7 Business AI Insights Every CEO Must Know

Sharing is caring: Facebook | Twitter | LinkedIn | Google+

Business AI has broken into the mainstream

2016 was the breakthrough year for personal and business AI regarding the mainstream press coverage. Probably the biggest news was Google’s AlphaGo AI beating a human in the game of Go. The reason why this is such of an achievement is that Go is brutally hard for computers. It involves almost an infinite amount of different combination of moves, which makes it next to impossible to use traditional brute force solutions that calculate all the possible combinations then picking the best course of action.

The saga of AI beating expectations continued in early 2017 when AI was able to do another feat deemed near impossible for our current AI tech. Artificial intelligence beat professional players in Texas Hold’em poker. Compared to Go, poker introduces a whole new level of difficulty because it requires a certain amount of intuition and imagination. It’s not only finding a way to compute the next move efficiently; it’s about creating an AI that can guess reliably what cards the other players might have.

The ability to have intuition and imagination in AI has profound implications for business applications of artificial intelligence.

How to assess artificial intelligence and its progress

Artificial intelligence can be divided into two camps: narrow AI and general AI. Also, the names weak and strong AI are being used. When you watch films like Her or 2001 Space Odyssey, what you see is general AI. We are not even close to that, and many argue that general AI will not even be necessary. It is also fairly obvious from the numbers where all the action is.

> Current predictions tell us that only 0.5% of cumulative revenue share between 2016 and 2025 will occur in companies developing general AI.

Many pundits predict that we will soon enter the Gardner hype cycle's “the trough of disillusionment”. If that happens, it will come mostly from the public’s expectations of AI being general AI when almost all companies in the industry are working on narrow AI.

Narrow AI is essentially highly specialized AI for a specific purpose. Imagine the most brilliant mathematician or developer who literally faints if there’s a need for social interaction. They are stunningly brilliant in one domain and completely fail at others. That is what narrow AI is.

If you are new to AI, it might come as a surprise, but narrow AI applications already surround us. If you have ever listened to music from Spotify’s “discover weekly” playlist, viewed an item based on Amazon’s recommendations, or talked to your mobile phone’s assistant, you have already worked with narrow AI. They are not always brilliant, but they are here.

People often expect some godlike abilities from artificial intelligence, but in practice, for AI to be adopted it only needs to be a little better than humans at performing the same task.

The real-life requirement of AI being just a bit better than human counterpart is why self-driving cars and other types of servant robots are inevitable. They do not need to be better than human beings in every way; they only need to perform the job better and cheaper than a human counterpart would.

The author, Arthur C. Clarke once said that any sufficiently advanced technology is indistinguishable from magic. This is why we will not notice narrow AI until it is everywhere and we start to reflect on the near past.

If you look at your dishwasher now, you might think it is something incredibly basic, and you remember how annoying it was to wash dishes by hand in your younger years. But if you could get your hands on a time machine, jump to the 1920s, and demonstrate your dishwasher to somebody in that period, they would be amazed beyond words by this magical robot. This exact same phenomenon will happen with personal and business AI.

I call this phenomenon invisible adoption of artificial intelligence.

1. The invisible adoption of personal and business AI

The invisible adoption of artificial intelligence will make it look as nothing is happening, and suddenly within few years, we wake up only to find out that AI is in the majority of products and services we consume.

Businesses will use artificial intelligence to deliver their everyday value propositions, and we won’t even blink twice when our only contact with a company is with an AI-powered customer service.

Think at the end of the 90s and how mobile phone adoption now seems like it just exploded — suddenly everybody had one. The same thing will happen with personal and business AI, but this time it will happen inside products and services, so we will not notice it so clearly.

Business owners will face the incredibly hard challenge of transforming their products and services to benefit from business AI.

The leap that companies will see coming but majority will not take in time

During this decade we have already seen how hard it has been for companies to implement “digital transformation.” Most companies ordered a website built on top of CMS, started using online CRM, installed visitor analytics, bought some online advertising, maybe hired an agency to do search engine optimization, and called it done. AI will make it painfully visible that digital transformation means digitalizing all company processes, not only extending company’s presence into the digital world.

Companies without proper digitalization will not be able to benefit from AI because all AIs live and die by the data they have.

I predict that we will see a record amount of bankruptcies in the coming ten years driven by markets adopting business AI one way or another.

If you want to stay updated about all things that you as a business owner need to know about business AI, leave your email below and I’ll send you an update when something important happens.

[wd_hustle id="future-updates-inline"]

Now that you understand the difference between narrow and general AI, and you know the basics of how AI adoption will happen, let’s talk about what you need to know about business AI.

