I don’t know about you, but as much as I am fascinated by Artificial Intelligence (AI), I am hard pressed to give you a definition of what it is, exactly.
I can throw out an alphabet soup of acronyms and jargon with the best of ‘em: IoT, BI, RPA, IVA, AR, VR, E-I-E-I-O (not to leave you in suspense: Internet of Things, Business Intelligence, Robotic Process Transformation, Intelligent Virtual Agents, Augmented Reality, Virtual Reality, Old MacDonald Had a Farm).
But what do all of these acronyms really mean? If you asked me to give you a meaningful one or two sentence definition of Artificial Intelligence, I would struggle. Could you do it? I don’t think it’s just me. I tried Googling AI explanations and just came away more confused.
More specifically, when talking about AI and business, what does AI look like at work today? What is being sold to your company? What is the impact on people and what are the current adoption challenges? I was having trouble forming this picture clearly in my mind.
Enter Peter Goedegebuure, an old colleague of mine who has been in the high-tech and new-tech field forever, it seems. Both of us being more interested in the practical and immediate, we spoke about current commercial applications of AI, as opposed to all the exciting, promising technology still to come.
And he helped me form a clearer picture of AI is within 5 minutes. And so I thought I would share it with you in this blog.
It turns out that part of the reason that the term AI is confusing is because it is an umbrella term that covers many different technologies. According to Peter, the current AI market consists of just three key technologies: Process Automation; Business Intelligence; and Intelligent Virtual Agents.
Process Automation
This is also known as Robotic Process Automation, or RPA. When we think of robots we typically form an image of a mechanical worker on a manufacturing assembly line. And that is accurate, but it is incomplete. The assembly line robots are now supplemented by software robots that perform many of the routine tasks related to information flow.
Because of that, these software robots are found everywhere, and not just in a manufacturing environment. Many processes in Finance, IT and HR, among the various corporate functions, are now in the process of being automated through the use of software robots.
For example, need invoices created? You send a spreadsheet via email to get this done. But instead of a person processing this, a robot can open the email attachment, scan the fields, populate the invoice, request additional information or a correction to information received, and ultimately create and invoice for approval and sending to a client.
Implementation and adoption challenges
The promise of RPA is huge. It is one of the main reasons that we have seen a reduction in offshoring and outsourcing. There has been a significant increase in capability of robots to perform series of routine, rules-based tasks.
However, implementations have not been all smooth, nor should you expect them to be. In some ways implementing RPA has the same challenges as implementing an ERP solution, with a few differences that give an RPA implementation a special flavour. Here are the main reasons why a company might face additional adoption challenges with RPA.
Threat of job loss/redundancy
Robots replacing people at work and taking over has been the topic of sci-fi thriller movies for a long time. Very few of these movies paint a happy scenario for us humans, and that is the dread that workers feel as well. And that is going to continue to be a major cause for resistance impacting adoption. To lower this sense of dread, be prepared to communicate, communicate, communicate.
Adjusting to new processes
Implementing new processes has been happening in a significant way since Michael Hammer’s “Process Reengineering” movement in the early 1990’s. But what is different now is that the “people” operating the new processes are robots following strict rules. Want to negotiate a minor exception to that invoice for a special client? Good luck. Preventing major issues here will require detailed use cases and lots od training.
Amount of work required to implement
When working with automatons, you have to incredibly specific and detailed. You can tell an employee, “You know what I mean, go and figure it out!” Well, with robots, as with computers, that’s not possible, at least not yet. The sheer amount of effort required to develop detailed and defined rules for robots to follow is frequently vastly underestimated.
And ultimately it tends to fall on people who know the current processes really well (read: your most valuable and busiest employees). And their task is to define rules that may lead to their own redundancy (see first challenge above). Backfilling, and having a retention strategy for your most valuable employees, can go a long way to help here.
Amount of work required to maintain
Things change, and that means rules need to be updated now and then. Who will do that going forward? Software needs upgrades and fixes. Who will do that? Both of these usually require dedicated, and sometimes additional, resources. This requires thoughtful role definition through job and organizational design.
Plus, with assembly line robots that handle physical tasks, let’s remember these are highly advanced, sophisticated pieces of machinery. It’s like having a bunch of Ferraris in your garage to drive around. Something broken down? Get the platinum credit card out.
Business Intelligence
Business Intelligence, or BI, and also known as Big Data, refers to processing large reams of data and making business sense out of connecting dots in a new way. When done well, it’s a new version of gold mining. Extensive volumes of data hide important clues that can help a company do better. A system that can process this large and complex volume of data can literally identify strategic solutions that can be worth the weight of gold for a company.
Implementation and Adoption Challenges
An earlier blog covered one of these technologies, Tableau, that is revolutionizing the way data is used to guide decisions. Here is the link. And as mentioned in that blog, as exciting and promising as that technology is, it too has its implementation challenges.
IT Department Staff
Members of a company’s IT department tend to be naturally hesitant in these situations. They have very reasonable concerns regarding external consultants from the technology vendor, such as: whether the consultants have another, hidden, agenda; what the consultants will do with their data; and whether the consultants will burden them with additional work to maintain the technology afterwards.
