Since I have entered the real world after graduating from my engineering school, I realise a few things about Data Science and AI outside the academic world.
And one thing that’s strikes me is the separation between the innovation that people wish to have in their organisation and the actual innovation that is happening.
So this is what I want to talk about today. What makes an organisation successful at innovating ? What are the myths surrounding Data Science and AI ? What are the prerequisites you need to deliver on your expectation ?
Facing the hard truth
Nowadays with the constant innovation and progress in Data Science and AI, we feel compelled to move forward as quick as possible. Otherwise we feel that we are going to be left behind and while the competition move faster than us.
By doing so, we are assuming primarily 2 things that set us up for failing.
- Lack of digital maturity
The first one is about the lack of objective assessment of the current state of maturity of your organisation.
What is your current level of development ? How well do you know your processes ? Are your data centralised and consolidate ? Are you using a Datalake or a Dataware House ? Why do you use them ? Does they fit the way you use them ? Are they easily accessible ?
Those are the kind of questions that should be asked in order to first assess your level of maturity.
This is a mistake that a lot of organisation make. By over-valuated there level of maturity and assuming they can handle more & faster innovation growth. Where actually they can’t because of their weak foundation.
Building a tower requires first strong foundation. So when you do elevate yourself by innovating, you know everything will not fall apart as you go.
In the next section we will talk about how to assess your level of maturity and how to build a solid roadmap toward innovation using AI and Data Science.
- AI Myth – “We can do anything using AI ? Right ?”
This one come as a consequence of the first one. Because even if you are not fully aware of the lack of maturity in your organisation, you know there is a lot left to be done.
At this point, the idea of having a “End To End” AI model that can solve all your problems at once could sound really appealing. And even if consciously you know it’s not possible, part of your subconscience wish for it to be true.
This is another big danger to let that kind of thought getting into your mind when you start thinking about innovation. But this can be fixed easily.
See this myth emerge from a core lack of understanding of what AI and Data Science does. And I am going to make it simple so you never make this mistake ever again.
What you want to build to innovate is a complete system. And some part are going to be Systematic and other are going to be Discretionary.
Systematic elements of a system are the part that can be developed using basic algorithmic and rule based system. Given the set of input that we have, we know exactly what set of output we are going to get. All cases are handled systematically and we know exactly what to expect from our system.
In this part, AI & Data Science are not require. I would even suggest to avoid using them at all in the systematic part of a system. Because they are used to introduce Discretionary into the system.
Now let’s talk about Discretionary. The discretionary part of a system is going to be the part that until the rise of AI was primarily handled by human. Those are the area where experience and learning from past example come into play. But more than that, this is where we don’t know for sure or we would need millions of rules in order to decide.
This is where AI and Data Science are going to deliver the most value. Where we are going to build Statistical and Probabilistic model in order to help make the same decision that was previously made by humans.
So to wrap things up on the “End To End” AI model. It doesn’t exist. You first need to build a system that handles the Systematic part of your business processes. And once this system is built, you will be able to identify the gray area where innovation using AI and Data Science will actually bring you the most value.
The 7 Stages of automations
Ok great, but know how do I actually assess my current level of maturity in order to innovate the right way ?
Glad you ask. This is where the 7 stages of automation are very useful. Those are:
- Manual
- Scripted
- Tooled
- Automated
- Assisted & Auto-Triggered
- Intelligent
- Global Intelligent
Those stages start from a level of no automation where everything is done by hands. All the way to the 7th stage, where we have a completely autonomous system that requires little to no human intervention into the decision process.
Let’s talk about each one of those stages:
- Manual
This one is simple. This is basically the business that is doing its job. The process is known by the employees through learning and experience. And they spend most of their time executing the process.
The issue with this stage is that there are a lot of tasks to be performed, but all of them don’t have the same value. And often lot of lower value tasks need to be done and eat up a lot of time, which could be used toward more valuable task.
At this stage organisation are primarily focused on getting the job done. Generally short term thinking and focus on hiring more rather than automate more.
In some rare cases this can be enough to run a business, when most of the task that need to be done have equal value. But in most cases, if the business wish to grow over the long term and stay competitive, they need to move to the next stage.
- Scripted
In this stage the business has standardise and systematise the process. It is known by all people and clear for everyone. Each employee can discussion the process and bring idea in order to improve it.
This clarification of the process allows the teams to work on systematic script in order to automate the most repetitive and low value task.
This as a result free the time of the employee so they can focus on the more valuable task they have on their plate.
In this stage a good way of going about it to ask: What are the 20% of the task that eat up 80% of the time of the collaborators ? The 80/20 principles is a very effective tool here.
- Tooled
Ok, you have identified the task to automate and you’ve written the script that automate them. What do we do from now ?
at this stage, the organisation will start facing some scalability challenges. As they build more and more script to automate more and more task, it becomes time consuming to manage them all. While making sure everything is running as smoothly as expected.
This is where the need to build or use tools to manage your automation pipeline emerge.
The target of this is to centralise the management of your system. So everyone knows where and how it’s done. And they can access it quickly and efficiently.
In addition, of saving time in the management of the overall system, this also makes your system more resilient to a change in employee.
With the good tools in place, you are not fearing anymore that your favorite lead tech leave you with a new one that is going to make 6 months to learn everything again.
