The building blocks to an AI powered organization

Artificial intelligence is reshaping businesses and organizations, but not at the scorching pace one generally assumes it is. Why is that?

15 September 2020


It is true AI has begun doing everything from predicting crop harvests, weather to deciding on bank loans, or seen in our automated customer service. Plus, the technologies that enable AI, like platforms, processing power and data storage, are advancing rapidly and also becoming affordable. Then there are industry pundits who estimate that AI will add $13 trillion (2019 estimate) to the global economy over the next decade and that the time is ripe for companies to capitalize on AI. But this too doesn’t explain the sluggishness in adopting AI systems.

Last year the Harvard Business Review published a report about how companies use and organize for AI and advanced analytics. Their data showed that just 8% of these companies had showed widespread adoption of AI core practices. Most firms had run only ad hoc pilots or were applying AI in just a single business process. The study also had revealed how there was a failure in most companies to rewire the organization--that AI initiatives face daunting cultural and organizational barriers. So, it is mostly for the leadership in these companies to take steps to break these structural-organizational barriers and effectively capture AI’s opportunities. Most of the time it is seen that companies commit the biggest mistakes by viewing AI as a plug-and-play technology with immediate returns. It is not a magic wand.


In this competing environment two research papers are quite significant to help managers or leaders in companies understand and navigate through AI influences using efficient tools.  Research scholars from Swedish institutes-- Ulrich Paschen Lulea University of Technology, and Christine Pitt from Royal Institute of Technology Stockholm-- and Jan Kietzmann from University of Victoria, in Canada developed a typology or a set of analytical tools for managers grappling with AI’s influence on their industries.  Their core argument is that by combining different building blocks and applications of AI, firms can generate different value-creating innovations.


The typology considers the effects of AI-enabled innovations on two dimensions:


             A)      The innovations’ boundaries and their effects on organizational competencies:  The typology’s first dimension distinguishes between product-facing innovations, which influence a firm’s offerings, and process-facing innovations, which influence a firm’s operations.




           B)      Describing innovations as either competence-enhancing or competence-destroying:  Competence-enhancing being those that enhances current knowledge and skills, while competence-destroying renders existing skills and knowledge obsolete.




The researchers have deconstructed AI systems into it’s six building blocks:


         a)      Structured data :


It largely deals with organised data, both intern and external. A few examples the authors quote inventory figures, sales data, or production levels, while external structured data may be found in web-browsing metrics or stock-exchange data




        b)      Unstructured data


Unstructured means non standardized or unorganized; and it is therefore more difficult to analyze. The data that falls under this category are  social media, IoT, and mobile devices falls in this category like blogposts, reviews, or tweets, any of which can contain text, audio, or video files




        c)       Pre-process


Preprocessing of unstructured data involves data cleaning, transformation, and selection, so that the data can be further processed. Two examples, the authors of the study have quoted are  natural language understanding (NLU) and computer vision.


NLU deals with human spoken and written language. When AI is applied, as a first step human language is transcribed into text; this allows AI algorithm to recognize spoken words but not allow it to ascribe meaning to them, like we humans do in our daily lives. This is because ascribing meaning would mean context, dialect, jargon, or dialogue which also brings in significant ambiguity with it.



So, most of the times, most NLU applications uses a lexicon and a set of grammar rules to analyze the data structure, the relationships between the data points, and the context of words or phrases in an assigned natural language. Then the nearest meaning of spoken text is then established using statistical modeling and machine learning. Right now the prominent uses of NLU includes “text summarization sentiment analysis, and relationship extraction.”



Computer Vision : Computer vision transforms images so that the representations can interface with the AI system. This processing is a very demanding task for computers. One of the leading examples is facial recognition used on surveillance video footages for sifting criminals in a public place in real time. Here the computing abilities of a machine is tested by seeing whether it recognizes patterns and then extract meaning.




          d)     Main process


This largely includes three types of intelligent behaviour : problem solving, reasoning, and machine learning.


Problem solving : It largely means “selecting a solution that best achieves a desired goal.” But when there is no single best solution, the divergent problem solving abilities of the AI evaluates alternative solutions that may be equally valuable.


Reasoning: Here the AI system uses logic to reach conclusions from the available data. This is done when patterns and rules is applied by the AI system to both present and future problems. Here AI systems go beyond traditional capabilities when it comes to reasoning under uncertain conditions. It involves a deductive or top-down reasoning attempts to reach new conclusions based on hypotheses that are believed to be true. Inductive or bottom-up reasoning seeks to generate general propositions from individual observations.


Machine learning: This process lets AI systems enhance their performance without depending on any predefined rules that is saved within the system. Consequently, AI researchers developed algorithms able to extract new knowledge from huge quantities of data, making advanced or ‘deep’ machine learning the key component of contemporary AI systems.




        e)     Data storage: Knowledge base


Intelligent behaviour relies on the storage of past data, information, or knowledge so that the experiences reflected in that knowledge can influence subsequent behaviour. In AI systems, these representations can be unstructured data, structured data, or data from pre-processing, as well as information generated by the system itself for the AI processes described in the previous section.



         f)        Information:


Once the processes are complete, an AI system must relate the meaningful information derived  from these processes, either as a basis for human decision making or these could be fed into inputs into other information systems.  AI generated information can also be used for nonhuman tasks in a variety of business applications. Some of the examples worth mentioning here would be



Natural language generation:  One of the best examples the authors have quoted here is the latest generation of Google Assistant which has capabilities of engaging in meaningful two-way conversations that were superficially indistinguishable from human interaction. So


Natural language generation (NLG) is essentially the reverse of natural language understanding; NLG produces complete conversational narratives as output. These can either take the form of a written text, or you can have AI system turn large data sets into reports and business intelligence insights, or they can take the more sophisticated form of generated speech.



