Phase 3

The line-of-business problem owner must agree that the model accurately represents the business problem and see that results are traceable and accurate and make sense.

Outputs must be surfaced in endpoints such as apps, websites, and dashboards. This usually requires a full-stack developer — another costly, scarce resource — to perform complex coding work, which can dramatically slow model deployment.

Models must be continuously monitored, managed, and retrained, because, over time, they will drift, resulting in less accurate and/or potentially biased results. As more models are deployed for automated decision-making, drift will happen faster than humans can detect, resulting in serious business and customer consequences.

Democratize analytics with no-code/low-code tools so more people can act as developers.

For example, SAS Viya’s no-code/low-code tools empower business analysts closest to problems to jump-start model development.

Make analytic results easier to interpret.

For example, with SAS Viya, non-technical users can right-click on any variable and view explanations in layman’s terms, get instant business context to understand the validity and relevance of results, and use embedded generative AI for real-time interpretations of data and graphs. 

Generate AI results that are highly interoperable so they can be quickly deployed in diverse endpoints without the help of a full-stack developer. 

SAS Viya enables one-click deployment of models into endpoints for shorter time to value realization. It also offers the App Factory, which enables users to embed model results into common endpoints such as apps and websites, using a built-in developer kit.

Preserve the value of models over time. 

SAS Viya continuously monitors models for degradation or bias, so IT knows when to retrain or retire them. It also provides report cards, complete with actionable recommendations on how to address detected biases.