More than just a trend or buzz words, Artificial Intelligence (AI) and Robotic have become a true opportunity companies must embrace and leverage to strive. But introducing AI and/or Robotic in the organization is not just a technical exercise, the impact on the human workforce can equally cause failure as the technical complexity. Should a company want to enjoy the benefits coming along with this next generation of tools, it must implement a strong change management strategy. It all starts with the most important question: “What are the benefits for the employees?”.
Where does the “Rise of the robots” come from?
Let us start with a little bit of history to understand why AI and Robotic are more than a short term trend soon to be outdated. Understanding the past is also wise if we want to reflect on the future.
One of the first references to “intelligence” produced by a non-living being can be found in the 50s with programs able to demonstrate mathematical theorems. At the time they were very much limited by the computing power. Moore’s law would tell us later that this limitation was temporary but would take decades to be lifted. In the late 70s, early 80s, progresses in algorithmics proved that AI was useful to mankind. Programs were able to solve simple practical issues. For example, in 1980, the Xcon system allowed the optimization of computer configuration automatically, without human intervention. But these first ‘expert systems’ were not flexible and very expensive to maintain. Towards the end of the 80s, budgets allocated to AI were cut and researches in this domain reduced, these days have been called afterward the “AI winter”. In the 90s, AI slowly crawled its way into human society through applications in logistics and industrial robotic. In 1997, a milestone was reached showing that a human reasoning could be mimicked and even exceeded : Kasparov, undisputed chess champion, was beaten by the “Deep Blue” computer. Starting in the years 2000, the computing power had reached sufficient levels to allow much more complex applications, machines started to be able to adapt to unknown environments. For example, in 2005, a robot built by Stanford was able to drive autonomously in a desert for 131 miles without having previously analyzed the path to be followed.
By 2010, organizations had realized they had access to massive amounts of data and that the computing power was sufficient to develop machine learning techniques (ability for a program, or machine, to autonomously adapt its algorithms to the data collected) later refined with deep learning (allowing to define analysis criteria and variables beyond human’s ability to do so). Computing power, data availability and storage no longer being constraints, applications began to multiply. Let us choose three of them as illustrations:Robo-advisors : developed for the financial sector, the investor provides information on her/his risk profile and the expected return. The robot then selects the best financial product matching the preferences and constraints. This “first level of support” can help financial advisors focusing their added value on more complex structuring requiring their expertise and spending more time with their customer. Chatbots : these programs allow customers to discuss (or “chat”) with a machine providing answers in real time. One of the key challenges is in the capture of the intent: how to match the sentence provided by the customer with the question (with a different phrasing) configured within the chatbot. Applications can be found in multiple domains such as IT support, selection of a restaurant, HR recruitment process, etc.
Robotic Process Automation (RPA) : The first levels of RPA do not necessarily involve Artificial Intelligence. These programs can use graphical interfaces designed for human interactions, take decisions and automatically perform tasks normally processed by humans. They are available for deployment nowadays as they have become mature, reliable and scalable.
 Moore’s law : Empirical law stating that the transistor density (number of transistors placed on a given surface) doubles every two years ; the computing power follows the same logic. The growth is therefore exponential.
 It is actually the second AI winter, lasting from 1987 until 1993. The first one took place in the mid-70s.
Implementing AI/RPA, not so easy after all
Today, most organizations are aware of the existence of AI/RPA, and a good majority of them is convinced that these tools should be incorporated in their strategy to stay ahead of the game… or to simply avoid becoming irrelevant. Being convinced that implementing AI/RPA is a must and being able to do so are two separate things. Even if technologies are mature, experience on the field tells us that implementing them is not straightforward.
One common mistake we see in AI/RPA implementations is what we call the “All or Nothing” approach. Organizations want to go big and to go fast. Once the preliminary “Proof of Concept” is done, the implementation begins full throttle without taking the time to understand the impacts on the daily activities these new tools bring. The lack of preparation results in an inadequate understanding of where the added value can be rolled out. It all eventually creates confusion and frustration.