In addition, there is the personal uncertainty whether they themselves, used to working with older IT systems and apps, can move on to one of these new technologies. Building engagement with the IT staff is crucial here, and that takes effort and empathy.
Busy Executives
Business Intelligence is meant to provide decision support for management. That means that in these projects, management needs to be even more engaged and leading the charge than in most other types of projects. Unfortunately, time with management, especially senior management, is at a premium. So, getting these busy and often distracted decision makers to focus on reviews and approvals can be a challenge. Leadership engagement activities will help here.
Data Accuracy and Availability
Most companies run on the assumption that the data they are using to make business decisions is current and accurate. It often takes a project such as this to burst that bubble, resulting in more work to clean up data than was initially envisioned. That often translates into higher project costs and longer timelines
It also results in additional stress on already busy individuals, those who really know their business and understand what the data variations mean. Again, backfilling and having a retention strategy is an important part of yoour change management strategy here.
Data Maintenance
Once the data is cleaned up, it has to be maintained. That means having dedicated resources who focus on keeping particular data elements updated, as opposed to allowing several individuals to make changes. Ultimately this results in companies needing to establish new roles through job design, as well as new processes to approve and maintain changes to data.
Intelligent Virtual Agents (IVA)
Depending on your perception, this is either new technology, or technology we are familiar with that is evolving into the next thing.
All of us are familiar with automated response systems when we dial a number (and many of us wish we weren’t – if you agree with statement, press 1; if you don’t, press 2; if you would like to speak to a real person, press 0 or hold the line. I am sorry you are having problems. Please call again later. Click. Bzzzzzzzzzzzz.)
From an evolutionary perspective, Intelligent Virtual Agents are automated response systems that can actually adapt to a customer while responding to their questions. This is significant because it allows going beyond answering simple, rules-based queries. For example, it allows handling of much more complex interactions such as sales and marketing, payment collections and other functions.
You have likely already experienced talking to this kind of machine, sometimes also called “chatbots” (although, as with all else high tech, there is disagreement whether these two terms are exactly the same thing).
One of my favorite uses for this technology is RoboKiller, which is anti-robocall application you can use to stop the growing number of spam calls to your cellphone. A particularly delicious feature is that they are able to keep scam callers on the line as long as possible. And thereby they waste their time and prevent them from calling other potential victims.
Implementation and Adoption Challenges
Of the three AI technologies in this blog, this is the newest. Because they are still so new many companies are cautious when it comes to buying them, which is normal – that is the definition of having a product early in the sales maturity cycle. And the reason for the reluctance is that managers don’t understand how easily their employees will adopt the new technology. Without knowing that it is impossible to know how much implementing IVA will really cost.
In addition, the company may be contemplating IVA for one of its client-facing processes, and that always increases risk exponentially.
But if we look at just the internal user adoption challenges, what are they for this AI technology?
Fear
Because this technology is still so new, the number one challenge, more than in the other two AI examples, is fear. And this is not just from front line staff. Remember, these are “intelligent” technologies, meaning that if you are a middle manager, or even higher, you could be vulnerable too. You just don’t know. And being human, what you don’t know is infinitely scarier than what you do know, even if what you know is very, very bad. To overcome this fear be prepared to overcommunicate.
But let’s look at some of the companies that have implemented these chatbots. What implementation challenges did they face? And what can a company do to overcome these?
Integration in Current Processes
Since Intelligent Virtual Assistants are “intelligent”, it requires looking at them more like you are hiring a person, rather than putting in some IT. What will this new IVA person be doing exactly? Does that make sense considering how it is done today? What specifically will they not be doing? Defining clear workflows and use cases becomes paramount, as do job design activities. It has to make sense from a people and business perspective.
Training
Not your employees (although you will need to do that too), but your Virtual Agent! Yes, these are intelligent machines who can learn. If they can learn, well, someone has to teach them, at least something, initially. A chatbot should know something about the business context and their audience. As part of training, you will also need to have a person backing them up. Someone needs to jump in to course correct if they begin to go off in the wrong direction with a human user. Finally, as things change, you will need someone to provide remedial training to keep your Virtual Assistant smart, not just intelligent.
Clear Roles
Again, in the continuing theme of thinking of the Virtual Assistant a bit like hiring a person, their roles need to be clear. The same goes for the roles of those human employees working with or through the chatbots. I have seen many (non AI project) instances where a project defines new processes without defining the revised roles, or at least not clearly enough. There is an assumption that people will sort it out, which results in some pain before they eventually do. But with this new technology that will not be an option. Clear role definition has to be done, and done properly.
Slow and steady does it
Overall, like with all new things, it’s best to implement IVAs incrementally, get used to them, and then grow their application. It is also probably prudent to reduce the risk by implementing them internally within a particular function first. An example could be within HR to assist with candidate scheduling, or booking flights and hotels for staff travel.
Final Words
In the future, as other AI technologies become commercially viable, there will be more products and services that we’ll see. Then it will be time to consider what new implementation challenges will come up. For now, when it comes to practical concerns of preparing your organization for using AI technologies, these three are the only ones you need to be thinking about.
*The acronym “WTF” stands for “What The Fargin” as per Roman Troy Moronie, who is still in Sweden, as far as I know.
Like this blog? Share it with your colleagues. And sign up and you won’t miss another one!