2. AI in business applications is starting to happen

The biggest push we are witnessing right now is AI being adopted by business applications. There’s virtually no industry where applications could not be more intelligent.

The dreams of artificial intelligence assisted work have been floating in business world’s collective consciousness since the 70s, but now they are finally becoming a reality.

Envision a world where your tools give you better recommendations based on your preferences; predictions and insights are automatically extracted from your business’ data; your AI accountant takes care of the financials; you enjoy effortless, automated AI scheduling; AI prioritizes your daily activities in the most optimal way based on how you like to work, and your sales meetings are analyzed by an AI to find where to improve the pitch. Sounds good, right?

Intelligent apps will not only make business processes more efficient, but they will also make your employees happier. For example, AI can already help with unconscious discrimination. You can delegate repetitive tasks to the AI-powered machines and staff can focus on tasks that require human judgment.

What tasks people are doing will change radically in the very near future. Provided with enough training data, narrow AI surpasses humans in many prediction tasks already. As the time passes, algorithms improve and computing resources increases, AIs will be able to predict longer and longer chains of events. Jobs that require taking an input and performing a quick prediction will be turned to AIs.

Jobs that require more comprehensive judgment about what the prediction means and what actions should be taken based on it will remain in the human realm for a while still.

Gartner predicts that significant majority of the world’s 200 largest companies will use AI-driven features in the very near future. The majority of the apps will be AI-driven by 2020.

In the next few years, implementing AI in services & products will become a required core competency for the vendors. Business AI won't be just "a feature" in business applications, it will become "the feature".

3. Artificial intelligence in customer service

Chatbots made some splashes in the media in 2016 when Facebook started to promote their Messenger platform for bots heavily. The marketing push was a little premature, and Facebook caused much damage to the image of chatbots. The technology was not there, and the new paradigm of conversational user experiences turned out to be hard to pull off for UX designers who were used to working on apps.

Bad user experiences caused chatbots to become almost a joke in the press and consumers did not see the benefits. A year later, it is rare to find somebody who prefers to use any of the over 10000 Messenger chatbots available, but there are few examples already where artificial intelligence in customer service works.

However, it would be a mistake to forget chatbots because of the rough start. Jim McHugh, Vice President and General Manager of NVIDIA predicts that in 2017 there will be a chatbot that produces such human-like responses that an average person cannot tell if it is a human or a machine. This advance will cause an AI avalanche in the customer service industry.

I predict that we will see the first fully AI-driven frontline customer service team by the end of 2017. After the initial hurdles, customers will not notice the change from humans to AI because they will not be able to tell that they are communicating with a machine. At the beginning there will be many limitations to how much AI can do for the customer. We still haven’t figured out how get AI to follow up a long chain of related events with enough reliability for an actual real-life business use.

How it works in practice, is that once a chatbot cannot figure out what to do next, it will pass all the data it has collected on the customer’s case so far to a customer service person who will then solve the problem. There are already companies like alphablues who provide this functionality with their chatbots.

In a survey done in the UK, 60% of responders felt that live chat was the appropriate communication channel with a retail business.

The general public is getting comfortable with chat-based communications with companies more and more every year. This is good news because in a phone-based customer service, bouncing customer back and forth between representatives can cause friction and frustration for the customer. In the live chat, the change from chatbot support to human support happens in an instant.

At the end of 2016, Microsoft’s AI team announced a historic achievement of reaching human level accuracy in recognizing the words in a conversation. Now, AI teams all over the globe are focusing on developing AI that can understand the meaning behind the words.

Once our AIs can understand the meaning behind our sentences with enough accuracy, we can automate certain customer service interactions altogether.

On a more worrying note, Forrester predicts that chatbots will create lost wages of $262.7B annually by 2021. They also predict that AI and other intelligent automation will replace 16% of US jobs by 2025. At the same time, they estimate that AI could create new jobs equal to 9% of the workforce.

I think 9% is a very optimistic prediction. Historically, the number of new jobs created by new technologies, especially disruptive ones, has been tiny compared to jobs that suddenly are not needed anymore.

As discussed before, most of the jobs that will turn obsolete will be jobs that either requires only predictive skills that AI can replace easily, or muscle/action based skills that robots powered by AI can replace easily. Jobs that require a valuable judgment of some sorts will be safe.

3. Harnessing business AI in decision making, predicting and forecasting

The year 2016 will be imprinted in the history as the year when everybody predicted US election results completely wrong and the candidate who was not supposed to have any chance of winning ended up winning. Even though AI did not play any part in the actual presidential race, it was a very public prediction failure for many respected research organizations which use AI and various statistical techniques to make predictions out of huge piles of data.

