Your company is NOT ready for AI

Cyprian Vero
The Startup
Published in
7 min readMay 28, 2019

Artificial intelligence (AI) and machine learning (ML) are the new sexy tech terms quickly becoming synonymous with innovation. Experts from all industries call it the future of technology, the true disruptor.

This much buzz and excitement makes every executive eager to jump on the AI innovation wave and ride it as fast as possible. The problem is, they don’t know anything about “surfing”! ;-)

AI is not a technology that you just plug into your company and expect it to magically push you in front of your competition. It requires large quantity of data, time for experimentation and an organization culture that embraces learning from failure.

So can you just wake up tomorrow, go to the office and say “From today we will use AI!”? The answer is NO and here is why.

Successful implementation of AI in an organization rests on the following 3 pillars. Without one of them your efforts will fail.

  1. Knowledge / Data
  2. Expertise / Talent
  3. Culture / Experimentation

Knowledge / Data

The foundation of every AI system is data. Without it, the AI can’t learn. The quantity, quality and freshness of your data will greatly impact the performance of your AI systems.

Even though the idea of collecting and using data to drive business decisions has been a hot topic for more than a decade, many companies today still lack understanding of how to use it. They simply focus on collecting as much data as possible and brand them selves as Big Data companies.

Dan Ariely a researcher and an economist famously said:

“Big Data is like teenage sex: everyone talks about it, nobody really knows how to do it, but everyone thinks that everyone else is doing it so they claim to do it too…”

It’s great if you have data, but is it centralized and accessible via API’s? Do you know why was it collected? Is it a full set or a subset that was processed and filtered?

For a Machine Learning model to work accurately it needs data that is free of capture errors, definition errors, processing errors, coverage and sampling errors. It’s not enough to have data, it needs to be correct data collected specifically for the purpose of solving your business goal.

So what should you do to get your data ready for your AI revolution?

Start with the business goal. AI is used best to automate processes that are time consuming and repetitive. Your business runs on top of multiple processes. Some of them run in parallel others in series. Find a process that is simple in scope but important enough that it greatly impacts your business performance. Then investigate what data you already have about this process, check if there is documentation describing why this data was collected and if it is a full representation of the process or a filtered subset. The more time you spend on quality of your data the better your ML models will be.

Remember that nothing is constant, your business changes and adopts to the market condition and your customer needs. Your ML models will need to adapt with those changes. It is important that your data is always accessible and fresh. If your data is a static representation of the past that is just stored in an old database your ML models will immediately become outdated and will predict the past that you already know. To avoid this issue, make sure you invest in building a centralized data and technology infrastructure that will provide you ease access to the real-time representation of your business performance.

Expertise / Talent

Although the research about AI has started in 1950’s with Alan’s Turing concept of “Turing Machine” the AI industry hasn’t yet produced many experts in that field.

In the recent “Global AI Talent Report 2019” JF Gagne a CEO and Founder of Element AI in Montreal, Canada reports that as of 2018 there are only about 22,400 qualified researchers in the field of machine learning, an increase of 19% from 2017 and roughly 36,500 people who are qualified as self-reported AI specialists on LinkedIn.

According to Reuters analysis of AI job-seeker market, the demand for AI talent grew double in the last 2 years.

“AI job postings as a percentage of overall job postings at Indeed nearly doubled in the past two years”

https://www.reuters.com/article/us-usa-economy-artificialintelligence/as-companies-embrace-ai-its-a-job-seekers-market-idUSKCN1MP10D

Most of the existing AI talent is absorbed by giant tech companies like Google, Microsoft, Amazon or IBM, and the growth rate of new talent remains small. This may however change as the same companies build new self service AI tools like TensorFlow by Google, Microsoft’s AI Platform or Watson Studio by IBM, and partner with universities and online curse providers like Udemy and Coursera to train future AI workforce. Just this month (May, 2019) Microsoft announced that it is partnering with companies such as General Assembly to upskill and reskill 15,000 workers in the field of AI by 2022.

Unless you are one of those AI experts the chance is that as of today your company does not have machine learning expertise.

Finding and attracting talent that is scarce will take time, which is another reason you can’t just immediately jump into implementing AI strategy for your company. Use this knowledge as an opportunity to train your current employees and reach out to AI consulting companies like Stradigi AI or ElementAI to help you figure out your AI implementation plan.

Culture / Experimentation

Having abundance of data and ML talent will not suffice in successfully implementing AI strategy if your company culture does not embrace experimentation and data-driven decision making.

The unfortunate reality is that more often than not business executives embrace intuition-driven decision making, solely relying on their gut filling and ignoring all available data signals simply because they do not support their vision. If your company direction is driven by a hunch and HIPPO (Highest Payed Person’s Opinion) your AI efforts are bound to fail.

In an experimentation culture decisions are made after the hypothesis is tested. It is a culture that embraces failure and rewards those that are not afraid of trying. Jeff Bezos the CEO of Amazon perfectly describes his failure philosophy in a statement to share-owners:

“One area where I think we are especially distinctive is failure. I believe we are the best place in the world to fail (we have plenty of practice!), and failure and invention are inseparable wins. To invent you have to experiment, and if you know in advance that it’s going to work, it’s not an experiment. Most large organizations embrace the idea of invention, but are not willing to suffer the string of failed experiments necessary to get there…”

If your company culture does not embrace learning by experimentation and A/B testing is not an integral part of your managers KPI’s your AI implementation will be long and difficult.

It is not only at the lower level of your organization that you will face challenges. Public companies struggle with the overarching control of the board that pushes executives to deliver quarterly results. The desire to meet those objectives drives executives to base the company strategy on the short-term initiatives that bring immediate revenue. AI on the other hand is hard to predict and needs constant optimization before it can show its positive results. If you do not have a buy-in from your board and secured budget for the R&D, AI will quickly lose it’s funding to more immediate short term wins.

Having the buy-in from the board and an executive alignment will not be enough. Your AI efforts can easily be sabotaged by your front-line employees fearing that the AI and automation will make their job obsolete and middle management fearing they will lose their bonuses if they don’t deliver on their current KPI’s.

AI strategy does not end at implementation. Your new tools will directly and indirectly effect your entire organization. You need to choose AI champions inside your business that will address your employee fears and help them transition into the new reality that is focused on decision making rather than repetitive time consuming processes.

Conclusion

The reality is that only a small percentage of companies today are ready to implement AI in their organization. Before you decide on your own implementation consider the following questions:

  • Have you identified a manual process that is simple in scope yet provides a significant benefit if automated?
  • Do you have API access to data relevant to this process?
  • Do you have documentation explaining how your data is collected?
  • Do you have a board buy-in and secured funding for AI R&D?
  • Does your organization culture embrace experimentation and data-driven decision making?

If you answered “no” to any of the above questions, make sure you address them before moving forward with your AI implementation otherwise you are less likely to succeed.

To learn more about this topic I recommend you read Applied Artificial Intelligence: A Handbook For Business Leaders by Mariya Yao, Adelyn Zhou and Marlene Jia as well as The Innovator’s Dilemma: When New Technologies Cause Great Firms to Fail by Clayton M. Christensen.

--

--