Image generation: Image generation is the reverse of image recognition. It outputs complete images, even if there is incomplete information available is incomplete. Even though this technology is not yet very sophisticated, ‘drawing bots’ can already generate images from text descriptions. Today you have AI technologies , though not sophisticated at this stage, like ‘drawing bots’, that can generate images from text descriptions.



Robotics: Robotics lets machines use information to physically interact with and alter their environment.


A typology of AI-enabled innovation and their effects on competencies



The potential of AI is immense. It can introduce significant changes to organizational functions, processes, and product functionalities, as well as alter firms’ competencies and change how they compete within their set of industries.


As an analytic tool for managers, this typology considers the variability of AI enabled innovations and their potential effects on two dimensions: the innovations’ boundaries and their effects on organizational competencies


Competence-enhancing process innovations: Instead of going into definitions, its worth explaining this by an example. The elevator company KONE, provides services to over a million elevators and escalators globally. The company’s technical team undertake regular maintenance of these elevators, including inspecting the material, parts, and components critical to passenger safety. KONE installed Internet of Things (IoT) sensors that use AI to analyze data about each elevator. When the application identifies a possible problem, technicians are notified, perform further diagnostics, and initiate required preventive maintenance. As such, the process competencies of the technician are enhanced through the use of AI.


Competence-enhancing product innovations: The example here will be when car manufacturers incorporate AI in navigation apps, automatic breaking, or alert systems, they improve the driving experience for owners. However, these improvements do not fundamentally alter the set of relevant competencies in the automobile industry, such as the many skills involved in vehicle development or production.


Competence-destroying product innovations : The best example here will be autonomous buses powered by AI in the urban transportation sector. AutoBus pilot is currently underway in some regions of Switzerland. If the project is successful, bus drivers for public transportation system can be replaced by AI-powered driverless vehicles, just as taxi drivers face the looming threat of driver-less cars.


Competence-destroying process innovations : In the advertising world, for instance, firms are using AI to systematize digital-media buying. What that means is that AI-powered machines are making autonomous buying decisions based on multivariate testing of audiences, bids, keywords, targeting, domains, and placements to understand what’s working and what’s not, and what has the greatest potential to reach the goals of a campaign.


Applications of Artificial Intelligence for managers


              1) Anticipate how AI-enabled innovations might affect organization’s competencies: When managers investigate which of their processes or products could be affected by AI and whether the outcomes would be competence enhancing or -destroying, they can understand more fully their firm’s exposure to AI-based risks or opportunities and plan accordingly


                 2) Map how AI affects the competencies of their strategic partners : This can be explained with the example of a retailer who predicts customer demand more accurately with AI. This can not only improve cash flow and inventory management but also enact better plans for transportation firms that supply to the retailer’s warehouse.


                3) The typology as a framework to discuss their organization’s strategic innovation portfolio : A consumer-goods manufacturer can use AI and might invest in initiatives which enhances existing competencies for products, processes, or both.


                4) Managers could use a tool to help them track how AI innovations can introduce value-enhancing or -destroying change to products and processes within their industry.




When is the time to make a shift to AI?


While building AI competencies require cutting-edge technology and talent, it is very important to align a company’s culture, structure, and ways of working to support broad AI adoption. But since most businesses aren’t born digital, traditional mindsets and ways of working can counter those needed for AI


Interdisciplinary collaboration : AI can have the biggest impact when any company begins developing cross-functional teams with a mix of skills and perspectives. This will mean having people on the business development and those concerned with operations, work side by side with analytics experts will ensure that initiatives address broad organizational priorities, not just isolated business issues


Data driver decision making: When AI is adopted broadly, employees up and down the hierarchy will augment their own judgment and intuition with algorithms’ recommendations to arrive at better answers.


To be agile, experimental, and adaptable : Getting early user feedback and then incorporating them into a new version will allow companies to correct minor issues before they become costly problems. for this organizations must shed the mindset that an idea needs to be fully baked or a business tool must have every bell and whistle before it’s deployed. A test-and-learn mentality will re frame mistakes as a source of discoveries, reducing the fear of failure.


Anticipating unique barriers to change: Some obstacles, such as workers’ fear of becoming obsolete, are common across organizations. But a company’s culture may also have distinctive characteristics that contribute to resistance. For this innovation are required. For instance, Harvard Business Review (HBR) cited the example of a financial institution with a strong emphasis on relationship banking,  where it showed how AI can help enhance ties with customers. The bank created a booklet for relationship managers which showed how by combining their expertise and skills with AI’s tailored product recommendations could improve customers’ experiences.