Case Study: Optimizing Recommendation Workflows for Corporate Insurers through Structured Content
The project
A global corporate insurer sends field engineers to warehouses, manufacturing plants, hospitals, and more to conduct inspections and deliver risk reports. International engineers write reports for international clientele in English, which creates the potential for error.
The reports are channeled through brand assurance (BA) to manage the review process. Each time BA discovers a substantive error, a risk report reissue must be released.
- A large portion of reports require changes
- 10% of all reports require reissues because of errors
- Fixing these errors costs the insurer a staggering amount of money in BA, outsourcing, and external resource and time expenditures
Business intelligence metrics and stakeholder interviews show that more than half of these expenditures are spent fixing what I defined as low-severity and medium-severity non-technical errors. In fact, we were able to determine the exact amount from BI reports.
Low-severity and medium-severity non-technical errors include:
- Style
- Terminology
- Spelling
- Grammar
My role
My role was to build a plan that could reduce the overall risk report cost. I worked with the following colleagues:
- Field engineers (10+ interviews)
- Content advisory groups
- UXC Writing Style groups
- Legal advisory group
- Engineering advisory groups
- External Acrolinx team
- External Tridion CCMS team
Tools
- Figma
- Acrolinx
- Tridion
- Proprietary tools
Process
The first step was to give the engineers tools to make their jobs easier
Using the BA style guide, which was updated frequently, we would create a library in Acrolinx that engineers could incorporate into the writing process.
The library, in addition to Acrolinx’s other applications, would be available as an API feature, which would be a straightforward lift for IT. Engineers could then self-correct while writing, based on BA best practices, instead of saving the edits and fixes for BA only.
While this wouldn’t fix everything, estimates show this step alone could save almost 5-figures worth of hours of BA work per year.
Once this was in place, we could continue to iterate with content components
We discovered during the research process that the engineers used to have a database of “canned text” that was no longer useful because no one managed it.
Using Tridion, we created a model that uses content components that come from managed data sheets, word-for-word, repurposed to replace the current canned text db. We divided risk reports into custom and templated sections. Two of the five sections could be templated, meaning that they could be completed with content components.
These components could be included in the reports, with little to no touches from the engineer, based on metadata from the location, report, and the area of risk.
At this point, generative AI began to pick up steam
While at this time, GenAI was new for content professionals, we decided to find how we could build conditions into our project that would make it easier to incorporate AI advancements that were forthcoming.
The plan was to continue to create clean content for model consumption so that AI responses could use the existing components to help scale out updates as they could be released.
Impact & lessons learned
Unfortunately, before I could finish this project, the UX content department was restructured, and my two-year contract ended without any further renewal. While I was very disappointed I couldn’t see this project through to fruition, I learned a lot about stakeholder management.
- Operating without authority means fostering relationships with stakeholders from different areas of the business. They have to trust you first.
- Understanding other concurrent projects in UX can help you find ways to collaborate. Always be on the lookout for opportunities.
- Content is a hard sell. For some types of organizational cultures, content represents more risk than reward. Content, for example, exposes proprietary information. Know the values of the organization you are working with and find a path that works for them first, and then go from there.
- Change is also a hard sell. The more crystalized your value proposition is for the enterprise as a whole, the better your chances of getting your project approved.
There are several technological, circumstantial, and economic reasons we didn’t get to finish our project. My biggest takeaway is that change is incremental, and you have to start by operating within the confines of the enterprise’s culture. The position of a vendor or third-party contractor means teamwork first. You foster collaboration opportunities. With clients as with customers:
- Take care to understand what success means to them
- Understand the conditions and requirements of each project
- Find collaborators, stakeholders, and champions and nurture those relationships
- Build enterprise-wide value opportunities, find small projects as MVPs