The heart of this article is about the implementation of intelligence features in user decision-making support. This blog post will explain why the implementation of intelligence features has a natural tendency to bring about additional complementary intelligence. More specifically, the focus will be on professional business software manufacturing and where and how intelligence enhancements can support in high-stake business decision-making.
Professional software products like Enterprise Resource Planning (ERP) software are made to support enterprises in their business processes. End-users working with such software are usually professional experts inside their job roles, for example Accounts Receivable Accountants, Human Resources Administrators, Operational Purchasers, or Warehouse Managers etc.
All enterprise roles benefit from a supportive ERP software product since this helps these experts with their daily work: The right information displayed at the right time allows users to make appropriate decisions. For instance, one such business decision could be an Operational Purchaser selecting the best supplier for a requested product.
Integrating software intelligence for business processes will support users in making the best user decisions more quickly. The structure below (pic 1) attempts to categorize, abstract, and simplify AI/UX interactions into different interaction types. Two main categories can easily be distinguished from one another: Automation and augmentation.
It’s important that the software intelligence is consistently displayed in a recognizable manner, so that the system intelligence quickly gains the trust of the user.
Situation Handling is a SAP S/4HANA product which combines all these facets. If you’re interested in learning more, you can follow up under this link.
(pic 1 – SAP AI/UX pattern categorization)
Like G. Marcus and E. Davis point out in Rebooting AI: Building Artificial Intelligence We Can Trust, a “master algorithm” with almighty software intelligence is “barking up the wrong tree”.
Segmenting software intelligence into separated purposes, sections, and capabilities, for use in solving and supporting clearly defined tasks and setups is - from a user experience perspective - a much better approach. It is far more manageable to apply these categorizations of software intelligence to business processes, as AI models are much easier to train for clearly defined narrow purposes. At the same time, they support users well as these intelligence enhancements can be nicely integrated into an existing business process.
The vast majority of daily business decisions are high-stake decisions which ERP users carefully make while executing a business process. Following Google’s assessment criteria to determine the stakes of a situation, ERP users touch upon “financial decisions” on many occasions, meaning they are faced with high-stake decision-making with deep business process impacts. Google AI’s solution to such problems is advocating for “explainability and trust”. They try to solve these situations by being open and transparent about the proposal calculated by the system. This is a very important aspect, as the ERP end-users are job domain experts and are therefore aware of the consequences brought about by wrong decisions.
Bringing back the example of an Operational Purchaser again, this job role is responsible for investing large sums of money into material orders which - for example - is essential to keep production lines up and running. At the same time, the enterprise demands that the Operational Purchaser to make an optimal business decision regarding the best supplier and the conditions they offer (not only for the right product, but also for price, product quality, amount of product, delivery time and conditions, available rebates, etc.).
ERP software development teams are confronted with the challenge of supporting these job experts with software intelligence, and this support can occur in two ways:
Regardless of which support type is chosen, the software intelligence needs to know the complete process and user business goal in advance. In any of the cases described, the high-stakes decisions are special. For distinguishing these decision types, information has to be collected by the expert user from multiple resources until they are finally ready to make the business decision.
This detailed decision preparation, structure, and execution increases the number of design requirements for the supporting software intelligence:
Multiple as well as single software intelligence features are plausible in these sorts of scenarios. Using the ERP purchase order again, not only is a recommendation service suggesting a specific supplier useful, but so is having an intelligent prediction service for product delivery time. When are the products available for use? Furthermore, an AI explanation service is requested for both intelligence services. It should be presented in the business language of the user, so that the user can clearly understand which data the algorithms used to calculate the displayed results.
As demonstrated in the example above, software intelligence services have a natural tendency to support and build on each other. The key purpose should always be to keep the user informed, aware, and in control of the intelligence suggestions the system is providing.
Summary
From the easy purchasing order example, it’s clear to see that software intelligence has great potential to help ERP end-users with their daily tasks. Not only can it speed up processes, but it can also automize them. Furthermore, applying one type of system intelligence (example: recommendations) triggers the need for follow-up system intelligence. For example, explanatory intelligence could be used to then explain why this recommendation has been suggested. The process does however have the potential to quickly become highly complex. Therefore, it’s important for the user to be in full control of the system intelligence and able to interrupt these processes at any time – this is especially relevant in high-stakes business decision-making.
If this blog post has given you food for thought, feel free to leave us a comment – we would be interested in hearing your perspective on the application of AI in today’s business software!
References:
Professional software products like Enterprise Resource Planning (ERP) software are made to support enterprises in their business processes. End-users working with such software are usually professional experts inside their job roles, for example Accounts Receivable Accountants, Human Resources Administrators, Operational Purchasers, or Warehouse Managers etc.
All enterprise roles benefit from a supportive ERP software product since this helps these experts with their daily work: The right information displayed at the right time allows users to make appropriate decisions. For instance, one such business decision could be an Operational Purchaser selecting the best supplier for a requested product.
