You Feel Stuck Trying To Grow Your Company?

Good leadership is vital for organizational success, but even good leadership can be an obstacle. Successful leaders often struggle with a few common issues: difficulty trusting others, identity and worth and self-discipline.

Warehouse Management

12 Tips for Keeping a New Warehouse Organized

If you want to serve your customers quickly and efficiently, you’ll need a good system of organization in place. But with potentially thousands of items to keep track of, dozens of people to hire, entire systems to build from the ground up, how are you supposed to do it?

The Business Troubleshooters Ltd.

Dublin

Carlow

Cork

Ⓒ 2020 - All Rights Are Reserved

Sign Up To Our Latest Transmission

Dublin

3013 Lake Dr,
Citywest Business Campus,
Dublin, D24 PPT3
Tel: (01) 901 1310

Carlow

Enterprise House,
O’Brien Road,
Co. Carlow, R93 YOY3
Tel: (059) 910 0440

Cork

Acorn Business Centre,
Mahon Industrial Estate
Blackrock,
Cork, T12 K7CV
Tel: (021) 2021130

Share on linkedin