The failure to predict such a big event has rippled into 2017, and we are seeing a lot of research funding going into the area of AI prediction.

Interestingly, there was a healthcare startup called Genic, whose AI predicted that Trump would win over Clinton a few days before the election, despite all polls being against it.

The recent advances in infusing AI with more long-term memory and a capability of understanding multiple tasks will eventually lead to a whole new category of business AI prediction and forecasting tools. These tools will assist executives in complex strategic questions with answers and predictions that take into account all the data available.

Compared to humans, machines can focus on one task indefinitely. This means that while business owners are running out of steam after many hours of pondering a hard question, AI will continue crunching in the background. This will amplify and expand business’ ability to choose the most optimal strategic route in ways we can’t even predict yet. We are already starting to see examples how artificial intelligence is used in the finance sector to augment lawyers.

One of business AI's true value propositions is that it will enable companies to operate in unfamiliar territories without pre-existing knowledge. This means that businesses can use artificial intelligence to simulate possible scenarios without having concrete data of the circumstances. AI will simply use its existing data to extrapolate a different scenario.

We already have AIs that can take a chain of data and predict into the future. They are currently used mostly for translation and text generation, but in the future, the same underlying concepts will be used to simulate very complex business scenarios.

Improved prediction powers are also driven by the emerging field of BOI (Behaviors of Interest). Companies in the industries like customer experience, healthcare, engine performance, self-driving vehicles, manufacturing, and cyber security are in the constant need of better prediction capabilities to spot trends, anomalies, and other BOIs from the data generated by their products and services.

4. Artificial intelligence in HR for happier employees

Companies big and small are always struggling with HR related metrics and processes. A few years ago, when big data was all the hype, there was a push towards “people analytics” and other HRMS (Human Resource Management Software) systems. Collecting data and integrating systems to produce reports has helped businesses to have a better overall view of their employees, but very little has happened to decrease turnover and attrition.

Traditionally, HR has been very much led by psychologists applying science to workplace issues. With all the progress with business AI, more and more HR companies will adopt artificial intelligence to enhance their processes.

Large corporations using HRMS already generate workforce data in such large amounts that human HR specialists cannot keep up with it. AI is expected to help with data processing and creating personalized HR plans for each employee.

In 2014, AI called Ellie, demonstrated that humans like to open up to a machine better than to another human. In the test conducted by a research team, Ellie emulated a psychologist with 239 people. Half of the people were told they were having a therapy session with a person via a 3D avatar and half of them were told the truth.

When test subjects communicated with Ellie, it analyzed their facial expressions for sadness after it asked participants sensitive and intimate questions.

After the session, test subjects evaluated how comfortable they were at disclosing private issues. To help counter subjects' self-evaluation biases, transcripts of the sessions were also analyzed by three human psychologists.

The result was that the group who thought they were communicating with a machine felt more comfortable disclosing personal issues, and in the end, less sadness was detected.

The Ellie experiment demonstrates that humans can be comfortable talking to a machine as long as it helps them as well as a human would. From a business perspective, Ellie paves the way for AI in HR services, like AI work psychologists to help avoid burnouts or an AI to fight against discrimination in the workplace.

5. The great division of AI companies: AI startups versus AI conglomerates

The big five players in AI right now are Google, Facebook, Amazon, Microsoft, and IBM. The big players are all mostly aiming for AI solutions that are the most scalable like, for example, AIs for the health care industry and all sorts vision related AIs. These tasks are very computationally heavy and require enormous datasets, so only the big players have enough resources to tackle these challenges. There’s also been a wave of M&As lately targeting startups competing with the big players. More will follow when incumbent companies who missed the boat will desperately try to grab in-house AI expertise by buying promising startups.

While the big players are busy tackling challenges requiring enormous datasets and warehouses stacked with computing power, there’s a door open for startups to develop heavily specialized solutions for small and medium-sized companies. The majority of these solutions work mostly based on customer’s data, so the playing field is suddenly even between startups and big players.

A level playing field in the specialized AI solutions is excellent news for business all over the globe.

“We will move from mobile-first to an AI-first world.” -Sundar Pichai, Google’s CEO

Even though the big five have dominated the press lately, products and services from a rapidly growing AI startup scene are starting to blossom. Investors are investing at an almost exponential rate, and new AI startups are being founded with an extraordinary rate. A big part of AI startups is focusing on business AI which is great news for us.

If you would like to stay up to date with the latest and most exciting AI startups, leave your email below, and I'll send you an update when something new and exciting arrives.

[wd_hustle id="future-updates-inline"]

If the motto in the 2nd industrial revolution was to take a tool and make it electric, the motto in the coming years is to take almost any existing tool and apply AI to it.