Integrating software intelligence for business processes will support users in making the best user decisions more quickly. The structure below (pic 1) attempts to categorize, abstract, and simplify AI/UX interactions into different interaction types. Two main categories can easily be distinguished from one another: Automation and augmentation.
- Automation: This is a simple automation of a business process. Automation works without user involvement to a greater extent - no detailed explanation is needed. As seen in the structure below, we only have a few bridges between automation and augmentation.
- Augmentation: From a UX point of view, augmentation (user support) falls into several subsections which then divide up further into the following parts: Notifications, suggestions, predictions, explanations, and feedback.
It’s important that the software intelligence is consistently displayed in a recognizable manner, so that the system intelligence quickly gains the trust of the user.
Situation Handling is a SAP S/4HANA product which combines all these facets. If you’re interested in learning more, you can follow up under this link.
(pic 1 – SAP AI/UX pattern categorization)
Like G. Marcus and E. Davis point out in Rebooting AI: Building Artificial Intelligence We Can Trust, a “master algorithm” with almighty software intelligence is “barking up the wrong tree”.
Segmenting software intelligence into separated purposes, sections, and capabilities, for use in solving and supporting clearly defined tasks and setups is - from a user experience perspective - a much better approach. It is far more manageable to apply these categorizations of software intelligence to business processes, as AI models are much easier to train for clearly defined narrow purposes. At the same time, they support users well as these intelligence enhancements can be nicely integrated into an existing business process.
The vast majority of daily business decisions are high-stake decisions which ERP users carefully make while executing a business process. Following Google’s assessment criteria to determine the stakes of a situation, ERP users touch upon “financial decisions” on many occasions, meaning they are faced with high-stake decision-making with deep business process impacts. Google AI’s solution to such problems is advocating for “explainability and trust”. They try to solve these situations by being open and transparent about the proposal calculated by the system. This is a very important aspect, as the ERP end-users are job domain experts and are therefore aware of the consequences brought about by wrong decisions.
Bringing back the example of an Operational Purchaser again, this job role is responsible for investing large sums of money into material orders which - for example - is essential to keep production lines up and running. At the same time, the enterprise demands that the Operational Purchaser to make an optimal business decision regarding the best supplier and the conditions they offer (not only for the right product, but also for price, product quality, amount of product, delivery time and conditions, available rebates, etc.).
ERP software development teams are confronted with the challenge of supporting these job experts with software intelligence, and this support can occur in two ways:
- By supporting single business process sub-steps while placing an order (for example, by single prefilled form input fields made at the system’s suggestion via software intelligence).
- Through suggesting the end result of the business process in advance (for example, by suggesting the final order be placed at a pre-selected supplier as the ERP software intelligence was able to learn and conclude the desired setup and conditions).
Regardless of which support type is chosen, the software intelligence needs to know the complete process and user business goal in advance. In any of the cases described, the high-stakes decisions are special. For distinguishing these decision types, information has to be collected by the expert user from multiple resources until they are finally ready to make the business decision.
This detailed decision preparation, structure, and execution increases the number of design requirements for the supporting software intelligence:
- Firstly and most importantly, the user should always be in full control - the software intelligence support should be stoppable at the end-user’s request.
- The software intelligence needs to know the most important decision criteria in advance.
- The software intelligence needs to be transparent and explanatory, so that the job expert can understand what the software intelligence is communicating.
- To enable the software intelligence to learn, users need to be able to provide feedback at any stage of the process and on any detailed piece of the process.
Multiple as well as single software intelligence features are plausible in these sorts of scenarios. Using the ERP purchase order again, not only is a recommendation service suggesting a specific supplier useful, but so is having an intelligent prediction service for product delivery time. When are the products available for use? Furthermore, an AI explanation service is requested for both intelligence services. It should be presented in the business language of the user, so that the user can clearly understand which data the algorithms used to calculate the displayed results.
As demonstrated in the example above, software intelligence services have a natural tendency to support and build on each other. The key purpose should always be to keep the user informed, aware, and in control of the intelligence suggestions the system is providing.
Summary
From the easy purchasing order example, it’s clear to see that software intelligence has great potential to help ERP end-users with their daily tasks. Not only can it speed up processes, but it can also automize them. Furthermore, applying one type of system intelligence (example: recommendations) triggers the need for follow-up system intelligence. For example, explanatory intelligence could be used to then explain why this recommendation has been suggested. The process does however have the potential to quickly become highly complex. Therefore, it’s important for the user to be in full control of the system intelligence and able to interrupt these processes at any time – this is especially relevant in high-stakes business decision-making.
If this blog post has given you food for thought, feel free to leave us a comment – we would be interested in hearing your perspective on the application of AI in today’s business software!
References:
- “Rebooting AI” by Gary Marcus and Ernest Davis, p.119.
- SAP Help Portal
- Google: People + AI Guidebook
- SAP Design channel on Medium: SAP Fiori 3: How to put intelligence into your user experience
- SAP Design channel on Medium: Approaching Intelligence from a Design Standpoint
- Forbes: How Organizations Can Build Trust Into Their User Experience
- SAP Managed Tags:
