Hyperight

A practical guide to AI-driven transformation for Algorithmic Business

A practical guide to AI-driven transformation for Algorithmic Business
Bookmark

Gartner defines Algorithmic Business as “the industrialized use of complex mathematical algorithms pivotal to driving improved business decisions or process automation for competitive differentiation.”

That might be the definition of a business algorithm or rather an explanation of ‘how’ an algorithmic business works, but it is not the ‘definition’ of algorithmic business. Because it uses terminology that pre-assumes some degree of knowledge about the business algorithms and their applications. 

And I’d like to offer a simpler definition that can be understood by anyone. 

Algorithmic Business is the execution of highly automated business strategies based on dynamic data-driven policies in response to emergent or anticipated opportunities or threats in extremely short time frames in the order of hours, days, or weeks instead of traditional time frames of quarters or financial years prevalent in traditional business that rely upon human decision making.

And with this definition of Algorithmic business, it is much easier to understand Gartner’s definition as to how it works.  Moreover, we can use our new definition to further define the three core business capabilities of algorithmic business namely– Anticipation, Adaptation, and Response. 

Figure: Conceptual description of business capabilities of algorithmic business | A practical guide to AI-driven transformation for Algorithmic Business
Figure: Conceptual description of business capabilities of algorithmic business

Anticipation: 

Anticipation refers to the capability to detect opportunities and threats in advance. The farther the business can detect the opportunities or threats, the better are the odds of realizing/avoiding them. The detection capability is powered by predictions from data analytics and AI. Most organizations have started developing this capability in the last decade and most of the large organizations are already far ahead in their journey while smaller organizations are catching up faster than ever. In the future, this capability will be greatly enhanced with the use of synthetic data and one-shot learning techniques as the bridging of data moats will level the playing field.

Adaptation: 

Adaptation is the second capability for planning the response. It is the preparatory step where organizations flex their physical and mental muscles to identify the best course of action to accomplish the objectives and operating their tactical levers to formulate the response. Given the extremely short window of execution, adaptation is a function of readiness. Readiness is the state of being fully prepared for action and is often used in the context of emergency response where response times are extremely small. 

Emergency services proactively develop their emergency readiness through continuous training and simulating regular drills with different threat scenarios and environmental factors to avoid losing time in strategy formulation during real emergencies. 

In the business context, these drills can be replaced with simulations and experiments that serve the same purpose in developing readiness. In most organizations, this is still a new concept. It is not that they completely lack readiness. Every organization has some degree of readiness to effectively deal with opportunities and threats. But this readiness is implicit. It’s not developed to the extent where it can be readily accessed and deployed to achieve the desired objective in a time of need. Also, this implicit readiness cannot be easily measured and improved in quantitative terms. 

The key driver of adaptation is the underlying business agility and flexibility as expressed by the organization’s decision support and execution systems – ERPs, business processes, rules, heuristics, knowledge management, etc. The degree to which an organization can successfully adapt depends upon the ease of coordination of these complex operational systems. Slow and rigid systems greatly limit the availability of alternatives in planning tactical and operational maneuvers. Tactical adaptation results in a combination of several operational adaptations in multiple systems and processes to create a set of feasible and implementable policies.

Response: 

The response is the final stage where the most optimum policy is selected from the feasible set, deployed, and monitored for performance measurement. This capability is enhanced with practice and experience. When an organization starts its journey in algorithmic business, it has a limited inventory of policies catering to a limited set of objectives that it can successfully achieve. But gradually, it learns from its successes and failures and the inventory of successful policies grows over time. The priority of any organization at the response stage is to minimize damages and risks while testing out unproven policies. A targeted and calibrated response accomplishes bite-sized objectives with bounded risks backed by options to expand, postpone or abandon. 

For example, if one wants to test a new pricing policy, the calibrated response means taking precautionary measures like:

  1. Limiting the number of prospects and the size of the time window for the test price 
  2. Selecting the right channels and sales funnels to target the right profiles
  3. Deciding the evaluation criteria to measure the success or failure as early as possible e.g., target revenue rate
  4. Designing risk controls to abort the test automatically if something goes wrong

The Response is by far the most under-developed business capability in the algorithmic business except for pockets of excellence in marketing and supply chain. The degree to which response can be calibrated also depends on the level of business agility. 