Automation is in itself a process. So it also is dependent of the engineer that are running it. In the first 2 stages, we were more focused on moving forward in our automation pipeline. But now we need to take care of the automation process. In order to make sure we have a strong foundation before moving to the next stage.
- Automated
Now we go a step further in our automation pipeline. It’s time to start entering in the business’s head to separate the Systematic from the Discretionary.
Until now our focus was on the most obvious repetitive and low value task. But now we wish to go one step further and aim on automating the medium to high-value tasks.
Those tasks are generally performed by the employee. They can be grouped into multiple categories of task, where each one will have their group of systematic task on one side and discretionary task on the other.
It’s important to spend time with each employee to understand their thought process in order to build this classification of tasks.
Once it’s done, in this stage we want to be a humble focus on the systematic task (Rule based tasks) and put aside the discretionary one, knowing that we are not yet ready to handle them.
Then you start to build your systematic system. You test it, deploy it in production while making sure it fit perfectly with the current system of information.
But don’t rush yet into the next stage. You need first to assess what you have done, get the user’s feedback and improve upon that.
Because what you will realise is that, as you build and use the system, more and more systematic ideas will rise among your employees. As they use the system and give feedback, the system will keep improving.
And for some organisation this kind of automation is enough to ensure long-term growth.
- Assisted & Auto-triggered
But for some, it’s not enough. And the need for constant innovation and endless improvement on the processes’ efficiency might be just a requirement to stay in the game.
For those kinds of companies, moving on to this stage might be a necessity, not a choice.
In this stage, the use of data science and AI emerge in parallel to the automation of the automation workflow.
Let’s talk first about the automation of the automation workflow. The idea is to identify the events that can or should trigger any behaviors in the system. We have automate what need to be done, but so far that actual launch of job / application when an event happens has been put aside.
You want to find the events that can exist within your whole automation pipeline that connecting everything together. So when one is firing, everything else that needs to be fired, as a result of this main event, happens.
This is a chain reaction of events that run one job after the other. Here we understand why it’s really important to have built a strong foundation in the prior stages. Because when everything is going to run at once base on one event, you better be sure that everything works as expected.
Now let’s talk about how AI and Data Science will help in this process.
Our chain reaction of the event will only go as far as the first discretionary process occurs. So if we want to propagate the signal as long as possible before any human intervention and therefore improve on our productivity. We need to fill the discretionary holes that we have.
AI and Data Science are the tools you need to fill those holes. At this point innovations using those tools become a must.
And finally, the idea of implemented assisted system. There are discretionary holes that you will fill entirely with AI and Data Science tools. But that will not be in most cases because of the essence of AI and Data Science.
Using those tools you will develop Statistical and Probabilistic models. Those kinds of model aim to create an approximation of reality in order to take decision. But as I say, they are APPROXIMATION of reality, not reality itself.
So no matter how good your model is. It can be wrong from time to time. And depending on the level of risk that your business is willing to take or not, you may want to use those models as tools to help guide your decisions. Rather that letting it take decisions by itself.
At this point we are reaching the edge of automation and how much us as human are willing to let the machine take the decisions.
- Intelligence Stage
The last 6th and 7th stages are a natural evolution of the 5th stage rather than a big change in the automation process.
Here in the 6th stage, once everything is automated, Systematic as well as Discretionary. You will start to monitor the performance overtime you your discretionary models and compare the results to what human are actually doing.
The idea is to constantly analysing the performance and improving the models. Until you reach the point where you’ll have to take a decision.
Is the model performance significantly more important than the one human are doing ? If so, am I willing to let my model take decision by itself ?
To avoid spending time and money for a model that will never be used to this kind of purpose. You may want to ask this question directly once the discretionary model is launched by establishing a performance target.
And you will agree in advance, that once the performance level reaches this target, you will move to a complete automation of the process, removing direct human interaction.
Also, this brings me to another important point in AI and Data Science development. A model can always being improve: more data, new models, new perspective about the process … This is a never-ending process of improvement. Which comes from the very own nature of a discretionary system that tries to mimic human behaviors.
As humans keep learning and growing, the models we built can grow with us.
But in company organisation the goal is to allocate money effectively to maximise the profit. So this important to set a performance target so you know you’ve reach the stage of maintaining performance over time rather than keep improving it. Which cost less money.
- Global Intelligence
Let’s finish with the last stage, the stage of global intelligence. This is the ultimate stage of automation. But there is actually nothing left to do at this stage.
You’ve automated your overall process. Systematic as well as Discretionary part. No human intervention is required anymore. The job of your organization is now to monitor the current performance of the system to make sure there is no drift in the performance of your models. As well as updating and making the processes evolve to respond to the evolution of the market.
In this stage your employee are completely free to work on the most valuable task in their job. The one that requires the most creativity and high level thinking.
Conlusion
To sums up.
Make sure your data are stored properly, easily accessible, and that you keep a proper history of everything. Data is gold, without it nothing is possible.
Assess your current level of maturity using the 7 stages of automations. That way you will know how to build a roadmap for innovation in the 3 to 5 coming years.
And finally, take your time. The business game is an endurance game, the ones that last are the ones that win. Only build on top of what you know are strong foundation.
Hope this has been useful, or at least informative. See you.