6. Artificial intelligence will be in every product

There’s an incredible push currently underway to grab the dominant market share in the brand new AI specialized CPUs and purpose-built AI computing platforms market. NVIDIA, Intel, AMD, and Microsoft have all introduced their horses to the race. Unlike the traditional CPU and GPU market, the AI computing market is still open for newcomers, and because of this, relatively unknown companies like Cray are joining the race. Even Google has talked about their in-house crafted hardware solutions and the possibility of bringing them into the market.

Most product companies will not have any excuses for not making their products intelligent

Companies who want to apply AI to their products will need to get their hands on AI chips meant for embedding. These chips are specialized solely in running an existing AI models that produce output based on the user's input. Training your AI will still happen in the cloud or with your own computing systems.

Competing technologies and economies of scale will soon push the price of the AI chips so cheap that most product companies have no excuses for not making their products intelligent.

The demand for intelligent products is already visible. For example, in social media, many people have been amazed at how their children would rather play with Amazon’s Alexa than with their toys. The preference for "toys that talk like humans" gives a glimpse at how products that react to our feedback in the most natural way possible will take over the market. Most consumers do not currently wish for an intelligent hammer, but they will change their minds after they have tried one.

Intelligent tools will shift the burden of knowing what to do from the user to the tool

We are most likely the last generation that will have physical tools that don’t, at the minimum, collect data about how we use them or personalize themselves for our preferences.

At CES 2017, L'Oreal revealed a smart hairbrush that records the quality of the user’s hair and helps users to improve their brushing technique. For most people, a smart hairbrush sounds ridiculous, but try to see this as the first step of something bigger and start imagining few years ahead when we have smart combs that can detect changes in your hair that might indicate an illness. Suddenly, the small and seemingly ridiculous products start to make sense.

The bottom line is that whatever your product or service is, now is the time when you should start paying attention to how you can improve it with AI and sensors.

Another positive trend is the standardization and availability of AI development tools. The big players like Google and Microsoft have all released their AI development frameworks and accompanied them with cloud computing services where you can run your AI. The availability of free tools and relatively affordable platforms to train them on effectively makes AI development a possibility for everybody from high school students to Fortune 500 companies.

In the near future, we will see a sort of compatibility layer built on top of different tools to help developers combine various AIs constructed with various technologies into an AI that is more the sum of its parts.

7. AI in business intelligence

AI in business intelligence won’t be only about smarter predictions and insights. It’s also about making business intelligence easier to use for non-tech savvy people. BI tools with interfaces that can parse your questions into insights and predictions will be soon the de facto way of getting information about your business.

Companies who drop the ball with AI-powered business intelligence will soon find themselves with competition that can sell better, produce better, recruit better and do it all without breaking a sweat

Firms that are thinking about investing, or have already invested, in an in-house data science team to help with business intelligence know that around 70% of the team’s time is spent on cleaning the data. Rest of the time goes into selecting the right algorithms and then finding the correct parameters for the chosen algorithms for the data that’s being processed. Some machine learning practitioners describe tuning the algorithm parameters as a “dark art.”

Making data science more efficient is obviously a huge interest for the whole industry, and we are making progress towards the automation of certain processes. Research in automated machine learning has created several open-source projects that are now being used successfully to select algorithms and tuning the parameters automatically. The cost savings from reducing time spent on parameter tuning alone are enormous.

Automated machine learning has been slowly progressing, mostly behind the scenes. Now that business AI is in the news, big data has established itself as a norm, and companies are spending on business intelligence tools, expect to see much more happening in the field of automated machine learning.

Conclusion

2017 will be a fascinating year for business AI. While the public will be inevitably disappointed by AI due to wrong expectations set by the press, businesses have already started to jump on AI to protect their long-term survival.

More and more business applications will be powered by AI, and more and more companies will "bite the dust" when competition runs AI-powered circles around them.

From major areas inside a business, people related functions like customer service & HR will be the most likely candidates to experience radical enhancements from AI-powered solutions. Business intelligence will experience more gradual growth while marketing and sales' AI is getting an enormous push from small marketing and ad tech startups.

Products will start to have intelligent features, and they will begin customizing themselves according to user’s preferences.

Competition in the AI hardware market is heating up, and that will result in cheaper and faster AI chips both for embedding into products and for purpose-built AI computing platforms.

Developing AI will become more standardized, and cloud computing offerings are making it possible for everybody to develop their business AIs. Data science and business intelligence will enjoy more automation around hard-to-get-right processes, and companies will enjoy more efficient data science teams.

Source article: 7 Business AI Insights Every CEO Must Know at Artificial Intelligence in Companies / Machian Future