Resistance is futile. Artificial Intelligence (AI) is here, now.
Artificial Intelligence (AI) has become the software equivalent of the semiconductor —ubiquitous, unseen, and capable of changing the shape of society and business. And it is embedded into increasing numbers of systems and applications in a manner that is both transparent and transformational.
It’s here now, and its presence is only growing. Expect AI to be embedded in systems that deal with customers, suppliers, employees, machines, transport, and every other aspect of business activity.
We are not talking futures here. Most medium-sized and large businesses are already using AI-enhanced sales and marketing systems—whether they are aware of it or not. AI allows applications to process massive amounts of data in minutes or seconds and make decisions superior to those of people with decades of experience.
Everywhere you look, AI is assisting and displacing human effort. Legal research is being carried out by AI robots; MRI scans are analyzed with AI-supported systems for higher speed and accuracy; and investors are using AI robots to create investment analysis.
It should be evident that if AI can perform these sorts of tasks, then it might also help business managers in their decision-making and analysis processes. Traditional tools for analysis will soon become too burdensome due to their complex interfaces, their arcane methods, and the sluggishness of their workflows.
AI is revolutionizing analysis tools by making them more user-friendly, faster, more accurate, and cheaper to use. It’s an undeniable revolution and the emerging opportunities that AI presents are numerous and significant.
Business managers don’t have to understand AI technology, but it has become essential for them to understand what it can do and how it impacts business. As with all new technologies the winners are those who see the opportunities and utilize a technology before it becomes the norm. Everyone else can only hope to catch up.
How do we measure intelligence?
With all of its power, if we are going to use AI in our businesses, we’d be wise to have some measure of its intelligence. But how do we do that? We do so by taking steps to define a goal for every AI-embedded effort we deploy. Only by observing outcomes and comparing them with well-defined goals can we determine whether an AI application is truly intelligent. For example, if we use AI to place advertisements on various social media channels and tell the application to optimize clicks, if we find that fewer people click and engage as a result, it would be fair to say that the AI we’ve deployed is not so intelligent.
Setting goals in an AI application turns out to be more complicated than it first might appear because the goals themselves are complex. In the example above, our goal might be to achieve clicks, increase business, and lower costs, all at once. With that understood, we can use the combination of the three to provide us with the measure of intelligence we are looking for.
Since it’s not all that simple to determine how intelligent an AI application is, it’s a good idea to take the time to ensure its value. When dealing with suppliers who claim a whole host of benefits for their AI-embedded solution, for example, it is worth taking the time to take a step back, look at the broader picture, and consider whether the measures being promoted are the ones that are most valuable to your business.
The Four Types of Intelligence
Several components within AI help make it the powerful tool it is to address a broad range of business problems, beginning with machine learning (ML), but also including knowledge representation, search capabilities, and more.
Most of the attention that AI gets these days has to do with machine learning—a data-centric approach to building intelligence. ML uses algorithms that scan historical data and look for patterns of behavior that might be useful in the future. For example, it might discover that people within a certain age range, income bracket, and educational background are much more likely to buy a particular product than the population as a whole. Businesses can use information of this nature in a variety of ways, such as targeting market segments, reducing the cost of sales, and increasing revenue. The key word in the name of this aspect of AI is “learning”—machine learning makes it possible for systems to modify their behavior as conditions change.
There is currently tremendous interest in the machine learning component of AI. ML allows businesses to exploit the vast stores of data that they collect as part of their everyday activities to help them improve both their operational and their tactical decision-making. Building predictive models using ML is still a very labor-intensive activity. The variables involved are numerous, and the combinations can mean data scientists spend months searching for meaningful models. AI automates much of this work. Predictive models have a limited window in which they are valuable, making it critical to be able to refresh them frequently. In some cases, the refresh is measured in minutes, such as in algorithmic trading, and in other cases, it might be measured in several months. Either way, intelligent tools using machine learning can drastically minimize the workload.
ML isn’t without its drawbacks. But with skilled use, it can boost productivity and decision-making accuracy. Wherever there is a plentiful supply of historical data, ML can be used to sharpen decision-making.
Not everything can be learned from historical data. The ability to represent knowledge, store it in a knowledgebase, and use it to automate tasks plays a significant role in how AI benefits business. For businesses of all sizes, AI-supported applications can automatically schedule a wide range of activities. Airlines use AI to reschedule flights based on weather conditions, flight crew availability, the cost of refueling in various locations, and more. Only through the creation of an extensive knowledgebase can AI identify the optimal deployment of resources in real time, on demand. Knowledgebases are also used extensively in medicine, law, and other businesses where knowledge needs to be extracted, processed, and delivered immediately. It is used extensively in planning and reasoning.
Imagine a salesperson who, over the coming month, needs to visit 20 locations that are spread wide apart. A valuable question might be: what schedule would incur the lowest cost? While the problem might sound simple at first, its solution is actually quite complex. Ultimately, a search is needed to evaluate all the alternatives until the best one is found. Many problems in business are of this nature, where problems remain unsolved without the use of search algorithms. For example, the airline mentioned above needs to employ complex search algorithms while accessing their knowledgebase to find the optimal solution.
Communication with humans
For AI to be most useful to us humans, it needs to be able to communicate with us. A great deal of effort is now being directed at getting AI to interact effortlessly with its environment. In practical terms, it means that voice and image inputs and voice and text narrative outputs need to be part of the picture. To this end, natural language processing (NLP) also plays a significant role in AI development; it becomes especially important when people need to communicate with the apps and systems they use. Effortless communication between human and AI will determine how widely it is used in settings where human input is essential.
Any application that makes decisions either wholly or partially autonomously can be said to employ AI. When a search engine displays a list of suitable links, it has made a decision: its best guess at what will be relevant. Voice recognition is a process of matching waveforms to words and deciding which words are correct. And there are many applications in sales and marketing that use AI in general and machine learning in particular. Sales apps can interrogate vast amounts of historical data to suggest which prospects are worth pursuing and which are likely to be a waste of time. Marketing apps can learn which words and images are likely to be most effective in a marketing campaign. In both cases, the learning is ongoing, and the AI modifies the decisions that are being made on an ongoing basis. No human being could match the speed and volume of this kind of analysis.
The Need for Better Analysis Tools
As with ML, the workload associated with analysis using BI tools is often ponderous, with unworkable latency. Businesses are real-time and require real-time analysis. Data preparation—which typically comprises more than two-thirds of the analysis task— is just one candidate for AI. Others include exploration of variables and relationships, programming, arcane user interfaces, and the ongoing need to translate between business activity and technical tools.
Fortunately for us, the problems associated with existing BI tools and platforms are well understood. The fact that only one in four people use a BI tool even though they have access to one indicates that people find them unwieldy and impractical for their everyday needs. The incorporation of AI into BI has already started. Within five years, AI will be executing many of the tedious and time-consuming tasks associated with the use of BI, and the interfaces will understand the language of business.
Ultimately, we want to ask meaningful questions of our BI tools and receive useful answers quickly and accurately. AI will facilitate this.