Implementing A.I.

implementing A.I.

A.I. is gaining more and more traction both with organizations and with the public. Its use has gone up 270% in four years and the global A.I. market is predicted to reach a value of 267 billion dollars by 2027. (source: Gartner 2019)

62% of customers say they are willing to use A.I. for better experiences. As users, we like the algorithm in Spotify to give us songs that match our tastes, and this extends to any other similar service that can come close to what we would choose if we had the time and inclination to make each application our own. Businesses enjoy the benefits of greater insight and the knowledge that the huge amounts of data they gather do not go to waste. Yet implementing A.I. systems is proving to be a daunting process, and in some cases unnecessarily expensive.  

The obstacles are both human and organizational. Employees fear that A.I. will steal their jobs, that they are not valued enough, that they will become superfluous. Their resistance and a project that lacks clarity can end up costing much more than initially expected. Yet, worker replacement is not the goal, rather a reimagining of the work itself. 

At the same time, A.I. is not unlike all the other technologies we have been using so far, it is just new and has the potential to revolutionize the world of business in quicker, more dramatic ways. 

To deal with these issues a clear vision is necessary and an approach that pays attention to personalizing the process to the requirements of the company. Each new instance of A.I. software development implementation will be different. The sky is the limit when it comes to how it can help, so finding the right solutions is the key to benefitting from the change. 

Ways to make A.I. implementing easier 

As humans we are defined in a big way by how sophisticated our tools are and the innumerable things they helped us achieve as a species. Human and machine have been working in tandem for a long time. When we have been honest about the things we cannot do without assistance, we have managed to come up with ways to expand our abilities and maximize the talents we do possess, compensating for the others with technology. A.I is only the newest addition to our toolbox. So, we have to learn how to function at the intersection between the abilities of the machine and those that are natural to us. For this we need new roles at work and fresh perspectives. We also need to understand that implementing A.I. is not like plug and play technology and that it can take time for things to work the way we envision them. 

Understand the tech 

Knowing what each type of A.I. does and how to leverage it for your business is the first step into a journey of effective implementation. All of them have strengths and weaknesses, various limitations and advantages. Being aware of these and how they can be used for your internal processes will maximize the gain. A good way to do this is to decide what you want to use A.I. for. Is it for automation purposes? Or maybe to gain insight? Or perhaps for engagement purposes? 

Decide if the lack of transparency in deep learning processes is something that worries you or not. Do you need to explain the algorithm’s decisions? Is the area where you use it high stakes? Do you have room to experiment, fail and try again? 

Choose the right project for implementing A.I. 

Any project where data is available to be inputted is a good place to start. Find out if A.I. is going to bring value to the domain you want to use it in. Analyze your current processes to find out if the effort to introduce it is worthwhile. Do you have processes with a great number of steps? Does the process experience delays? Are there data fluctuations, or do you deal with missed opportunities, like in a logistics and scheduling scenario? Could A.I. bring substantial benefits?

Initially, you should choose a smaller project so the investment can pay off sooner. Starting slow is a good idea. Test as much as you can to make sure the technology is doing what you expect it to do. Keeping things low scale and not doing a huge initial leap might be a more fruitful approach. Starting small pays off in many ways. On top of all that, it is crucial to make sure that you use both machine and human resources at their maximum output.  

Reimagine how you work 

In implementing A.I. do not expect to be able to attach it to old, rigid organizational structures and processes. The whole area needs to be redesigned and you must redefine its purpose if necessary. A new paradigm must be at the center of the novel process you are creating, one that will integrate A.I. seamlessly and that can be hopefully reused. If data you use in this project overlaps with data in other parts of your activity, then you will be able to implement the new way of doing things much more easily because you have already gone through all the testing and hardship the first time. 

Map out an ideal process 

What you need to remember when it comes to such an endeavour is that the better you understand what you are doing the better you can mitigate the challenges. Start from first principles, or design thinking and go backwards to reach the root cause of the issues you are trying to address. Break down the problem into its smallest parts. This will help you to thoroughly understand it and the data that you need to reinvent how it works. It will also tell you what type of A.I. you need and what data science techniques can solve the issues. 

Because A.I. requires a high level of precision it is essential to specify all business questions as precisely and accurately as possible. Also do not forget to include any other business requirements. 

Define the business expectations so you have an idea what to measure and how to achieve success. 

Measure 

In implementing A.I. it is very important to be able to compare your old process with a new one, and see if there are differences.  Defining a set of benchmarks to reach along the way makes the process less difficult because you get to see progress as you go. 

Organizational changes 

The introduction of A.I. systems means the whole department will need to work in newer ways that fit with the demand of the technology you have chosen. Mindsets and procedures have to adapt as well. Key roles must be well defined and there are no siloed departments.  

The people 

A.I. is a form of human augmentation. So, one of the biggest and most important resources you have to gather are specialists and people that can work well together across fields to achieve your business goals. The system does not come already functional, it needs to be molded to your processes and procedures. This means a number of iterations and of failures. It also means a degree of risk and of perseverance. A team that can bring results is one that contains translators, working at the intersection between business and tech, data architects, engineers, scientist, visualizations specialists and business analysts. Such a team is well equipped to address the challenges of implementing A.I. as they arise and to exploit all situations to bring value and improve the system. 

People need to be comfortable with the new system and learn to trust its predictions. One way to do that is to reskill your employees both to work well in interdisciplinary teams and with the technology. Another way is to encourage their use of it for a period of time. Reward the ones who are open and willing to work with the A.I. versus the ones who turn to old methods when things are not clear. To make people comfortable working with it they need reskilling to both function well in interdisciplinary teams and use the new technology.  

The organizations that succeed are the ones where leaders in IT, analytics and business cooperate well. These people are also able to motivate and support their teams throughout the whole process. 

Invest as you go along 

In general, implementing A.I. does not need you to completely update your tech. In some cases, you don’t need any investment in this direction. That will come as you progress and expand your machine learning systems. Make an inventory of all the tools that you have and those that you require and start somewhere, then work your way through the list. Going all at once might be burdensome financially and detract you from focusing on the important aspects of the digital transformation. 

Conclusion 

Implementing A.I. is akin to taking digital transformation seriously. The organizations that succeed are the ones able to build interdisciplinary teams, the ones that are ready to reinvent themselves and are not afraid of change. Realistic targets are essential since this process is bumpy and full of stops initially. If you do not expect immediate success, but look at this as a way to safeguard your organization in the face of future, unexpected changes, by making it more resilient, success is already at your fingertips. 

  

  

Leave a Reply

Your email address will not be published.