A second type of issue is the lack of enthusiasm in the various departments. It can come from the management, underestimating the cultural dimension of the project and not sufficiently involving employees in the implementation. Key individuals are therefore not providing insights on where to improve and keep valuable ideas for themselves. But most of the times the lack of enthusiasm comes directly from the employees taking a passive behavior, observing without wanting to understand the potential benefits to their daily work. Most project managers agree that any transformation project is doomed without a proper change management approach. This is even more true in the context of AI/RPA where humans feel threatened by this new “competition” they see as ready to steal their jobs!
The key to success : placing the human at the center of the approach
Only focusing on what we want to implement is not sufficient, how it is implemented is key. We are talking about technologies that are commonly perceived as threats to the human employees, so it is essential to address right from the start the human aspect of these projects. Organizations should not impose AI/RPA but rather dedicate some time to create adoption. “Easier said than done” some would be tempted to say. Well it is not so hard if you consider some simple principles that we will now review.
Let us begin with the first one: Optimization vs. Augmentation. AI/RPA projects are often conceived as Optimization initiatives, looking at the way tasks are being produced and searching for ways to replace as many as possible with automation. This approach will typically create fear and barrier against change as employees are being directly challenged by AI/RPA. Looking at them through the Augmentation lens can achieve the same results with much less negativity. By “Augmentation” we mean trying to understand, in existing processes, which tasks allow the employees to bring value and which ones are to be considered as burdens, wasting their time. The latter should definitely be automated to allow employees doing more of the first. With this approach AI/RPA allow a same population of employees to create more value for the organization.
The second principle is to always ask the same question, over and over: “What is the benefit for the employees ?”. If the communication around the AI/RPA project does not answer this question, protective reflexes will come back and employees will feel threatened, leading to resistance to change and eventually to failure. For this reason, we encourage our clients to start their implementation journey with a “Proof of Value” rather than a “Proof of Concept”. The true challenge is to understand how the AI/RPA implementation is bringing value to the organization rather than making sure that it is technically achievable to implement it.
The third principle we recommend is to go “step-by-step”. There is no obligation to transform the whole organization in one go. Increasing gradually the penetration of AI/RPA will ease its adoption and raise the maturity level of the human workforce. People have a natural tendency to fear and reject what they do not know or understand. Let them see for themselves the benefits of the new tools and barriers will decrease. Doing so will also make the identification of new use cases much easier as employees understand where AI/RPA could help them.
The fourth principle is to promote a collaborative approach between departments. AI/RPA will deeply impact the way the whole organization operates. Thinking that it can be limited to a given part of the organization will lead to difficulties. AI/RPA should not be imposed by top management, should not be fully rolled out by IT, nor be owned only by the business operation teams. All parties must be adequately involved and represented in the project to make it a success.
We will finish with a fifth principle: articulate a change management strategy from the beginning, giving a clear focus on communication and training. Ignorance is the worst enemy of AI/RPA implementation, organizations have to counteract with a strong information policy.
Implementing AI/RPA the Consulting 4.0 way
Sia Partners is pioneer in the concept of “Consulting 4.0” and strongly believes that consulting firms will also deeply be impacted by the transformations brought by AI/RPA. To remain relevant, we must be capable to embrace AI/RPA ourselves, go through the transformation within our culture and leverage the potential of these tools to deliver our consulting services. We must be pioneer in the transformation of our sector, this is one of the key concepts of Consulting 4.0.
Applying the first principle we just brought forward, Sia Partners has created the concept of “Augmented Consultant”. We empower our consultants with bots to avoid wasting their time on low added value tasks. Data collection for example was among the first tasks to be robotized. Deep learning and Natural Language Processing allow us to build knowledge bases much more effectively and rapidly. Robo-advisors are being developed to address specific and recurrent client needs.
Implementing AI/RPA, whatever the industry or company, requires to position the human factor in the center of the approach. The key to success is to think about the added value and not to be obsessed with the technical feasibility and the cost reduction. An AI/RPA project about to be kicked-off should articulate its deployment strategy on this principle. A project already started should take a step back and reflect on the positioning of the human factor in the way the project is executed. This could be the difference between success and failure.
Publié le 09 août 2018