How Salesforce Promotes Transparency and Trust in AI Applications

Pretty much everyone, both inside and outside the Salesforce ecosystem, is hooked on the latest AI advancements. Since OpenAI and ChatGPT burst onto the scene in late 2022, almost every single enterprise is investing in their AI capabilities and showcasing what they can do via artificial intelligence. Salesforce is no exception and while the CRM giant had been developing AI for years, it was at Dreamforce in 2023 that a string of ambitious AI-driven products were announced in full force. AI is enhancing and transforming our experiences of Salesforce, and the world around us, but many workers still don’t fully trust it. Arguably, this is one of the biggest barriers to AI adoption and to this technology achieving its full potential. Salesforce is passionate about ensuring its workers and the public develop greater trust in AI. In addition, Salesforce itself is a key player in developing AI technology worthy of this trust. Read on to find out how Salesforce is setting about building confidence in the most trailblazing technology of all.

Salesforce Trust and Compliance

Salesforce has long had a reputation for upholding the highest standards of security. Salesforce compliance certifications extend far and wide with the #1 CRM adhering to every major framework that exists including GDPR, HIPAA, ISO 27001, 27017, and 27018, as well as SOC 2. Salesforce has built a legacy as the gold standard of customer data because this information is comprehensive and most importantly, safe. In fact, Salesforce considers security so vital that they have invested $18.9 million in funding ethical hackers since 2015. These bug bounties have been awarded to hackers who can anticipate the actions of malicious hackers and so help Salesforce avoid potential threats.?

As the most trusted data source in the CRM industry, Salesforce is ideally placed to combat some of the misgivings around privacy, security, and trust in the new era of AI.

AI in Salesforce: Einstein Trust Layer

The Einstein Trust Layer is built specifically to support and improve the privacy and security of AI throughout the entire Salesforce ecosystem. This Salesforce AI trust layer comes with a comprehensive set of features designed to ensure that AI is leveraged responsibly in Salesforce. These features include:?

• Grounded prompts - this ensures that LLMs (Large Language Models) have reliable context to generate precise results. With Einstein Trust Layer this is achieved with secure data retrieval and dynamic grounding to ensure AI results are properly aligned with prompt intent.
• Crucially Salesforce does not store any of the context around AI prompts and this belongs only to customers. The LLM is designed to forget all context and one of Einstein’s core safeguards is zero data retention.
• Data Masking, which masks PII (personal identifiable information) and PCI (payment card industry information) when interacting with third-party LLMs is available in some geographies, languages, and for certain features. Unfortunately, Data Masking is not able to detect 100% of all data patterns across countries and regions so this feature still has room for improvement.
• Salesforce has also added toxicity scoring which prevents toxic or harmful outputs by AI. It scores the toxicity level of outputs and logs them in an audit in Data Cloud.
• Prompt defenses in the Einstein Trust Layer also stop potentially malicious outputs from being generated. These are grounded in system policies, which of course can vary depending on the context. ?
• This trusted enterprise AI architecture adheres to Salesforce’s own AI Acceptable Use Policy, which outlines the core values and principles of AI. This includes a stipulation that end-users are always made aware of when they are interacting with artificial intelligence.

In the Einstein Trust Layer, Salesforce and AI are working to implement the highest level of data security in an ever-changing world…but what else is Salesforce doing to ensure AI is trusted??

How is Salesforce Making AI Trusted and Ethical?

As AI in Salesforce continues to progress, the CRM has run several surveys to gauge worker sentiment toward this technology. According to a survey by Salesforce and YouGov across almost 6000 workers and nine countries (including the US, UK, Ireland, France, Germany, India, Singapore, and Switzerland), 63% of global workers would trust AI more if a human was overseeing it. While some respondents trust AI to perform tasks such as writing code and communications autonomously, most workers are confident when AI and humans collaborate together to carry out deliverables.

Salesforce is driving an AI strategy where humans are still at the “helm”, implementing checks and balances to ensure that technology retains its trustworthiness and integrity.

Ongoing Audits and Compliance Checks

Salesforce is committed to setting up ongoing audits and compliance checks for AI systems. Audit trails exist within the Trust Layer and beyond while Salesforce has pledged to ensure that humans intervene at critical stages in AI processes to make sure that systems are working efficiently and accurately.

Data Privacy Laws

As mentioned a little earlier in this article, Salesforce is already compliant with practically every important data privacy law. And this isn’t going to change anytime soon! Keeping in line with regulations such as GDPR and HIPAA is even more important in an era when machines are handling our data.

Monitoring and Maintaining Trust in AI Systems

Consistent monitoring of AI systems is also crucial to maintaining and building trust in how they function. For Salesforce, this requires a human hand to lead the process.

Feedback Mechanisms and Improvement Cycles

Interestingly, AI technology can be used to collect customer feedback on its own performance. It is, however, humans who need to check the final analysis of this data and do the hard work for AI to be positively received.?

Salesforce's Commitment to Transparency: Future-Proof and Universal AI Applications

Salesforce Automation has been taken to another level with the Einstein 1 platform and the flurry of activity around AI over the past year or so. While speed and efficiency are core drivers behind Salesforce Einstein AI, it is the steadfast commitment to trust, security, and transparency that sets Salesforce apart. They may have taken a slightly slower approach to AI advancements than some other major tech players but these steps have been cautious and accompanied by checks and balances along the way. While architecture like Einstein Trust Layer is industry-leading, it is unclear whether Salesforce (or any company) is fully prepared for the consequences of an even more intelligent or autonomous era of AI technology.

I guess we will all find out soon enough!