Short comprehensive guide on current status and what you need to know in order to properly adopt AI in your company.
What exactly is AI?!
Artificial Intelligence (AI) is the most compelling buzzword in tech. Just ignoring all of the alleged disruption that AI started to have upon everything, might seem like you don’t want a long and prosperous life for your business. The amount of articles, news, reports, conferences related to AI is just huge, almost becoming impossible for a starter to identify the meaningful content. All of this noise has also streamlined the creation of two rival ideologies regarding what AI might bring to the future of humanity: an utopian side, where AI will solve all problems and leave people unemployed, and a dystopian side that considers AI as the biggest threat to humanity.  
Nevertheless, as always, the truth is somewhere in between, so let’s have a closer look together.
The AI concept, that we refer to today, was invented in 1954 by John McCarthy and was used for the first time at the famous Dartmouth workshop. This was the moment when AI gained its name and its mission. The scope of the conference was “to proceed on the basis of the conjecture that every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it. An attempt will be made to find how to make machines use language, form abstractions and concepts, solve kinds of problems now reserved for humans, and improve themselves”
Starting from that moment, the concept of AI gains a lot of attention and resources started to be invested in further developments.
What we nowadays call AI represents a broad discipline, that borrows concepts from many other sciences such as: neurology, cybernetics, information theory and Alan Turing’s theory of computation. It actually pursuits the objective of helping machines learn by themselves.
Machine learning (ML), a subset of AI, represents the ability of machines to learn to perform different tasks or solve problems. Most people, when they say AI, they actually refer to ML. Furthermore, Deep Learning (DL) represents a set of ML techniques that can recognize patterns (e.g. such as image recognition algorithms).
For a more detailed description of what exactly is AI, please read this article. Although it is focused on cyber security, it provides enough generic insights.
What can we use AI for?
There is a bit of a hype around AI/ML. According to some experts the world is already being disrupted, as many industries are reshaped by AI.
Before jumping into such conclusions please have a short read into this article, written by Andrew Ng, one of the most reputable AI experts worldwide. His overall conclusion is that “if a typical person can do a mental task with less than one second of thought, we can probably automate it using AI either now or in the near future”. So, by using AI/ML, we can automate the little annoying things that keep us from focusing on the big picture. By all means, this does not mean that AI is the solution to everything. It can be seen mainly as a better tool.
I have recently run into this report, that I must admit was quite shocking. 40% of EU’s AI startups, don’t even use AI/ML at all. Talking about the AI hype, this is a clear sign that many people didn’t really get what AI/ML is. AI/ML are being used at this point as a differentiator in business, but one has to be very careful to identify potential abuses.
One McKinsey article from Nov. 2018 provides some good insights on the use of AI in different industries. Apparently, the lead is taken by Telecom, Hi-Tech and Financial services who use AI more often than others, mainly for service operations and product/service development. As a side note these industries are already highly digitalized, so we might draw the conclusions that these sectors will embrace first the AI/ML challenge.
Fig. 1 – AI usage per industry and business function (Source: McKinsey survey).
Another useful resource is the Deloitte State of AI in Enterprises, from 2018 (Fig. 2). Besides many other useful information, you can find a graph on the top use cases for AI/ML. As you can notice bellow, top uses are recorded in the Tech and Financial sectors, a finding that goes hand in hand with the results from the McKinsey study.
Fig. 2 – Use cases for AI/ML (Source: Deloitte State of AI in Enterprise)
Now, before starting an AI business case within your company, you will need to dig a little deeper. First of all, you will need to develop some kind of strategy, as to what exactly would you like to do with AI. As seen before, not everybody uses AI at this point, but if you really want to position yourself in the leader quadrant, you need to start drafting your strategy. This means identifying in what areas you want to use AI, and ultimately choose what type of algorithms you are going to use. There’s an abundant array of AI/ML algorithms at this point, each one being appropriate for solving rather particular problems, as opposed to one size fits all solutions.
But there are also certain aspects that you need to be aware of. First of all, you cannot expect AI to solve everything for you. We’re just not at that stage yet. What you cannot definitely replace with AI/ML is human creativity. You should not be drawn into such illusions where your company can be run only by AI (except if you have an extremely simple business model). “Human creative achievement, because of the way it is socially embedded, will not succumb to advances in artificial intelligence”. The quote comes from this article, and I couldn’t agree more with the author here.
Another interesting article from McKinsey, on the economics of AI, goes in a similar direction, mentioning that machines equipped with AI actually are becoming very good at prediction. But taking a decision, requires also judgement, that is usually done by a human. It is humans that put that prediction into context and decide on the next steps and the actions to be taken. So, what is truly happening is that “the value of human prediction falls, while the value of human judgment goes up, because AI doesn’t do judgment”. Taking as an example a food demand forecasting: AI will be worthless in the absence of a grocery store to actually buy that food; this represents context and actions.
