How a readily-deployable form of AI is transforming today’s enterprise
The enterprise world is abuzz with artificial intelligence (AI). We’re searching for a way to best use the technology to keep up with a turbulent marketplace. Chatbots, NLP, deep learning, and machine learning are concepts that have become commonplace in executive board rooms as business leaders search for a solution that best fits their needs. However, with so much hype around AI, enterprises are getting detached from what’s useful. There’s a disconnect between what the technology industry is pegging as ‘useful’ AI and what is applicable to enterprise businesses.
The goal of AI and automation in business is to provide a means to streamline processes with the end goal of reducing costs, improving customer experiences, and, of course, increasing profit. And this was meant to happen on a grand scale over the past 12-24 months. From revolutionizing minutiae business process to disrupting whole industries through automation, AI was meant to hit the mainstream.
So, why aren’t we seeing any such results yet? The problem is the lack of utility and functionality in the enterprise approach to AI. While many businesses understand they need to do something with AI, they’re not sure exactly what. In fact, Forrester research indicates that 58% of enterprises are researching AI, while only 12% are using a model. The necessity to explore AI is recognized, but there’s confusion as to how to move from a research phase into an actionable state of deployment.
“While companies like Apple, Facebook and Tesla rollout ground-breaking updates and revolutionary changes to how we interact with machine-learning technology, many of us are still clueless on just how A.I. is being used today by businesses both big and small.” – Forbes
Not all AI is the same
There are two forms of AI: open, or ‘pure’ AI, and pragmatic AI. Open AI is the kind of large-scale, deep learning and cognitive services developed by the major technology companies – think Google Brain, IBM Watson and Microsoft Cortana. It can often feel like science-fiction. And while this type of AI is awe-inspiring and has the potential to transform society in many impressive ways, its application to the business world is less straight-forward. And the reality is that making a machine ‘speak’ is still too demanding for even Google to do very well.
“Deep learning is catching up, but companies still need a large number of such transcriptions to find significant patterns.” – Wired
That’s why open AI has so far failed to make is mark on the enterprise. Instead, it’s paraded around the Silicon Valley technology conferences as a peep into what the future holds. But AI for the enterprise needs to be actionable, tactical and effective in managing or automating specific tasks. The use cases should be defined and measurable. And most importantly, there needs to be a definite path to ROI.
Enter pragmatic AI.
Pragmatic AI: built for the enterprise
Like its namesake suggests, pragmatic AI is the practical application of artificial intelligence to specific tasks and business processes. It combines elements of deep learning with core business processes, and can be deployed through readily accessible channels, resulting in a highly functional solution.
Where open AI needs to learn business processes to grow and be useful to enterprises, pragmatic AI is pre-programmed with the businesses existing set of processes, to make it usable from launch.
The predominant challenge of AI and its application in the business world is utility and determining how enterprises can put the technology to work to deliver results. AI as a solution on its own is weak. It’s the equivalent of buying a car computer and foregoing the actual car – the computer on its own isn’t going to get you from A to B. The same is true for AI – to be useful, it requires integration with backend systems, end-user channels to be deployed through, and a platform to orchestrate business process automation.
An example of pragmatic AI in action is the automation of front-line customer service inquiries using a chatbot. Here’s what that looks like:
1. The chatbot is programmed with current customer service scripts and analysis of the business process to determine in which conversations it will be most effective.
2. It’s then deployed across end-user channels, such as a messaging app or a web chat widget and automates repetitive customer inquiries.
3. The chatbot iteratively improves by learning where new processes and refining old ones. It starts to handle more complex inquiries and a wider range of questions based on ongoing analysis of the interactions.
“The state of the art in pragmatic AI is bright glimmers of intelligence that is advanced enough for enterprises to exploit now.” – Forrester Research
There’s a place for both open and pragmatic AI
Open AI is the technology that will propel the human race forward in profound ways. Open AI will help make medical breakthroughs and discover cures for cancer. It will transform society with driverless vehicles and autonomous travel. For these reasons it is invaluable. But for day-to-day business processes like customer service, billing and marketing, open AI’s applicability is complex, harder to define, and the overall ROI is questionable. In those scenarios, pragmatic AI presents a far superior solution that’s less risky, more easily deployed, and better suited to the enterprise.
Important distinctions between open & pragmatic AI
Long tail versus fat tail solutions
In any given enterprise, there are long tail and fat tail challenges to address. Long tail challenges are refined, unique and occur less often. These challenges are suited to open AI’s capability to go deep and unpack the finer details through continued learning.
Fat tail challenges are in abundance – they flood the enterprise. High volumes of repetitive customer inquiries or simple customer management processes are examples. Solving for these areas requires repetitive process that can be easily replicated by machines. This is where pragmatic AI is a fast and effective solution.
