Ai search onboarding


Leo Sarmento
3 min readDec 17, 2021
SmartHint's new AI search engine onboarding process

Working for a startup is completely different than working for design shop or a big client. Shorter design cycles, smaller cross functional teams and more hands on with the final product being tested directly in production.

Smarthint is a Software as a Service (SaaS) company that provides a service that generates products recommendations for e-commerce platforms using Ai and browsing data. Their team is in full creative mode all the time with freedom to propose new products and features.

Use data to ideate, materialize insights, measure to generate data and repeat

Think, do, check and repeat. This is the usual development cycle for a SaaS. Regular improvements releases and constant evaluation with measurement of success through business results.

This is just one example of those improvements that their devs are constantly cooking up.

The challenge

The product team, I mean, developers were about to deploy an Ai powered search engine and there were some hurdles to overcome in the install and activation process.

This was supposed to make a huge impact in the competition and be a big selling point to new and current customers.

Although the search engine and its configuration panel wasn't market ready they needed more data to improve the Ai.

learning models behind it. So an early deploy was necessary. Clients would surely need a hand to help installing, setting up and activate this unfinished new feature.

The process

The search engine was already being tested by some hand picked current users of their software. The necessary information regarding everything about the new feature was reachable.

It was only a matter of talking to one of those clients to gather data to compile a script and do some IDI's (in depth interviews) with the remaining testers.

The main finding was the difficulty to setup and activate the feature on Smarthints control panel. The feature itself was working and didn't need any immediate adjustment.

The solution

The ux team (of one, in other words, me) have decided to make a step by step guide and coach the user through the whole setup process. An onboarding process that would keep the user informed during the constant back and forth between the developers and the client’s team. Some steps would be critical and compulsory, while others more simple after the feature activated would be optional and could be skipped.

Also it was important to show how the feature would behave on the client’s online store before it was put on production. As the feature could be activated and deactivated using tags on the URL, we’ve figured out a way to continue the onboarding guidance during this test right on the client’s e-commerce website.

See it in action :


By the time the Ai search was released Smarthint had around 15k stores using its core services. Not all had enough SKU's to fully take advantage of the new engine. Analysing the clients profiles, there was around 8k stores that could use the feature. It was expected that 10% be interested in the search, and would try it in the first month. Being able see it actually working on the store before it goes on to production, even if it took some days to the Ai to have it ready, the onboarding process put client at ease to install and test it. Almost 4k stores installed and tried the search engine in the first month. In the months following the launch, there was an 8% increase in the number of new customers.



Leo Sarmento

15+ years trying to understand people as an UX Specialist. Let me tell you some stories about my work…