As an afterthought, the judgement above, puts the whole AI/ML job loss saga under serious debate. Indeed, jobs might be lost, especially the ones that are prone to automation, but other will be created. People might need a bit of reconversion.
The cost of adopting AI
Clearly, you’re not going to get a clear figure as what regards pricing. Although, you might find examples where prices are shown wide in the open (hereor here), going around 300k or more, usually intense analysis has to be done in order to determine the overall costs of implementing AI/ML.
As an advice, I would rather point you towards sources teaching you how to build your analysis.
This great article from ZDNET outlines that the true benefits of implementing AI/ML are manpower savings due to automation of parts of the operational and decision-making processes. Nevertheless, it is rarely that AI/ML will manage to automate the end-to-end business workflow.
But reducing operations duration, does not mean the products or services that you deliver will improve overall. You might end up creating other bottlenecks throughout the organization, as other business processes might not keep up with the one you automate. That is why AI/ML adoption should be seen as a company-wide strategy, with clearly defined objectives and ROI (return of investment) projections. Simply adopting an AI/ML chatbot will not bring extra benefits if, overall, your client service sucks. Therefore, you’ll have to go along the line and impregnate that new AI/ML into your company’s DNA.
Among the costs that you should consider, are the following:
- Infrastructure: Hardware, software, cloud, datacenters, energy related costs.
- AI/ML expertise: the people coming with this kind of expertise, must be payed, and quite a lot, I’m afraid!
- Actual solution development: the cost of the solution itself. There is quite a difference between COTS (Commercial off-the-shelf) and custom solutions. But if you find a COTS that fits your needs, grab it. Should be much cheaper that any of the custom solutions.
- Testing related costs: machine learning implies the use of big relevant datasets, so that the learning process can be done. If you own that data, you might spend some bucks to re-structure it properly; if you don’t own it, well, that’s another story. Learning, can also take a lot of time, as properly training an AI/ML system might take many months or even years until reaching satisfying results.
- Business process redesign: clearly some process revision will need to follow. AI/ML will definitely improve, remove or adjust different business processes that will automatically trigger the redesign of others. As mentioned before, a company wide strategy is something that you must consider.
- System integration costs: You surely do know by now that integrating any new solution, into current environment, is one big pain in the ass for companies. Just remember when you implemented your ticketing system and you needed integration with every department in the company. It might go hand in hand with the business process redesign, but it is usually run by another department. All those API’s and dependencies will need to be reviewed and updated accordingly.
Don’t take this list for granted as its non-exhaustive. It can be much bigger, and could change a lot based on your particularities.
How do you start?
AI/ML projects will fail, like any other type of project, if you do not do your homework properly. Many companies get trapped in the AI/ML hype and try implementing it just because it’s trendy or it sounds good for the business. Things do not go necessarily this way. Since AI/ML is a disruptive technology you need certain prerequisites before seeing any results.
And these prerequisites might also add up to the costs, so budget wisely! You should consider the following:
- Develop an AI/ML strategy at company level. This comes in several flavors:
Choosing the right technology: out of the many algorithms out there, chose the ones that fit you better. Choose also the other part needed by your new software (on - premise, cloud, software frameworks for development etc.)
Build up a team with the proper skills that can handle your AI transformation. AI/ML transformation goes beyond technology. It’s the type of technology that has deep implications on many levels, such as social media. That is why you need more than engineers to drive your strategy. Lawyers, psychologists, UX designers, communication and social media experts, customer advocacy etc. There is a big ethics debate on AI going on now. Don’t underestimate this!
- Build a pilot project. You will need to convince a lot of people on your good results and the extra funding needed. Start with something small, that can really demonstrate AI’s worth. Then go big!
- Identify the data needed for testing. AI needs data, a lot of it. And it must be accurate also. So be prepared to invest some money into getting proper data. Don’t hurry with the tests, let the AI system learn until you reach a satisfactory error rate.
- Get professional help. Engaging an expert consultant to do the analysis of your core business (Gun.io) to see where AI could be utilized can alleviate the heavy lifting and mitigate mistakes you might otherwise make without expert help.
Andrew Ng, the renowned expert mentioned before, has published an AI Transformation Playbook. That’s quite a useful playbook, although you need to scale it down a bit as its “tailored primarily for larger enterprises with a market cap from $500M to $500B”. But it definitely worth a read.
One important thing mentioned in the playbook is that if you add up AI/ML to any typical company you won’t get an AI company. “For your company to become great at AI, you will have to organize your company to do the things that AI lets you do really well.”
P.S. Here’s a list of useful resources. Some of them inspired this article. Enjoy!
Dan Tofan is an expert in cybersecurity, with more than 10 years of experience, gathered in EU level institutions or working groups, national governmental agencies as well as in the academic and private sectors. He holds a PhD in computer science as well as a number of international certifications in the areas of cyber security and project management.