How we built and deployed an AI web application at scale for a large international corporation.

This client requested a complex data mining application that can be comfortably used by their Business Analyst teams around the world. These analysts are domain experts, but do not necessarily understand data science nor are they proficient with coding. Our solution was selected due to our flexible, yet time- and cost-saving, approach. Other solutions, in-house and commercially available options (e.g. SAP), have been tested, but deemed insufficient, inflexible, or simply too time-consuming to get to a point where proficient usage is possible.

Large organizations often come with special security requirements and applications have to be hosted on specific cloud platforms using approved services and data access methods. To fully leverage our solution, hosting on our secure systems should be preferred but, as sometimes this is will not be possible, we are glad that we can fulfill clients' needs.

A coarse timeline

We always start with a series of initial talks to get to know each other, get an understanding of the problem, and, most importantly, find a common language to describe the problem. From here, we will quickly develop a usable prototype/ proof-of-concept (PoC) and demo it to the client.

If approved, we enter a regular (usually weekly) cycle of refinements. While the overall idea and specifications are usually clear, the nitty-gritty details of the user-interface design are generally not and are inferred on the fly with product managers and target groups involved in discussions and regular meetings. Depending on the complexity and clarity this will take the bulk of the time. Some positive side effects of involving the target groups are that the engagement is higher and employees are much more likely to accept the new solution, and they get trained on the product on the fly.

Finally, we organized an assessment center with all decision-makers involved to get the final approval for the rollout. The various iterations of security checks usually precede the assessment center until company standards are met.

After rollout, we hold the service line and will be responsible to keep the product up and running as well as respond to any questions that may arise.