Have you considered implementing artificial intelligence for your business? You’re here, so chances are you have, and there is some idea developing in your head. What does your dream AI implementation look like? And, quite importantly, where does it start?
It may come as a surprise but an AI project does not at all start with AI.
Seriously, it doesn’t. Here’s why.
The AI magic
It’s tempting to see AI as the fairy dust of the tech world. It will fix any issue, solve any problem, optimize, improve, boost, increase whatever you want bettered. But your business is not a cupcake that can be ugly but none of that matters when you put some frosting on it. With business, every step matters, so you can’t progress without a strategy.
Cassie Kozyrkow, Chief Decision Scientist at Google, says:
You can’t expect to get anything useful by asking wizards to sprinkle machine learning magic on your business without some effort from you first.
When you want to implement artificial intelligence, you have to do some work on your own. It doesn’t start with hiring a top data scientist or getting an AI degree. You first start with a vision.
Don’t be fooled, though, we’re still not doing any magic, and it’s not the Aladdin wishful thinking kind of vision – it’s a vision backed by strategy and business objectives.
So, number 1 on your getting started with AI to-do list is: VISION.
This means that you should know what you want AI to do. It’s a very general concept at this stage. You may know that you want AI to increase sales, but you don’t have to dive into technical nuances just yet. You got it, let’s move on.
Step number 2: decision-making.
It’s common for companies starting with AI to scatter the decision-making responsibility across different individuals or even departments. That’s because they’re trying something new, they may not know what to expect, and they want to be in this together. While a good team is of great value in any project, the decision-making power has to be assigned to one person. Is it you who calls the shots? Is it a manager, an expert from a given department? Make sure there is a person who’s in charge of making decisions, and make them a part of the process.
Understanding is the key
Normally, when preparing for AI implementation, I would suggest you start with education. It’s still a valid point: in order to make well-informed decisions about introducing AI into your company, you need to have a deeper understanding of AI technologies’ possibilities and limitations, of the use cases, the requirement for AI adoption, and so on. That’s the kind of knowledge that doesn’t just come to you naturally, you need to do some learning before you begin. That’s step 3: identifying the use case. However, if you don’t want to have extensive knowledge of artificial intelligence, there’s nothing wrong with that. When you’ve got the vision – the idea of what you want AI in your business to be – then you know what information to look for. If you want to recommend products, you don’t have to look into chatbots. Clearly, you prioritize: you choose what matters and what doesn’t and that gets you to step number 4 – the objective.
Your vision tells us what AI is generally supposed to achieve for your organization. The objective for your AI project needs to be more specific: it should express the intended goal. Important thing to remember: AI is a tool used to achieve your business goals, not a goal itself. You should only implement AI technologies for their business value, and not to fool around with expensive toys. When you identify the objective, consider quick wins as well. This is especially important for first AI projects when you need to quickly validate the idea of AI for your business, and you don’t want to spend years developing awesome algorithms just to later see that the market has already moved on. Start small, test your assumptions, mitigate the risk of wasting time and money. And always track the results.
As you can see, this step requires some strategic thinking. At this stage, you not only know what you want to achieve but also what success and failure mean to you. Identify metrics to track, establish success and failure criteria. This stage already combines business and technology, so you can use the help of a data science team or AI advisor to help you identify the right goals, metrics, etc.
You can notice that planning your AI project starts from the end – the output. What you want to know first is what you expect the end result to be, and then you move on to what logically comes first – the input.
What does it take to craft AI?
Exactly, what does it take? It takes data, that’s for sure. But what do you do with it? Who builds the model? Who even chooses what model has to be built in the first place? It’s time to do some reality checks – that’s your step 5. Ask yourself difficult questions and look for answers. Don’t lie to yourself – if you don’t have data, you won’t create it overnight, if you lack the know-how, you won’t wake up AI-smart the next morning.
What do you have to know? Here are some examples of the questions you should ask yourself:
This is a part of our AI readiness cheat sheet. Other questions include issues related to budget, data, company culture, and so on. Answering the questions in this test will help you assess where you are in the AI readiness spectrum and identify areas for improvement to make sure that when you start, you are well prepared.
Answering those questions will probably show you some things you’re lacking, and you will have time to think about how you want to bridge those gaps – which is step 6. You might be missing skills, which is a common problem in AI projects – it’s difficult to hire experienced data scientists or AI engineers, and even if you’re about to hire them, they’re quite expensive. If you don’t have any previous experiences with AI projects, you may not know what skills to look for and how to verify that someone’s worth hiring. And you may not even be sure that your project will be a success and in three months’ time, there will still be a need for an in-house data science team. It is commonly recommended to start first AI projects with external teams – they’re easier to find, they start working on your project right away, they’ve worked as a team before, so no time will be wasted. This solution is safer, more time-efficient, and allows for hands-on learning by other members on your team who are making their first steps in AI.
Another thing you might be missing is data. When you already have a data science team, they will help you identify the data required for the outcomes you expect. You may be able to collect the required data, obtain it from third-party providers, or scrape publicly available data. There are solutions to all (well, most) problems, you just need to get creative. And that’s why you need a good partner in crime – data science team. Just kidding, no partnering in crime. Don’t make you AI a criminal. Safe and trustworthy is the way to go.
Ready, set, go for AI
Let’s sum up what you’ve got so far:
- The general vision of AI in your company
- The decision-maker
- The use case(s)
- The objective along with quick wins, success and failure criteria, and relevant metrics
- Reality checks of the current state of your AI readiness
- A plan for bridging the gaps found during the reality check
What do you do next? Are we there yet?
Now, you can finally assemble your team. When you outsource data science, you still need to connect them with your team: the decision-maker, domain experts (e.g. from sales or marketing departments), perhaps your project manager. With the team on board, you can start with what you see as “AI work”: data cleaning and preparation, building, training, and testing the model, and so on – all in accordance with your data strategy. So the “real” Al work – getting the model – is the last stretch of the whole process. Surprising, isn’t it?
Because the process of starting with AI is not at all obvious and it can get tricky, we begin with a 2-day AI Sprint workshop. It’s the time to connect the business with the tech, crystallize the idea, to discuss the possible scenarios, expectations, requirements, issues, build the data strategy along with the project roadmap. It’s the first step, though a big and important one, towards a more seamless AI adoption – in contrast to hastily jumping into a project without giving it careful thought.
Starting with AI is a step-by-step process requiring you to ask yourself some difficult questions, think strategically, and in many cases work differently than before. It’s a new experience and as such is valuable. Starting small allows you to select the right pilot project and execute it successfully. The right execution of the first AI projects can pave the way for successful company-wide AI implementation. Being well-prepared matters.
Learn how to use predictive analytics to drive business results and revenue