Artificial intelligence (AI) has become somewhat of a buzzword: In a recent survey, 87% of IT decision makers see AI as a necessity to gain a competitive advantage in today’s business landscape.
As a business owner, you might interpret this information as a clear sign that your business needs AI. However, how do you sort through the sea of information to understand if your company should join the AI revolution?
If your data is poorly labelled and you are not sure what to use it for or how exactly AI is going to have a financial impact on your business, you might not be prepared to embrace the AI trend.
Below are 3 signs your company is ready for AI.
1. You Have a Well-Established Data Collection Process
AI thrives on data. In order to learn anything – from simple routine tasks to the analysis of customer data – AI algorithms need large amounts of data to train and understand the task at hand.
Getting good-quality data is one of the biggest challenges in the world of AI. But it’s not only about scraping the internet for data and cleaning it. Publicly available data is accessible to everyone, so the true value lies in your own data.
There is no template that tells you how much data a given AI algorithm needs for proper training. It depends on the complexity of the problem and the complexity of the learning algorithm. What it needs is good-quality, original data.
How do you ensure you have that?
- Format your data consistently. AI-powered algorithms can do wonders with big data, but it cannot generalize. A human can look at a few different spellings of a name and conclude it’s the same person, but an algorithm will classify every spelling variation as a different person. That can negatively affect the predictions.
- Keep your data up to date. This means that you delete old, irrelevant data and that the data has no gaps (missing values or outliers).
- Find missing values or document their absence. Many data sets will have missing values. The reasons can be developers’ mistakes, network malfunctions, another system failure, or a poor data collection process. Understanding the nature of missing values make it easier to fill in the gaps.
- Store the data in the same format. Chances are, a lot of the data needed for your AI project is unstructured (PDFs, text files, even physical documents). Although some algorithms can sort through such documents and derive value, structured data is much easier to work with. To minimize the amount of unstructured data in your company, aim for converting and storing your data in a searchable database. This entails quantifying and tagging your data to express it in numbers, dates, and groups of words.
If your data collection process is well-organized and up-to-date, AI might be a good idea.
2. You Already Know What Area of Business You Plan to Improve
Once you collect enough data, you need to figure out what you will use the data for. After all, machine and deep learning are about solving an existing problem, not uncovering a new one.
There are 3 areas where AI can be of help.
Building Better Products
A good example of how to build a better product with AI is the “Discover Weekly” feature on Spotify. This algorithm-based playlist consists of 30 songs picked based on the music taste of a user. How does Spotify do it?
- User behavior analysis: what songs people listen to
- Natural Language Processing (NLP) for music-related text online: what songs people currently talk about
- Convolutional neural networks: analyze new songs and classify them accordingly
Making Better Business Decisions
It can be questions like “How to predict when a piece of equipment will break,” “How to optimize resource planning and demand forecasting,” or “How to control rent prices.”
New York City recently faced the latter question. New York is notorious for its high rent prices: The average rent for a one-bedroom apartment is $2,940 per month.
Such conditions create a divide among residents based on their income and further class differences among neighborhoods. To combat that, New York City created housing vouchers to subsidize rent.
However, this led to landlords refusing the vouchers or harassing residents who previously used them. New York decided to crack down on offending landlords by sending out investigators to the most exposed neighborhoods.
Investigators found these landlords by applying advanced analytics to data describing things such as available housing, high-performing schools, low amounts of crime, suspiciously low use of housing vouchers, etc.
As a result, a record number of income-discrimination complaints were filed — 120 in total.
Many people don’t realize what AI-powered algorithms are best at: analyzing all available data for one specific task and getting really good at it. This makes AI perfect for automating routine tasks.
As we mentioned earlier, AI thrives on data. To start identifying processes and routines that can be automated in your company, ask the following question: What processes are rich in data or on what occasions do people have to search through and analyze big amounts of data?
For example, in health care, it could mean summarizing long blocks of text from medical journal articles, studies, and patient journals by identifying relevant information before making a diagnosis.
