Maximize AI Outcome - Essential Building Blocks

Artificial Intelligence continue to shape our world and imagination. Industry analysts continue to project increasing adoption rate in multiple industries. Algorithms we invented have led to huge advances in several different industries. While it touched our imagination with self-driving cars, it entered our lives unannounced through suggestions on what to buy and what to watch. Now it is touching almost all walks of life – gaming, healthcare, fraud detection, court verdicts, insurance claim, hiring and so on. Machines are getting smarter and AI is increasingly integrated in to our daily lives. But the path to AI nirvana for enterprises is often a slippery slope. Those who got the basic building blocks established were successful in driving business outcomes leveraging AI.

In part-1 of this series, I have discussed few use cases of AI. In this article let us explore basic requirements enterprises should possess to exploit the potential of AI and other digital technologies.

Leadership

The goals for incorporating AI in enterprises should match the organization’s overall vision and strategy. AI initiatives should be evaluated for outcome and co-led by both business and IT leadership.

AI introduction requires integration of business processes and many technologies like IoT, ML, NLP, Robotics, Edge Computing, Cloud, Security, etc. This requires experts from internal and external (partners) working together for use case identification, blue print development and road-map creation. The organization structure should enable innovation in respective business units to flourish. By establishing smaller cross functional project based teams, and integrating through the core teams, the workforce will feel empowered, and will create an agile organization.

AI should be an enterprise wide initiative with executive sponsorship and funding.

Change management

Automation is often met with resistance due to the job loss concerns. Change management plays a major role in reducing the friction while enabling human-machine collaboration for a successful outcome. Identify potential job loss areas early (for example Forklift operators, call center agents etc.) and work with the affected employees in retraining and retaining in other areas wherever possible. While improving efficiency and productivity, AI will also create new jobs. For example, some portion of the displaced workforce can be utilized to manage the autonomous bots or train them. Human resources (HR) team will need to be realigned from the ‘recruiting’ mindset to talent management. HR should establish feedback mechanism, and role development based on future skills that are needed.

Infrastructure

Organizations need to invest in specialized high performance hardware to handle AI/ML algorithms. The processing could be handled in traditional CPUs in the case of simple search algorithms, but complex neural networks might require high performance computing clusters, offloading of processing to Graphical Processor Units (GPU) etc. Some use cases requires scale-out - leverage elasticity of the cloud, and in some scale-up with high performance clusters. For real time actions, intelligent edge (edge computing) could be the approach. Such high performance computing infrastructure at data-centers requires appropriate investment in power and cooling as well.

Enterprise-wide AI initiatives will drive investment in storage exponentially. A strong data foundation is key in realizing full potential of AI. Often due to the legacy environment, and proprietary data formats, data aggregation and analysis could be a major challenge for companies.  There should be infrastructure in place to support the data deluge from Internet of Things (IoT) devices - sensors and actuators, as AI and IoT goes hand-in-hand when it comes to business problem solving. Company needs to invest in DMSA (Data Management Solution for Analytics) like data lake to store and interpret structured and unstructured data.

Solutions like robotics at the factory floor or warehouse might require facility remodeling to support autonomous robots and cobots, so that humans and machines can share the work-space safely. Robots requires real time communication to back-end systems for quick remedial actions. Very low latency communication network is instrumental for accurate and safe operations of bots. Real time response requires computing power closer to robots. It might require investment in edge computing technologies to achieve this.

There should also be capabilities to test machine learning models, NLP prototypes and robots without impacting production systems. This often requires establishment of test-beds.

Skills

AI is taking leading role in reorganizing workforce by tackling simple and repetitive tasks and leaving the complex work to humans. It requires workforce with new skills. For example instead of forklift operations, one might need drone (UAV) flying skills. Data analytics skill is becoming mandatory and also AI/ML developer skills. Likewise natural language processing (NLP) requires familiarity with wide range of existing and emerging tools. IT team should be familiar with various conversational AI tools, programming language like Python, R etc., network design and administration of structured and unstructured databases. NLP solutions provides new interfaces for customer interactions that should appropriately be leveraged for better customer experience. Journey mapping and Customer experience (CX) interface development are another set of skills organizations should acquire to transform customer experience and differentiate themselves.

Cybersecurity

The usage of AI technologies to automate and optimize various processes requires connectivity to machines and sensors. Also it requires close collaboration with edge computing network, corporate network and cloud. This increases the attack surface. Moreover AI algorithms are complex to understand. A sophisticated attacker can gain control of the AI systems and manipulate it to behave differently. Enterprises should invest on cyber security measures, implement “need-to-know” framework and micro segmentation.

Data privacy is another prominent concern in spotlight. Algorithms need access to samples, test data sets, behavioral data and images. The solution might have to analyze data from camera footage, video, images, past history, behavioral data, etc. of employees and customer. Global regulations like EU GDPR and other industry specific regulations, must be taken into consideration while designing AI solutions. Company should explicitly state what data being collected, how it is used and discarded post usage to all its stakeholders.

AI can automate business processes, help gaining insight through data analysis, and transform engagement with customers and employees. Its capabilities are growing at an unprecedented rate. When it comes to realization of AI potential, the gap between ambition and execution can be bridged by focusing on these building blocks.