AI-powered Language Translation Tool

One of the 5 major Canadians banks engaged my team to develop an internal language translation tool powered by an enterprise AI model. The objective of the tool was to to streamline the process and decrease costs on manual human translation, which was currently performed by third-party vendors.

I acted as the project lead and service design / experience strategist working to build an experience prototype in collaboration with tech and data teams to build the product and service experience.

Research & Current State Mapping

  • We first interviewed super users of translation services, who often send documents (such as tech documentation, training documents, and internal team communications) to the translation vendor to be translated into French

  • Even though we wanted to decrease the use of the vendor and cut costs, it wasn’t going to be possible to eliminate use of them altogether so it needed to be a solution that would still enable this partnership. So, we interview third party translation team understand how they ingest documents and do human QA and translation

  • Using these insights, we were able to create a high-level mapping of the current process and touchpoints to define a target state product experience

Product Design

  • Wanted to identify a use case that would be the best suited to demonstrate how the product would work and designed the product user flow from there

  • The flow needed allow for users to input text, be translated with AI, but in some cases, use the vendor’s own technology, and have a human-in-the-loop check

  • The proof of concept we built needed to be something that would communicate :

    • Meets business needs - translate the right things to actually save time - (an executive communications would be a bad example, because it should still be translated by a professional human translator) but meeting notes can be translated with AI

    • Actually streamline processes - have human and AI touchpoints at the right part of the process and be as efficient as possible

    • Technically feasible - fit into the enterprise tech ecosystem and they could building right away, as efficiently as possible

    • Overall, clearly illustrate the value to executives to get their buy in and secure more funding to bring the product to life

  • Worked with UX/UI designers to create a clickable prototype to demonstrate how the product should look and function

    • We then tested it with users and iterated on the UX, validated with vendor partners to ensure minimal disruption to current workflows

    • Ensure adoption so that it would not become a costly investment with no payoff 


Outcome

Overall this was a very fun exciting experience designing an AI product to do what most AI products promise to do - streamline a process, save money in doing so by using human talent to do the things we need it to 

  • This was executed in a really short time frame - around 6 weeks, starting from research and identifying the use case, through to passing off the clickable prototype and story starters to their tech and data team for them to take this forward - while also socializing it across parts of the business, preparing for steering committee, and continuously getting buy-in on the project

Back to Portfolio