In the investment industry, where detailed information about the companies dictates the stock prices, automation of data collection from multiple sources (public financial reports, press releases, social media chatter, opinion blogs from industry experts, etc.) provides investors with better quality information that leads to better investment decisions.
Knowing what you want to improve with AI is important.
3. You Are Sure Your Investment Can Generate a Positive ROI
Any business investment should have a positive ROI. This also applies to AI.
Some might argue that any boost of AI is good, but such an approach is fueled by the “hype” surrounding AI, not sound business decision-making.
Focusing on the ROI for an AI project prevents you from spending money on something that may just be a superfluous add-on.
How do you calculate the ROI on AI? Before you start crunching numbers, consider the following Venn diagram that shows the qualities an AI project must have in order to bring any ROI: revenue-generating opportunity, a data advantage, and a strategy that’s appropriate for proven AI technology.
Now ask yourself – can the AI project you are about to embark on:
- Generate revenue? By this, we mean that it can solve a real business problem that is either wasting resources or that can create a totally new stream of income.
- Use AI technology that other companies have successfully implemented? There are a lot of proven technologies out there that can be put together to create something unique. The above example about Spotify is proof.
- Be trained with good-quality data specific to your company? An AI algorithm is only as good as the data that is used to train it. If you want your AI project to solve a problem that is specific to your business, you need to collect enough good-quality data that describes all aspects of an area you want to improve.
If the answer is yes to all three, you have an AI project with a possible ROI. Now you can actually calculate the numbers. The basic ROI formula is as follows:
The ROI is the gain from investment subtracting the cost of investment, which is all divided by the cost of investment.
The calculation of the cost of investment for an AI project might be straightforward. If you put an in-house team on the job, the cost of investment would be their salaries for the time it takes to complete the project and any investment needed in software or hardware.
If you hire a consulting company, the quote it gives you would be your cost of investment with a possible percentage on top of that to account for delays.
The gain from the investment is hard to calculate.
For example, if you are trying to automate processes, the minimum gain would be hours saved on that task multiplied by hourly salaries of people performing it. Additional gains would be expressed in how those saved hours are used.
Let’s look at x.ai, the company behind an AI scheduling tool Amy + Andrew, to see how you could calculate the gains of automating something as simple as meeting scheduling.
In 2012, the CEO of x.ai, Dennis Mortensen, scheduled 1,019 meetings. Out of those, 672 were rescheduled. With an average of eight emails per meeting, Dennis spent a lot of his time doing routine work.
With an average of 20 meetings per week and an average CEO salary of $160,000 (according to Glassdoor), the ROI on implementing an AI scheduling assistant for Mortensen looked like this:
With AI scheduling, Mortensen could save 24.4 hours and $3,032 per month – a 17,736% ROI.
The $3,032 saved per months results in an annual savings of $36,384, not to mention the value Mortensen brings to the company by spending that time on other more important tasks.
The cost of Amy + Andrew, in this case, is between $17 and $59, which is clearly more cost-effective than the cost of Mortensen of doing his own scheduling.
It's important to figure out if AI can generate positive ROI for your company.
Is Your Business Ready for AI?
AI is an undoubtedly fascinating trend, but it’s not for everyone. AI algorithms thrive on good-quality data. Without it, the predictions and analyses the algorithms do might be worthless to your company.
Even if you are in possession of big data, you have to decide what you need your algorithm to help you with before you train it. Like a toddler, an AI-powered algorithm needs to be told what to make of the world before it can make any predictions of its own.
Finally, we advise you to approach an AI project like any other investment: Look at the ROI. A farmer does not purchase a new truck simply because it’s the newest model, and neither should you invest in AI simply because it’s the newest trend.
About the Author
Valeryia Shchutskaya is a Marketing Manager at InData Labs, a professional services firm delivering AI-powered software and technical solutions to companies who want to leverage data and machine learning algorithms for business value. Valeryia writes about the ways AI and Data Science change modern business promoting data-driven decisions, processes automation and cost-efficiency.