Responsible AI: Cresta’s approach

A recent study from Mitre reflected that while conversations about AI are everywhere right now, actual trust in AI technology remains low: only 39% of US adults responded that they believe AI technology to be safe and secure. 

At Cresta, we believe that as we stay on the cutting-edge of innovation in generative AI technology, it’s our responsibility to do so in an ethical and transparent manner. Earlier this week, our CTO Tim Shi was joined by Cresta Product Marketing Manager Jess Stallings to host a webinar on what responsible AI means to Cresta, and our approach to ensuring the highest standards for the companies we work with. 

Below, we’ll walk through the four key pillars of our approach to ethical, responsible AI. You can also download the webinar on-demand today to hear more about these from Tim and Jess. 

Defining responsibility: Four key pillars 

These four principles guide our development and deployment of our technology, ensuring that doing so responsibly stays as the top priority. 


This pillar breaks into two separate layers: the model and the application. As Tim explains, care must be taken to ensure that the model is unbiased – that it is trained on a large enough and diverse enough dataset. Models can be further refined and customized using each individual customer’s proprietary data. 

The other element of fairness takes place at the application layer. This is ensuring that agent populations are being made more successful by the AI, taking into account diverse backgrounds and skill sets. For example, AI provides tremendous value to new agents in the ramping up process, serving up the right knowledge at the right time, and coaching on best practices. For tenured agents, the value from Cresta’s generative AI may look more like automation: making workflows more efficient, handling repetitive tasks to free them up for more strategic work. 

Our agent-facing desktop application can further support the needs and preferences of different individual users by making the components and layout of their workspace customizable. 


The second pillar is transparency; we clearly document the purpose, capabilities, limitations, and potential risks of our AI systems as they continue to evolve. As a matter of policy, this information is available to our customers. Cresta also invests heavily in R&D designed to make it easier for humans to interpret and understand the outputs of our AI models, including cutting-edge techniques like Chain-of-Thought Reasoning (CoT) and Model-based Critique. 

With CoT, we are exposing the model’s reasoning process to the user, allowing for human correction or guidance. This transparency builds trust between the product and the user – and ultimately makes the product more accurate. 

Privacy & ethics

Our third pillar is about the strict ethical standards we adhere to. Our generative AI models are designed to provide assistance to agents and managers that improve the efficiency and quality of customer and prospect conversations, but our platform does not make recommendations or predictions based on sensitive personal information, such as medical or financial data. 

Quality optimization & risk mitigation

Cresta takes a comprehensive approach to assure high quality and low risk, which can be broken into five components. First is data-driven design. By combining KPI data with conversation data, Cresta’s AI models provide valuable assistance to agents, supervisors, managers, and executive leaders – helping them enhance their productivity and deliver exceptional customer experiences. Learn more about how this works in our full manifesto

The second component involves deployment in stages. Working with Cresta and deploying AI across the enterprise is not a one-and-done situation, and we ensure that every stage is done thoughtfully and iteratively. 

Third is human-in-the-loop quality assurance and optimization and gets to the heart of what we believe at Cresta: AI’s role is to enhance the human, not to replace them. Our human-in-the-loop quality assurance measures involve continuous monitoring and evaluation of our AI models by experienced professionals. 

The fourth component involves implementing guardrails along every step of our processes, both internal and with customers. To read more about this in detail, please read our full manifesto

Finally, foundational to Cresta’s work is looking to the future, continuously evolving and innovating; Cresta is committed to ensuring that our customers always have access to the latest breakthroughs in AI technology.

To dig more into our vision of responsible AI and how it impacts the customers we work with, download and share the webinar replay today You can also read more about our commitment to responsible AI, you can read our full document here

Build vs. Buy: Pros and cons of building your own generative AI solution


Security as a competitive advantage: evolving with the threat environment


John Donovan on the Potential of Large Language Models in the Enterprise