Modern software design takes place from the end-user’s perspective. Current software development works with user profiles, use cases, and user tasks as core pieces. At the same time, we are all witnessing a significant enhancement of software with inclusion of artificial intelligent (AI) and machine learning (ML) capabilities. Viewed from the coding we are all aware of the changes and know what needs to change in coding and architecture to include the new functions and features.

What impact does AI/ML enhancement have on people using software? Focusing on User Experience engineering: Can we really go on and still use our old-style user profiles to plan and manufacture the software of the future?

The focus of my short blog here is: What happens to our good old user profiles when AI/ML-enhanced software becomes a reality?

Current Use of User Profiles

In traditional software development user profiles are rather static. They have their goals, responsibilities, needs, and pain points. In practice, single job tasks are matched against software features and functions. The software architecture glues this into an overall business process. In addition, topics like coding language, architecture, performance, security etc. need to be covered. Putting the final software project all together results in a software piece that matches the current market demands.

Accounts Receiveable Persona

Screenshot 1: Example of a generic accounts payable accountant user profile

AI Enhanced Software is a Game-changer

Let’s have a detailed look at the AI/ML enhanced software and its effect on businesses. It can support in two different ways mainly:

  • Complete automation: large chunks of complete business processes are running without any user involvement
  • Partial process automation: complicated, time-consuming, and cumbersome process sub-tasks are done with AI/ML help

In the case of partial process automation, user attention and actions are requested. The following qualities are needed in these cases:

  • Supervision and administration of the of the AI/ML support and results
  • Fast judgement on the appropriateness, fit, and correctness of the AI/ML results
  • Feedback on the AI/ML results to receive better and improved results next time

If we review and compare this with the traditional UX user profiles none of the characteristics listed above are part of it. These qualities need to be incorporated into the user profiles and enhance the main goals, job responsibilities, user needs, and pain points. They need to be mirrored in the user profiles, as future job profiles will demand far more of these qualities.

A Real-life Example from Business Software

All of this sounds rather abstract. I’ll give you a concrete example in which you can see proof of the ideas I have described.

Put yourself in the shoes of an accounts receivable accountant (AR). One of his daily tasks is to report the company’s cash flows by checking the bank statements. Here the accountant looks for the daily cash inflow. Basically, he looks to see if customers have paid their invoices.
In step one, the system acts automatically and solves all easy matches (invoice data matches received payments data). This automation does not work in all cases, so you have exceptions (for example, the customer paid less because of a damaged delivery or paid three invoices at once, etc.). The system can’t resolve these without manual help, so they are displayed to the accountant.

AR Accountant Interface for Balancing Bank Statement Items

Screenshot 2: SAP Fiori 3 labs preview screenshot of the SAP bank statement post processing – ML supports the matching job task here, finding the right customer payments for the fitting invoiced items

Working manually on these exceptions is a time-consuming task.
This is where AI/ML moves to the front line. It provides the AR accountant with solution proposals, showing him how to solve these exceptions (see screenshot 2, red box). In a table, AI/ML collects proposed solutions for solving the exception.
Now, the AR accountant’s supervision and business knowledge are required. The AR judges the AI/ML proposal and evaluates if it fits. If it does, he simply accepts and the system support has saved him a lot of time. If not, he has the option to provide feedback about the mismatch to receive a better proposal next time.

This real-life example proves the change in user profiles’ qualities. There is a shift from the AR having to manually handle lists trying to find matching items, to just having to do a quick spot check for correctness on a prepared system solution. During manual handling, a lot of experience is needed in dealing with bank statements, customers, payment behaviors, invoices, etc. The AI/ML-supported behavior needs a quick decision about correctness. If the system proposal fails, a feedback quality is required from the accountant. In the best case, the AR provides feedback about why the system proposal didn’t match.
What does this change look like in the daily work of AR accountants? Instead of the one carrying out the time-consuming manual work, the AR accountant becomes the expert that supervises and guides the AI to the right business outcomes.

As a takeaway for yourself: You can leverage the power of AI/ML features in your own daily life. Imagine your kitchen toaster breaks and you urgently need a new one – you’ll most likely make use of Google and Amazon’s AI/ML recommendation features. We are all somewhat aware that both companies use AI/ML technologies to analyze our search or buying behaviors. Next time you find yourself looking for a toaster or any other product, leverage the power of their algorithms. Select only the products you really like, and when continuing with browsing you will receive a better and broader variety of matches to your choices.

Summary

As you can see, the traditional and rather static UX user profiles need to be improved. Three new qualities have been identified: Firstly, supervision on the process where AI/ML is leveraged to support users. Secondly, a fast assessment and judgement on the AI/ML calculation results. Thirdly, the required quality is needed to provide the system with digestible feedback.

At the same time, job knowledge and experience are still needed to spot-check the AI/ML proposals. Automation has a key role in this context. With iterative positive feedback about solution proposals, the AI/ML will ask the AR accountant if it can automate the following process step and take even more complex job tasks off of him.

Not logged in