How can you prove your API-based data extraction tool is accurate and valuable when there is no conventional UI?
How it went
We needed to show-off the power of our API – which accepted documents, sent them off to machine-learning models for data extraction, and returned data in a JSON payload. Useful, but a workflow that is hard to visualize and parse within the context of a sales call.
I worked with a team of two engineers, another designer and a product manager to create a way for our sales people to walk people through a live demo of our product, and have the results be human readable at the end of the demo.
Our objective was to not just visualize a normally invisible process to help tell the story of our product, but create trust in our system with a demo that ran live with any documents the users wanted to share, and results we could share and send back to them.
It was the rarest of all starts to a project – I could tell my sales coworkers that they were indeed the user this time.
We started with research – I conducted user interviews with the people running the demo to better understand what they felt was important to show off and explain in a call. I also sat in on several sales calls as well to get a qualitative sense of what our prospective users wanted to see, and expected to see. We wanted to understand, as well, what could be a ‘surprise and delight’ moment in a B2B sales call.
We did look at what other, similar companies used for demos; either live or pre-recorded. We pulled on prior research we had conducted as well in trying to learn what pain points people had with technology like ours, and how we could mitigate that pain.
After some initial sketches, we had the groundwork for the basic needs of our sales people, how the demo needed to flow, and how we could surprise and delight prospective users. I drew out a journey map for the sales cycle to help understand the perspectives of all parties through the process.
We focused the demo experience on gathering some light data about potential user needs with a couple quick questions before uploading any document to be processed. This helped us understand what people needed from us, and also gave our sales people a chance to talk through the different options we provided.
We drafted out some wireframes to illustrate the intended path. Because the whole process could take several minutes, we had to have some visual information on system status, but not too compelling as to distract from any conversations happening between potential users and our sales people. Additionally – because no process is 100% perfect – we needed system notifications when our process was done, and a success; and when the process had failed.
Due to the pressing speed in which we were building our demo tool, for our first version of the sales demo we used images and messaging to share system status – with the intent on building something more visual and animated for future versions. Our first version does display our extraction results alongside the original document; and our sales people can toggle between the JSON payload and a table of results.
Currently, the sales live demo tool is running and was a key piece of our sales calls. There is now current work in progress to refine the user interface, improve the flow and make some enhancements to the back-end performance.
Update: BlueSuit was acquired in January 2023 – probably because the demo showed off our API so well.