Choosing which to focus on – long tail or fat tail – is up to the enterprise. The difference being that fat tail challenges generally have a bigger impact on customer experience and the bottom line due to their frequency and commonality. Long tail challenges are minor yet very unique, and therefore they affect the customer less.
Open AI programs have longer deployment times because they begin with little prior knowledge of the structures they’re expected to learn. Pragmatic AI is specialized to the application area (learning topic) so it can learn fast. Whereas open AI might learn Calculus by trying to invent its own system of logic and mathematics, pragmatic AI begins already knowing algebra and geometry.
It’s worth noting too that pragmatic and Open AI differ too in their controllability. If open AI breaks there aren’t many ‘knobs’ for the business to turn to correct in a manner consistent with the demands of everyday businesses. Pragmatic AI, indeed pragmatic, is built with knobs easily available for quick changes that come along with operating a modern business.
“Implementation [of AI] isn’t as easy as signing up for a SaaS product. […] deep learning benefits from very specialized, expensive hardware and expertise. Setting up effective models is also time-consuming.” – Wired
Pragmatic AI works on a functional level. It’s deployed with pre-programmed information – like call center IVR scripts – to create predictive decision flows for chatbots to follow. The decision flows are refined over multiple iterations during the development phase, which can be completed in a matter of weeks, or faster if required. The differentiating factor is the fact that the decision flows are refined with each team multiple times before deployment. This means the system can immediately get to work to resolve customer inquiries. Learning and development comes later through analyses when bots need to revert to live agents to resolve inquiries.
Make no mistake, open AI is a powerful tool. Its ability to sift through massive amounts of research and information is invaluable to society. But on the enterprise level, where day-to-day business challenges need quick solutions, open AI is less applicable.
This is where pragmatic AI has such distinct advantages: it is practical and performance-based. Once deployed, pragmatic AI is programmed to execute on a specific set of functions as determined by the business. From changing passwords, to canceling accounts, binding policies and tracking claims, if a human agent can do it, pragmatic AI can do it, too. And do it fast.
Pragmatic AI doesn’t put customers on hold, it doesn’t take lunch breaks, and it doesn’t exist behind a 1-800 number and automated phone menu. It can be arbitrarily scaled up on demand. If it’s 2am and a customer needs something, pragmatic AI can spawn up to work. If it’s 7pm and 1000 people are on the line, pragmatic AI can spawn up 1000 chats.
With its rapid deployment and impressive utility, pragmatic AI lends itself to many enterprise use cases. The use cases fall into three main categories:
1. Customer service: this is the most obvious application of pragmatic AI. Chatbots can easily manage the majority of repetitive inquiries that flood a customer service department. One of the key benefits in this scenario is the ability to eliminate customer wait times and provide 24/7 customer service without the additional cost of additional resourcing and labor.
2. Marketing: pragmatic AI is also an effective tool for outbound engagement and marketing initiatives. Chatbots can be programmed and developed with rich media and content to inspire audiences or used to deliver promotions as part of a wider campaign.
3. Billing and transactions: this feeds into functionality. Consumers expect immediacy and the ability to get things done quickly. The flurry of bots hitting the market without transactional ability and other important functionality components means many businesses have shied away from using the technology. But with pragmatic AI, transactions can be processed without ever leaving the messaging conversation. This allows businesses to automate billing and purchases with chatbots to further streamline repetitive customer interactions and processes.
“Facebook said it was “refocusing” its use of AI after its bots hit a failure rate of 70 percent, meaning bots could only get to 30 percent of requests without some sort of human intervention.” – Digiday
As with any new technology, ROI plays a huge role in determining its ongoing value to the business. Open AI represents a challenge in this respect. The outcome of deploying a large-scale AI system is not always certain. Measurable results or progress even after 6-12 months of an open AI program, is not as common as you would like to think. This is in part due to open AI’s lack of core functionality – determining ROI is therefore difficult.
Pragmatic AI offers far more transparency in this respect. Its ability to reduce costs through automated business processes means ROI can easily be calculated from day one. It’s also less expensive to deploy and less resource-intensive for the enterprise to get started than open AI programs.
Ignoring the hype and choosing what’s right for the enterprise
With AI is dominating the headlines and conversations in the boardroom, it’s important for enterprises to remember that not all AI is built the same. While impressive, open AI may not be the technology to bring solutions and actionable insights to the enterprise. On the other hand, pragmatic AI – with its increased functionality, fast deployment and ability to streamline business processes – provides an appealing opportunity to propel enterprises into the digital era and beyond.