AI Script Generator
How might we help Investment Relationship teams prepare for Earnings?
--
For an Investment Relationship Officer (IRO), one of the most stressful and hectic times in their quarter is preparing for Earnings. A key part of an IRO’s job is to manage communication between a company and the financial community, including its shareholders and potential investors. At Earnings, the company will report on their previous quarter — their financials, and whether or not they met their goals — and they will reveal plans for the coming quarter. This is a crucial moment of communication between company and community, when investors can be won and lost.
There are so many moving parts to preparing for this, including filing government reports, scheduling meetings with shareholders, managing press releases, and planning the earnings call.
This problem space is broad, but as a company geared towards helping IROs in every facet of their work, Q4 was determined to solve as much of it as possible. In order to move quickly, many teams were formed and the problem spaces were split to increase focus and specificity. A small team of scrappy developers were assigned to validate if we could utilize AI to create a draft of the earnings script for an IRO. Once this concept was validated, I was brought in to manage the UX of the product in preparation for release.
Design challenge
Determine areas in need of immediate design improvement and work with the team on roadmap items to be released in three sprints.
Design Audit
I was brought onto this project completely blind with no prior insight into the work that had been done. This was for the best, as I first needed to look at the product with fresh eyes so that I could generate a list of UX priorities to be fixed.
I was given various artifacts to onboard myself, including demo recordings, product documentation, and access to the product in our staging environment. I began by logging in blind so that I could experience the product for the first time as a user would — with no other context then the goal I was setting out to achieve. I went through all the steps of the workflow, intentionally breaking things along the way so that I could pinpoint areas where the user would get stuck. Ultimately, I came up with a list of 36 items to be solved, ranging from minor to major. I categorized these into 4 categories — new features, UX, UI, and copy — and prioritized them into a major list.
From there, I began to review the documentation provided to me and refined my list.
Integration pains
As I became integrated on the development team, it was clear that we were going to need to do some work to trust each other. I am a big advocate for cross functional collaboration, but this team had been working without a designer or a product manager for months. The product manager had joined the team a few weeks before I did. We had worked together before, so that was an easy relationship to continue.
I was immediately onboarded onto ongoing tasks, and began to provide designs for sprint work. It became immediately clear that my work was being taken as suggestion. Despite design reviews and grooming sessions, when the developers began working on something they went with their gut over my designs.
It was clear to me that I needed to gain their trust quickly if we were going to be successful.
Show and tell
When I am working with developers, especially senior developers, I have a list of tried and true methods to build rapport and collaboration. One of the most successful tactics I’ve employed in the past has been bringing them into the design process. By showing work early and often, and by allowing them opportunity to contribute to the final design, we build shared ownership of design artifacts and things are more likely to be built as designed.
With this team, this was moderately successful. It was a very small team of only three developers, so only the lead developer could be spared for these discussions. He was vocal with his feedback, which helped drive discussions and build the foundation of trust. He was incredibly knowledgable in the problem space, which was a huge help to me who had just been brought on the the project.
Misaligned
After the first sprint was completed, I found that the designs that I had provided were not build as designed. This has happened to all designers at one point or another, but on a project with such a short timeline this was particularly difficult to work through.
The design was concerned with allowing users to upload files for the AI to pull data from for the script. The step in the workflow included the upload of these files, and text only notes from the user. In the design exploration of this, we explored many different ways to handle this based on the types of files and notes that users were looking to include. As a triad, we had landed on the following design:
After the sprint, the following design was delivered:
This became the best opportunity to build trust with the development team. I was able to indicate my concerns with a real example, and to set my expectations with the team.
We discussed the communication failures that resulted in the misalignment. It turned out that there was some confusion about the function of the designs that indicated a takeaway for me to improve handoff.
We discussed the outcome design, which resulted in a confusing user experience. In the intended design, the notes and files sections were separate and equal. In the delivered design, the hierarchy made it seem that the notes were a child of the file uploader.
Finally, we discussed expectations moving forward. We agreed that delivered work in the future would exactly match designs unless changes were explicitly agreed upon. After that one hiccup, we were able to move forward as a team, and I had finally won their trust.
In conclusion
This work is ongoing. We continue to have discussions with customers, and to work as a triad to collaborate on priorities and solutions.
One of my devs put working with AI in the best words: “many people think that AI is smart. But really, it’s like working with an army of really hard working dumb people”. In our discussions with users, it’s clear that they think AI is magic, and that it can solve all their problems. In fact, this creates a very fun playground for us to work in. We are able to explore what data and prompts will result in the outputs that will best solve the user’s needs!
