Artificial Intelligence (AI) is becoming embedded in our lives more than ever and is offering considerable benefits to early adopters in a wide range of sectors and applications.
In fact, a recent survey of 250 companies that are already using cognitive technology showed that three-quarters of believed that AI will substantially transform their businesses within three years.
AI is growing as a tool that helps IT leaders achieve their digital transformation goals. Together with main stream adoption of DevOps and Continuous Delivery processes, IT organizations are seeking real-time risk assessment throughout the various stages of the software delivery cycle.
Although AI is not a new concept, applying AI techniques to software development and testing was not popular until just a few years ago. This article discusses how your organization could benefit from becoming AI-driven, some of the benefits of AI, and where to start.
Why Implement AI
A recent survey by Deloitte asked companies about their goals for AI initiatives, with more than half saying their primary goal as to improve existing services and products. Only 22% cited reducing headcount as a reason for the introduction of AI.
It is useful for companies to assess AI in terms of business capabilities rather than worrying about the underlying technology.
In general, AI supports three important business needs: automating business processes, gaining insight through data analysis, and engaging with customers and employees.
Currently, the most common application of AI is the automation of digital and physical tasks, typically back-office administrative and financial activities.
Such robotic process automation (RPA) is becoming much more advanced that earlier business-process automation, with RPA technologies increasingly able to act like humans by inputting, looking for and consuming information from multiple IT systems.
Example tasks include:
- calculating a customers’ service charges by extracting billing information from across multiple systems and multiple document types
- using natural language processing to read legal and contractual documents in order to extract and summarize the important provisions
RPA is one of the least expensive and easiest to implement of the AI technologies, and typically brings a quick and high return on investment. It is well suited to working across multiple back-end systems, although it’s the least “intelligent” as mostly these types of applications aren’t able to learn and improve.
Another common application is to use machine learning (ML) algorithms to detect patterns in huge volumes of data and interpret their meaning.
Example tasks include:
- predict what a particular customer is likely to buy
- identify credit fraud in real time
- analyze warranty data to identify safety or quality problems in products
These tasks are usually very data-intensive and detailed, and the ML algorithms are typically trained on some part of the data set. Typically, the models self-improve over time as their ability to use and interpret new data, categorize things and make predictions steadily gets better.
AI-driven applications that engage employees and customers using natural language processing chatbots, intelligent agents, and machine learning are also steadily growing in use.
For example, an internal company chatbot could be used for answering employee questions on common topics including IT, employee benefits, and HR policy.
Top Business Goals for AI Technology
A survey of businesses in the process of adopting AI technologies identified the following top goals and objectives for AI:
- Enhance the features, functions and performance of products (51%)
- Optimize internal business processes (36%)
- Automate tasks in order to free up employees to do more creative work (36%)
- Make better decisions (35%)
- Create new products (32%)
- Optimize external processes such as marketing and sales (30%)
- Pursue new markets (25%)
- Capture and apply knowledge where needed (25%)
- Reduce head count through automation (22%)
Top AI Benefits
AI facilitates collaboration between enterprise systems and the people that work for the business. Human resources are not made redundant, instead their efforts are empowered and enhanced by emerging AI technologies. In fact, AI technologies can free up valuable human resources for higher-level tasks.
Some benefits of AI are as follows:
- AI can reduce the time taken to perform specific tasks, can enable multi-tasking and ease the workload for existing resources.
- AI reduces the cost of executing both complex and repetitive tasks.
- AI can improve quality and reduce human error.
- AI operates 24×7 without interruption or breaks and has no downtime.
- AI enhances and augments the capabilities of differently abled individuals.
- AI facilitates faster and smarter decision-making.
- AI can unlock new capabilities and enable expansion into new areas of business
- AI can deliver and better and more positive customer experience.
- AI enables improved business monitoring by processing massive amounts of data in real time, highlighting issues, recommending actions and, in some cases, initiating corrective actions.
Top AI Challenges
Research by Boston Consulting Group has suggested that one of the key challenges with implementing AI at scale across a business is not just technical, it’s also in understanding the need for organizational changes to support different ways to make decisions and the investment required in people and processes to achieve meaningful value.
Early adopters that have managed to successfully scale AI across the business typically dedicate 10% of their AI investment to algorithms, 20% to technologies, and 70% to embedding AI into business processes and agile ways of working. In other words, these organizations invest twice as much in people and processes as they do in technologies.
Where there is a lack of investment in people and processes, AI initiatives quickly lose momentum. It’s easy to launch a series of successful pilots, but without a focus on change management throughout the organization from top to bottom it’s nearly impossible to achieve AI at scale.
Some other AI challenges are as follows:
- AI is cutting edge, it’s new and as yet unproven.
- Lack of understanding of AI technologies and limited expertise.
- The computing power required to process the inflow of massive amounts of data and increasingly complex algorithms is costly, whether deployed on-premise or in the cloud.
- Unstructured data can unlock immense business value, but processing and analyzing this data is a big challenge. Types of unstructured data include audio and video files, images, texts, web content, and so on.
