2024 Digital Awards Winner: 3M – Order-to-Cash

By Vin Kumar, Michal Boba, and Josh Bodine
December 10, 2024
Season 6, Episode 11

The Hackett Group’s “2024 Digital Award” winners are transforming complaint management with artificial intelligence (AI). On this episode of the “Business Excelleration® Podcast,” The Hackett Group’s Vin Kumar discusses the award-winning AI project from 3M. He is joined by Josh Bodine and Michal Boba. The project was aimed at revolutionizing the email complaints management process in the order-to-cash cycle. Facing significant challenges in handling customer complaints, short payments and disputes, the team developed a sophisticated AI solution to categorize and analyze these issues, streamlining workflows and driving operational efficiency.

Welcome to The Hackett Group’s “Business Excelleration® Podcast,” where week after week we hear from experts on how to avoid obstacles, manage detours and celebrate milestones on the journey to world-class performance. This episode is hosted by Vin Kumar, AI and Digital Operations practice leader and managing director at The Hackett Group. He is joined by Josh Bodine, global issue resolution specialist, and Michal Boba, AI technical product manager, at 3M. In today’s episode, they discuss the award-winning artificial intelligence (AI) project from 3M that uses AI to handle the email complaints management process in the order-to-cash cycle.

To begin, Josh and Michal talk about the challenges that their Customer Issue Resolution (CIR) team faced in terms of complaint management, disputes of invoice, and managing the collection activity when customers fail to pay on time. The team developed a sophisticated AI solution to categorize and analyze these issues, streamlining workflows and driving operational efficiency. They wanted to bring those challenges into one centralized team and be more productive while focusing on the most value. Their goal was to make this part of the process easier and reduce the focus to the basic data to do the rest of the job. They wanted to be more proactive and less reactive to what’s happening.

This case was identified as a generative AI (Gen AI) solution, but AI wasn’t new to 3M. It started to be used in other avenues, and they wanted to use AI toward emails coming in from customers. They focused on identifying, extracting and bringing that data into the process so it can be analyzed correctly. The solution was not as easy and straightforward. It was a big challenge and had multiple steps from a technical point of view. The data integration was taken from multiple systems and they transformed that data. They automated the entire process and significantly sped up the integration using an AI system message prompt.

Next, they discuss what components of AI were used and what skills were needed for this solution. It was a hybrid task with the best services and practices. They started learning the engineering process and had a four-month output with clear instructions and gave examples. They continuously improved and adjusted the system prompt. 3M also used the top services and companies, and changed services many times. The information technology world is very dynamic in today’s situation, and you need to constantly be aware of what is going on. They built the solution in modules, so they could unplug modules and create independence when needed. They are very open for any future perspectives and options, and could switch anytime for something better. Michal talks about an example of an approach with their first proof of concept and MVP customer translations of the emails. After the translation, they focused on output and extraction.

In addition, Josh and Michal discuss the feedback on using Gen AI to solve this issue. They worked on identifying the specific reason that a customer is contacting them specifically with their complaint. They monitored the success in two ways through overall volume and the success rate. There would be a percentage with AI that doesn’t get all the information they need and they classify those emails at a 69% success rate, but the other 30% is situational. For that 30%, they need to go back to the customer to contact them and find out what is going on. They jumped up through improving the reason matches now to a 79% success rate. They were able to categorize over 59,000 emails correctly. They can classify 69% of the volume used with AI to identify those emails. This case management system has saved them around 1,400 hours since it’s been deployed, and that’s just the first step with categorization. For the future solution, they want to look at other methods in which the customers complain to them and how to bring those documents into the solution. They want to expand these documents into a portal and link what they are already doing there. The next steps will be around the extraction and doing lots of validation steps in the process while looking more at improvements.

In closing, they talk about the end target for this solution. The target goal was to fully automate email complaints from the intake categorization triage, then analyze and finalize the resolution step. They are 60% of the way there, and their final two targets will be in analysis and putting short payments into the process. They want to shift their team’s focus from day-to-day work to the bigger picture.

Time stamps:

  • 0:12 – Welcome to this episode hosted by Vin Kumar.
  • 2:56 – Challenges that the CIR team faced.
  • 6:50 – Why this use case was identified as a Gen AI solution.
  • 10:12 – How did you find the solution for this case?
  • 13:30 – What components of AI were used and what skills were needed?
  • 18:22 – Feedback on using Gen AI to solve this issue.
  • 23:29 – The future solution for this.
  • 26:47 – What is the end target for this solution?