The Last Mile in Becoming a Data-Driven Enterprise

It is fair to say that right across Europe, the last twelve months have been hard with many countries’ economies flat lining if not shrinking. It seems we are now in an era where companies, more than ever, need to maximise value from data and AI.  If that is to happen, data and analytics needs to underpin every business function and mass execution of enterprise business strategy is needed to enable people at all levels to execute on a common business strategy so that decision making at every level contributes to common outcomes.

However, for most organisations, that has never been possible mainly because the majority of employees are not using BI tools to access insights to support decisions. Most surveys indicate that it is rare to see more than 30% of employees with such tools. Generative AI may help increase that with the arrival of natural language queries, but nevertheless there is still a large number of employees that don’t have access to insights. So how do you solve this? Besides, BI tools there are three main ways to maximise value from business analytics, machine learning and AI to achieve data-driven business optimisation. These are:

• Embedded analytics
• Intelligent applications
• AI automation

Embedded analytics is where a user has the ability to analyse data and make use of insights, AI-driven recommendations relevant to a specific decision or decisions while using an application to perform task(s) as part of their normal workflow. There should be no need to switch to a BI tool or another application. A good example of this is a customer service representative (CSR). CSRs have no time to switch to a BI tool to analyse data about a customer they are communicating with in their headset or via webchat. Customer insights and recommendations need to be embedded in the customer service application in use. They just have to be there at the time they are needed, and they need to change each time a CSR starts communicating with another customer.

What do we mean by intelligent applications? Consider a fictitious company called XYZ Eats which offers a restaurant delivery service. They have a mobile app which allows people at home to order food, that restaurants cook and have a driver deliver to them. The business is reliant on getting multiple restaurants to cook to order, multiple drivers to be available and customers using the mobile app which is connected to an enterprise application. If the application is static, it will have code that specifies rules such as selecting the nearest driver for delivery. However, as the business scales with more customers, more restaurants and more drivers, this rule turns out not to work well and developers constantly have to change the code to provide more complex rules. However, if the enterprise application was an intelligent application, it would invoke a self-learning prescriptive machine learning model to dynamically recommend the action needed so that application logic does not have to change as the business scales. In other words, the application itself becomes data driven.

AI-automation is application of advanced technologies like process mining, machine learning (ML), Intelligent Document Processing (IDP), Robotic Process Automation (RPA) and Generative AI to augment workers and automate processes and tasks in ways that are significantly more impactful than traditional approaches. It requires intelligent orchestration of several technologies and combines pre-packaged automated tasks (skills) and organisational knowledge to perform routine and mission-critical tasks faster.

All of these are needed in addition to business intelligence if companies are to become data driven.  In fact, most businesses want business intelligence, generative AI, predictive and prescriptive ML models ‘wired into’ every application to improve productivity, reduce costs, and shorten time to action for better profitability. It is not just a case of doing this on a stand-alone case-by-case basis. What most companies want is to understand what is included in the collection of business intelligence reports, generative AI, predictive and prescriptive ML models to maximise contribution to common business outcomes and precisely where do these needed to be deployed across the business and at what levels to make this possible?

When it comes to BI, ML and AI integration with applications and processes, it is clear that one size does not fit all. In other words, the approach to integrating BI / ML / AI into the natural workflow of a user that needs it may differ depending on their role in the business and the task being performed by a person in that role at the time.

Why is a single approach not enough? Let’s consider four different examples from retail to demonstrate this. These are:

• A customer service representative
• A self-service customer facing mobile application
• A regional store manager
• A distribution centre manager

A customer service representative is tied to an application. They cannot leave the application they use. Yet they need insights about, and personalised recommendations aimed at the customer they are speaking to in their headset or chatting with on webchat.

A self-service customer facing mobile application needs access to on-demand insights and automated personalised recommendations specific to the customer using the application at their time of use to personalise the customer experience.

A regional store manager needs access to BI about the stores and region they manage. They also need automated alerts about staff shortages on any day and automated recommendations to re-optimise personnel to provide good service in the stores they manage. BI, and automated alerts and recommendations on inventory is also needed. In addition, they need access to transaction systems to act on any automated alerts and recommendations received e.g., to reallocate personnel.

A distribution centre manager needs access to BI about store demand, supplier deliveries and inventory. They also need to receive automated alerts about events such as in-bound supplier shipment delays that may cause issues and automated recommendations to get around these problems in order to avoid loss of sales in stores.

AI-automation bots can also help if they are linked to real-time on-line and in-store orders / sales data to automate distribution centre operations and outbound distribution.

Looking at these four examples, people and applications at different levels need to contribute towards achieving the same specific business objective and outcome. However, it is clear that in order to make this possible, different roles require different closed loop strategies to fit with their natural workflow.

The diagram below shows a few ways in which this can be achieved.

Therefore, there are several questions that you need to answer to get this right. These include:

• What is their role in the business?
• What operational and/or managerial process tasks do they perform?
• What applications do they use to perform these tasks?
• During what tasks are insights, recommendations and/or AI- automation needed?
• What insights, recommendations or AI augmentation / automation do they need to help them contribute to the common objective?
• In what form do they need insights, recommendations and/or AI-automation? E.g.,
  • Interactive reports / dashboards embedded in another application
  • On-demand recommendations integrated into another application
  • Alerts in a BI tool / an application / a mobile device
  • Automated action
  • Generative AI and augmented AI-assistance to perform a task faster
• Do they have time to use a BI tool or not?
• What actions does a person in this role, or an application need to take?
• Do people need to collaborate with others before taking action?
• Is the action expected to be automated?

The reason why we need to integrate BI / ML and AI-automation with business processes is to improve efficiency and effectiveness. We can do this to:

• Completely automate mundane repetitive tasks via AI-automation
• Assist people in speeding up tasks via AI-augmentation and generative AI
• Guide people to make more effective and timely decisions during their natural workflow
• Help people and applications contribute towards improving business outcomes
• Make “right time” automations, insights, and recommendations available to every role as they perform specific business tasks during their natural workflow
• Continuously observe business activity to ensure the business remains optimised
• Utilise BI, ML and AI-automation wherever it makes sense to improve specific outcomes

If you are tasked with undertaking this challenge and need help, please join me on my new education class on Embedded Analytics, Intelligent Apps and AI-Automation that I am running for Technology Transfer in April 2024.