Read the transcription of the episode:

Don’t rush your Automation.

Hello, Today, let’s have a look at automation of processes, actions, production or any other sector of your company activity. It is receiving lots of attention in conference presentations and lobby conversations. I think it is one of the most popular and trending topics of the emerging new business reality. Such terms as Artificial Intelligence and Machine Learning appear in almost every investment presentations of ambitious startups and fast growers. Many of such founders hope to create unicorns that aim for the world. In most cases, they proudly flash these key words in front of their potential investors, trying to convince them that automation is their best choice. Partly, because investors turn a hungry eye toward anything potentially big and easily scalable. On one hand, it is natural that big projects are investors’ first choice. On the other hand, this approach carries great risks that I would like to bring to your attention in this episode. Are you ready? Let’s do it!

I have recently heard a very good explanation of how to tell Machine Learning from Artificial Intelligence.

It is commonly believed that if it is written in Python, it is Machine Learning, whereas if it claims to be Artificial Intelligence is usually delivered in Power Point, so, frankly, doesn’t exist yet.

Just an idea stage.

Whatever it really is, it is merely called AI because it is a buzzword that makes a project sound more attractive and groundbreaking. I like this explanation because it illustrates how vague and superficial the beginners’ knowledge of automation happens to be. In other words, AI often exists only in Power Point, whereas Machine Learning is that real something that is being born nowadays. Yet still quite often it is more or less advanced algorhythm, not nearly as advanced as the true Machine Learning can be. Realistically, the true AI might not be a feasible investment option for most businesses for another decade or longer. It doesn’t mean that AI will not pop up here and there. However, it is a vast, complex and advanced programme, regarding data, know-how, processes and architecture. In most cases, only the global giants, with tens of thousands of employees and billions of dollars to invest, can venture into it. For most startups and beginners, such resources are out of reach.

Even Machine Learning, by definition, is no more than input-based data aggregation system. Before machines become intelligent, they will just continue to learn on the basis of data input. When they do become intelligent, they will just develop their own learning techniques, but still remain reliant on data input. There has to be a source of data for them to learn how to do it automatically in the future. When I talk of data, I mean massive amounts – not tens or hundreds of data records, but hundreds of thousands or millions of them, that our computer, with our help, could learn from. To develop such capabilities, most companies have to operate for a long time or grow extremely fast for a few years so as to accumulate enough data to get anywhere near machine learning.

There is a huge amount of ways in which things can be handled between manual work and AI or Machine Learning-based automation. Believe it or not, businesses existed before AI or ML. They functioned not only thanks to the genuine human intelligence, but owing to the algorhythms and protocols that for decades have enabled people to do things faster and more cost-efficiently.

Indeed, whether it was during SaunaGrow sessions or while listening to startup presentations as a Black Pearls VC investor, I’ve had the opportunity to witness the same scenario: a company has barely taken off and is already planning to scale its business, automate it using AI and Machine Learning and replace everything they can with machines. But the issue here is that it happens way too fast. Why? There are 3 elements that need to be thought over while considering automation.

Firstly, there are the expenses of automation. Planning and technological costs are usually very high, compared with manual work expenses. It often turns out that a project can be continued for months, cheaper, by hiring 2, 3, 4, 5 or 10 extra staff, without automating its technology.

Secondly, in many cases, organizations don’t have full knowledge of what the process should actually look like, until it has been repeated tens, hundreds or thousands of times. Investing in automation to soon, we run a high risk of wasting the resources on things that don’t need immediate attention at a given stage. Having lots of months or years of experience, we achieve a more complete picture of what our automation process should look like and which sectors are really worth automating.

Lastly, we must remember that every automation widens the distance between us and our customers. What often undergoes automation is the customer service that is replaced by self-service where our customer is supposed to obtain something by himself. This is not really automation, as the workload or a duty is simply dumped on the customer. Of course, owing to our improved functionality or solution, the customer may find something easier to do, but equally often automations detach us from our customers. This is not desirable, especially at the early stage of a company’s development, when we need the contact with our customers to learn how our product is perceived, how it should function as well as to know if we are really moving in the right direction. If we go automated too soon and lose touch with our clients, we are quite likely to lose the live picture of where, in what way and how fast our company should expand.

At a certain stage, most organizations grow up to a moment when automation becomes justifiable, but it is usually much later than most entrepreneurs, the young ones in particular, believe.

To give you an example, at AirHelp, which helps airline passengers’ compensation claims for flight delays and cancellations, most processes were still handled manually by the operation team, when we had as many as 3 thousand clients per month. What’s more, the first process automations appeared at the 10-50 thousand monthly client turnover, whereas some processes remained manually dealt with at the 100 thousand case rate per month. Only after almost 3-4 years of hyper growth in operation and millions served airline customers, we eventually felt the actual need of a more thorough automation. In our case, the full-swing automation process began about 3 years ago, when we had successfully supported not just hundreds of thousands but millions of clients. Even today, with tens of engineers at our product & tech department, it is still not always evident which process elements really need to have their subsequent parts automated.

At the beginning, it was easier and more efficient time and moneywise to just extend the team of our operation specialists who contacted our customers and airlines. That was the time when our staff grew from one person to four hundred people. The 4 years’ period, when that growth happened, gave our firm the opportunity to learn a lot more about the bottlenecks where certain elements or stages called for automation. At that stage, we had lots of suggestions in the company how to optimize those bottlenecks and facilitate our processes. By the way, the topic of dealing with large numbers of innovative ideas in the face of priorities is discussed in a separate podcast “MULTITUDE OF IDEAS”. I hope it will shed more light on what can be done to organize and evaluate them.

The priorities evolve in time. Therefore, the people in charge of action prioritization are of key importance. At AirHelp, we have Natalia Laskowska, definitely the best Operational Department Vice President in the job. Her comprehensive knowledge and understanding in the field of customer service and claim processing has always been the widest in the organization and cannot be overstated. A person in her position has to see the whole process from the beginning to the end, with its requirements, challenges and nuances of claim processing for 400 different airlines. Her expertise and competencies have helped to make the right strategic decisions as to where the company should be headed.

In the automation process, the human factor is just indispensable. We may hear a company talk of AI as their near or distant future goal, but it is the human wisdom and ingenuity that makes a priceless input here. It can’t be forgotten how vital it is who we work with and how this affects the company future.

To sum up today’s automation topic, in my mentoring sessions I always advise to carefully weigh up the pros and cons of it.

I encourage people to reflect whether the automation costs are justifiable against alternative solutions.

I suggest considering spending technological resources on some shorter-term facilitations rather than downright automation. Maybe optimisation is more worth the effort than flat automation which might irreversibly kill off the genuine human-to-human customer relationship. It is highly advisable to give it a thought as a business case before scheduling another automation task. It is better to defer the technological progress in the name of process optimization, before moving on to full automation.

I also recommend reading the article on this issue.

I hope my reflections will help you think through and plan the best time for your automation. Get to know your processes well and, when you feel like going automated, think twice, because following your impulse usually means doing it way too soon.

Thanks and have a nice day.

Click the picture to listen more or record a voice message.