Today, if someone asked for thoughts on artificial intelligence (AI), your mind might paint a pop culture informed picture of a dystopian machine-ruled future or a chatbot with the lexicon of a seven year old.
That’s the problem with new technology. Our vision of the future is coloured by the realities of today, or in many cases, what we witnessed fifteen years ago in Minority Report.
Viewing AI through the lens of a futures market
It’s not a risky proposition to think that the value of AI will rise in the future. IDC suggests that $41 billion will be invested in AI systems for enterprises by 2024, and Forrester projects 13.6 million new AI jobs will be created in the next decade.
If you believe that AI will play a bigger role in business in the future, yesterday was a good time to begin the journey. Some businesses justify inaction by suggesting the technology is unproven; it introduces reputational and financial risk to a business. Why not sit on your hands for three years, and wait for the technology to mature.
Doing nothing is a high risk strategy. To begin with, first movers benefit massively from scaling their internal capabilities ahead of their competitors, particularly in a white hot recruitment market.
Second, to do nothing and to be seen doing nothing invites aggressive competitors to actively target those companies and their customers. Third, allowing competitors to shape the market is to defer to a process you have no control over.
Three areas of AI application
There are three immediate areas of business application for AI. The first is development of virtual assistants, designed to act on behalf of humans in order to better achieve our goals.
Today, there is a fast-growing trend for chatbots. This is perhaps unsurprising given the global popularity of instant messaging (IM) platforms. The format is familiar to anyone who has used IM, and with IM platforms being more popular than their social media equivalents, there is a large tech savvy audience.
WhatsApp alone has sends more messages than SMS globally. Consumers like the fact that messaging works both as an instantaneous, as well as an asynchronous, channel.
Today, chatbot adoption is fighting on two fronts. From a consumer perspective, if a chatbot is not a convenience upgrade on existing alternatives (such as Google search or mobile app functionality) the novelty value of chatbots will soon wear off. From a usefulness perspective, companies struggle to keep up with consumer expectations.
When Capital One launched one of the first Alexa skills in March 2016, customers immediately thought that they could conduct all their banking needs through it.
Capital One are early adopters of the platform and have learned a great deal in the last 18 months, pushing Amazon and the limits of the platform, in terms of ontology size and complexity, along the way.
Whilst chatbots may fade in time, the role of virtual assistants is here to stay. Whilst today many chatbots are merely the equivalent of a call center, interactive voice response (IVR) menu system or an FAQ knowledge retrieval system, over time their ability to handle more nuanced requirements and provide informed advice will grow.
Building successful virtual assistants requires a combination of magic and logic. Magic to build compelling experiences that change consumer behavior, and logic to build smart algorithms that continue to learn and improve decision making.
A second area of immediate business benefit is automation and augmentation. Automation of manual processes, particularly in legacy businesses with legacy technology, has significant cost base implications.
Whilst robot process automation (RPA) is nothing new, the smart application of machine learning to not just convert a manual process into an automated one, but to do so in an autodidactic way, constantly improving the effectiveness and efficiency of the process, is a prime application for AI.
Augmenting workforces with AI driven applications is another source productivity gain. Many forms of customer service interaction are now a combination of human and machine response.
Machines can make individual service representatives more productive by automating repetitive tasks and automatically prompting responses to commonly asked questions.
Inhibitors to value creation
Ultimately, the killer application of AI is the invention of new business models, products and services. It is alluring to think that a firm’s data contains a map to some hidden treasure of a previously undiscovered business model.
The reality is somewhat more mundane. Only those companies with access to the right analytical firepower, coupled with an ability to free their data from the shackles of legacy siloed databases, have a shot at legitimately creating new value from data. Both are serious undertakings with minimal shortcuts.
Talent availability is a serious inhibiter of AI growth. Without a sustainable capability model, businesses are struggling to attract people with the relevant skills, particularly when trying to compete with Google, Amazon and Facebook. Given the low supply, high demand nature of the AI labor market, workers are well compensated, with average salaries of $170k according to Paysa.
Legacy technology is the other hindrance to the implementation of AI applications. Identifying previously unknown relationships within data requires the integration of disparate data sources. Silos are the enemy of integration.
Those companies that have migrated their data to the cloud, have built robust APIs and have reached a higher degree of digitisation are generally in a better place to generate value from their data.
The clock is ticking
There are two ways of looking at generating business value today from AI. One is to get tactical. Developing proof of concept prototypes, getting real consumer feedback, and developing the opportunity to upskill colleagues and learn by doing.
The process of creating a backlog of prioritised use cases along with their respective business cases can help to focus development in small achievable chunks, with each new application building on the underlying knowledge model.
>See also: Is business data AI compatible?
The other is to take a longer term view, and begin to create the structure required to exist in a more AI mature world in three to five years’ time. While developing internal data analytics capabilities, migrating data from silos into an extensible cloud solution and building key strategic partnerships may not provide visceral evidence of progress in the short term, it is vital to long term sustainable success.
Either way, inaction is risky. As the world has been digitised, AI has begun to take off due to the exponential growth of data, reductions in costs of cloud computing and the scalability of virtual machines. Those that adopt an AI first mind-set early are in the best possible position to take advantage of this burgeoning field.
Sourced by Simon James, global lead, Data Analytics at SapientRazorfish