There we have it! A simple yet actionable definition and outline of the algorithmic business pattern. Additionally, organizations aspiring for algorithmic business will also need a whole new set of artifacts like value metrics, billing, and metering models, distribution models, rewards functions, objective functions, and core services like demand learning, price optimization, market testing as agile processes, systems, and capabilities. Algorithmic business consumes the insights from Data and AI-led anticipation, but it also needs new artifacts and core services to formulate policies and responses.

The current level of maturity in these core capabilities and the ability to design and develop core services and artifacts needed by algorithmic business implies an urgent need for organizational transformation in business capability. 

AI-driven transformation for Algorithmic Business

The current organizational structure and ways of working have evolved from the industrial era, and it is based on the division of labor and job specialization inspired by Scientific Management techniques popularized by Frederick Taylor. The core assumption in Taylorism is that the repetitive task or ‘Jobs’ are the fundamental units of industrial value and standardizing jobs drive efficiency. But today, we are in the information era where the norms and assumptions of the industrial era no longer hold. 

Today, Objectives and Key Results (OKRs) are the fundamental drivers of business value while jobs and procedures are dynamic. In algorithmic business, we define the business objectives and outcomes clearly and let data and intelligent systems figure out the best possible solution/strategy to achieve those objectives in the available time frame and assist in its execution. Defining policy objectives and corresponding reward functions is a complex subject with ongoing research. What we know for sure right now is that it requires: 

  1. A lot more people than we have today with a much broader set of skills to experiment, discover, develop, and test these policies before deployment and then monitor their performance after deployment.
  2. Agile systems and processes and adaptive business models to capture and extract maximum value from an emergent opportunity or threat.
  3. A new business framework for policy design based on insight-based execution instead of a plan-based execution.
  4. And more importantly, it requires a change in culture, communication, and beliefs. And that’s the ultimate meaning of AI-driven digital transformation for algorithmic business. 

Additionally, most organizations today lack the maturity in adaptive business models. And it is not surprising because before we attain adaptability, we must first invest in business and process agility. Processes are the ‘code’ of the business that runs on operational systems. And agile processes need flexible and automated operational systems. However, while organizations invest significantly in developing analytics and AI, there is little investment in developing business and process agility. Business Agility, therefore, is the logical starting point for algorithmic business and must be a strategic priority.

Finally, organizations’ decision systems are in desperate need of transformation. They require new models and methodologies to handle changing business environments and multitudes of new scenarios that business has neither experienced before nor does it possess the ability to manage should it materialize tomorrow. Therefore, organizations must also upgrade and enrich their business operating system that contains their tactical decision logic. 

Given the available know-how, changing strategic priorities, limited investment budgets, and the amount of work needed, organizations need a new approach to prioritize and manage this transformation as a continuous and agile value stream.

Commercial Value Streams for Algorithmic Business. 

Commercial Value Stream (CVS) is a sequence of collaborative activities for developing agile business processes, automated systems, services, artifacts, and core capabilities for commercial implementation and deployment of adaptive business models and dynamic policies needed by the algorithmic business. Every CVS has its dedicated infrastructure, ways of working, and resources. Just as technical value streams generate agile code for infrastructure and application, CVS generates the ‘business code’ for algorithmic business. 

CVS provides a robust, focused, and structured mechanism to upgrade the commercial DNA of an organization and facilitates transition into the algorithmic business. As commercial and monetization responsibilities are increasingly falling on the shoulder of Data and AI leaders, organizations must empower them with the right tools and resources to achieve business goals and outcomes. Data and AI leaders must also firmly request the investments and resources for setting up the CVS. In my upcoming talk, I’ll dive deeper into the setup and operation of CVS and its intended output and benefits.


Somil Gupta will be presenting at the Data Innovation Summit on the topic Setting up Commercial Value Stream: Data/AI Commercialization to deliver business outcomes and returns for Data Science and Analytics Managers, focusing on why Data/AI managers need to set up and manage a commercial value stream for Data and AI initiatives and discuss the methodology for doing so.

Learn more about the Data Innovation Summit

About the author

Somil Gupta – AI Strategy and Monetization Advisor | Intakt AI (Part of Svara AB)

Somil Gupta – AI Strategy and Monetization Advisor | Intakt AI (Part of Svara AB)

Somil Gupta is an AI Strategy and Monetization Advisor based in Sweden. He specializes in developing commercialization and business development strategy to grow digital business and monetize investments in data and AI. Before starting his consulting practice, he led business development for digital and AI solutions for Bosch Nordics.

Read more related articles

Add comment