- AI still struggles to reach human-level performance for some tasks, for example, recognizing whether an image is of a dog or a cat.
- Data privacy and security is a concern when so much data is being collected, particularly when this involves people’s personal information such as financial details or health records.
- Data scarcity and bias. If limited, incomplete or bad data is used for algorithm development then this can result in badly biased AI models and flawed systems.
- Trust and Ethics. AI, like any technology, is open to abuse and being used for bad purposes. To be accepted there must be complete transparency in the use of data and fairness in algorithms.
What AI cannot do
There’s still plenty of stuff that AI can’t do, so humans aren’t redundant yet (phew!)
- AI suffers from a lack of creativity, and is as yet unable to create, conceptualize, or plan strategically. While AI is great at narrow, given objectives, it is unable to choose its own goals.
- AI lacks common sense, which prevents AI systems from understanding its world, communicating naturally with people, behaving reasonably in unforeseen situations, and learning from new experiences.
- AI lacks the ability to learn continuously and adapt on-the-fly. Today, the typical AI development process is divided into two distinct phases: training and deployment. On completion of the training, the AI model is fixed and then put into deployment. This generates insights and new data, but if you want to update the model, retraining has to be done offline before redeploying. But humans are able to dynamically incorporate continuous input from their environment, adapting their behavior as they go. In the context of machine learning, humans “train” and “deploy” in parallel and in real-time. Today’s AI cannot do this.
- AI cannot understand cause and effect and lacks the ability to form and test hypotheses about the effect that an intervention will have in the world.
- AI lacks empathy and is unable to interact with human feelings such as empathy and compassion. AI is therefore unable to make a person feel understood or cared for, and it would be extremely difficult for AI or robots to be accepted in environments that require empathy and a “human touch”, such as care and social services.
- AI and robotics lack dexterity and cannot accomplish complex physical work that requires precise hand-eye coordination.
- AI cannot reason ethically and lack any real conception of “right” and “wrong” and is unable to evaluate the ethical significance of real-world decisions. Human values are nuanced, amorphous, at times contradictory; and cannot be reduced to a set of definitive rules. This is precisely why philosophy and ethics have been such rich, open-ended fields of human scholarship for centuries. The challenge of building AI that shares, and reliably acts in accordance with, human values is a profoundly complex dimension of developing robust artificial intelligence.
- AI can’t deal with unknown and unstructured spaces, especially ones that it hasn’t observed or previously modelled.
How to Implement AI in Your Organisation
Committing Resources to AI
Build an in-house AI team
A small team that can act quickly to test applications within 6-12 months
Give the team it’s own budget
Promote a CAIO (Chief AI Officer)
- someone who understands what AI can and cannot do
- Work across functions with other business leaders and stakeholders
Other team members
- Product Managers
- Machine learning engineers
- Data science
- Applied scientist
- Data engineers
Data Collection for AI
Data drives everything, so being able to collect and store the correct data in time, or even in real-time, is an essential pre-requisite for implementing AI.
- Define the AI task and the data required:
- what data needs to be collected
- Is the data digitalized or are things still written down on paper
- Is data being collected from the correct processes
- Is the data accessible
- Is the data easy to understand or is it open to misinterpretation
- Be aware of data security, ethics and privacy concerns
- Translate the Data into Information
- Has context
Organizing AI Projects
There’s really no right or wrong way and any of the existing project planning and implementation methodologies can be used, such as:
- Traditional Waterfall approach
- 4-D approach; define it, design it, do it, deliver it!
However, the following are some best practice recommendations in implementing your AI project:
- Develop a high-level AI strategy for the business as a starting
- Start with a small scale pilot
- Provide broad AI training to the rest of the business
- Provide in-depth training for other in-house engineers
- Review learnings from the small scale project
- Implement a large scale pilot
- Update your strategy and provide further training based on your experience and learnings from the pilot projects
- Large Scale Execution
- Changing the organizations culture is the main challenge
Developing an AI Strategy
- Develop the Strategy Towards the End. You will not be able to develop a thoughtful strategy until you have some basic experience, and learnings from initial pilot projects. The more you understand the more likely you are to be able to implement an AI strategy that delivers real business value.
- Aim to learn and adapt continuously, as you learn and gain experience drive any required changes into the next project iteration or new pilot project.
- Be brave! Not all AI projects will be successful but the knowledge and experience gained will prove to be invaluable so review and analyze project failures, capture the knowledge and move onto the next pilot project.
- Develop as Many AI Projects as You Can that are broadly aligned with the overall AI strategy. This makes it hard for competitors to replicate what you are doing and, if you projects are specific to a particular industry, will result in you becoming recognized as an industry leader.
- Specific to Your Industry – become an industry leader
- Develop internal and external communications to let others know that your enterprise is leading the way in AI innovation, developing new tools and assets, and showing what’s possible.
Becoming an AI-Driven organization is not easy but it offers many benefits and AI technologies can transform the way that your business operates, challenge existing hierarchies and orthodoxies, engage your customers, and open up new markets and opportunities.
But remember that AI projects are usually higher than normal projects. AI is cutting edge technology with many uncharted waters – but the possibilities are endless!