5 reasons models of he need more than good intentions
The model of the model has been moved from “nice to have” to a “non -negotiable”. While organizations decide in industries such as health care, banking and the government, the demand for transparency, confidence and accountability is stronger than ever.
Sas Briana Ullman, product marketing manager and Vrushali Sawant, data scientist, discussed what it looks like in SAS INOVATE 2025. Their conversation epitomized the credible idea that was to do the right thing and build the right systems to show your work and support it.
Here’s what else I got:
1. Poor governance of it is a matter of security in the real world, especially in health care
Ullman began things with a sober statute: The Institute for Emergency Care Research (ECRI) listed the “insufficient governance of it” as a patient’s security concern in his annual report. This is ahead of many long issues in health care.
She also mentioned the American Medical Association, which found that 60% of doctors are concerned about him in health plans, especially denials of care, unnecessary waste and avoidable damage. Shares? Life, faith and equality in care.
“Only 16% of hospital executives have governance policies throughout the data system and he” noted Olman. “We have serious concerns, but a slow approach to readiness.”
2. We need a “nutritional label” for him
Ever tried to choose something healthy in the grocery store? This nutritional label helps you quickly understand what you will consume. Ullman argued that he needs something similar – standard, interpreable and accessible.
“We don’t have a nutritional label for him. And that’s a large part of the problem.”
For now, stakeholders technical and business stakeholders often speak different languages. Leaders want to make informed decisions, driven by the data. In the meantime, everyone has to do the job to translate into business value and prove that it is reliable.
This is a heavy elevator without better governance frameworks.
To overcome this gap, SAS has developed model cards – clear, standardized documentation that acts as a nutritional label for its models. These model cards help everyone involved understand the purpose, performance, risks and governance of the model, making responsible responsibility easier and more transparent.
3. Model Governance Bridge of Communication The communication gap
Governance is not just a checklist; Howt how we build confidence between teams.
Ullman clearly put forward: model governance empowers technical teams to show that their models are reliable, explainable and aligned with business goals. He also gives business leaders a way to prove the systems he without becoming overnight coders.
This is about transparency, accountability and clarity, not just compliance.
4. Data scientists need government tools that complement them where they work
Sawant chose the session with a direct demonstration to show how data scientists can introduce governance directly into their workflows using SASABOUT ViyaABOUT and Python.
Using Sas Viya Workbench, she launched a Jupyter notebook to go through a practical, reproducible example: training a model of decisions tree in the SH.BA registration data from the UCI machinery teaching warehouse to predict if an individual earns more than $ 50k a year. While the issue of use was simplified for the session, the process reflected the types of challenges that attendees often face when operating models in regulated or high environments.
It started with standard steps known to most practitioners – data cleansing, exploring variable distribution and engineering features to reduce noise and improve the efficiency of the model. But the concentration quickly shifted to what happens after a model is built: Establishment of governance in the course of work.
Sawant showed how to record the model in the SAS model manager, document its variables and assumptions, and pack it for placement using ordinary tools.
She stressed that the governance begins early, even when preparing data and engineering of features. Whether you are cleaning categorical variables, reducing possible prejudices or standardization of the workflow, those basic steps matter. From there, the process continues with the model registration, version control and creating model cards-all managed through SAS Model Manager, which supports both SAS and open-sourced models like those built in Python.
“Responsible innovation only occurs when every person involved in your life cycle acts responsibly,” Sawant said.
5. Governance is not just for regulators; is for resistance
Very often, governance is seen as something you do to check a box for legal reasons or compliance. But Ullman and the Sawant make the opportunity that the model governance is the way you escalate it safely and consistently.
What is it that enables repetitiveness. Ensures that models do not move without disclosure. It keeps you ready for audits-but the most important, for the consequences of the real world.
“We believe that the responsible innovation begins with responsible innovators,” Ullman said. “But how to get there? With the right frames, the right tools and the right mentality.”
Leave feedback about this