How Agentforce Guardrails Protect Your Brand

While there are areas of AI risk, such as brand safety and business compliance, I’ll preface this article by emphasising that 99% of the time, AI tools are incredibly helpful for driving customer engagement and streamlining internal processes. AI horror stories, though they gain a lot of traction online, are few and far between.

This is particularly true for Salesforce AI because the company has included multiple safety guards around its own AI features, including Agentforce.

The Einstein Trust Layer is one example.

The Einstein Trust Layer is a framework designed to ensure Salesforce AI tools are:

• Private: Data is protected and handled in line with global privacy regulations like GDPR.

• Reliable: AI models are rigorously tested to deliver accurate, high-quality recommendations.

• Ethical: Tools are designed to identify and reduce bias, ensuring fair and unbiased results. 

This gives you:

• Confidence in insights: You can trust that AI recommendations are based on clean, accurate data.

• Customer reassurance: Knowing their data is secure builds trust and loyalty.

• Compliance without the hassle: Salesforce AI helps you meet regulatory requirements without extra work.

When it comes to Salesforce Agentforce, which enables marketers to build marketing agents to represent their brand and handle much of the legwork alongside internal teams, Salesforce has introduced another safeguard; agent guardrails.

Think of guardrails as firm instructions on what your Agentforce agent can and cannot do.

They are designed to ensure responsible, secure, and effective use of AI in customer interactions. For example, guardrails can specify when an AI agent should escalate to a human, meaning your team can take over any complex or delicate communications with customers.

Guardrails exist to help marketers ensure that AI behaves in ways that are consistent with their goals, brand identity, and compliance requirements, while maintaining a positive, risk-free customer experience. 

That being said, here are some specific examples of how they do this:

1. Ensure compliance with privacy regulations

Marketers handle sensitive customer data and must comply with regulations like GDPR or CCPA. Agentforce guardrails can limit the AI’s access to personal information and ensure it only uses data allowed by compliance policies. For example, an AI agent can be restricted from accessing or storing sensitive data like credit card details.

2. Maintain brand voice consistency

Consistency in tone and messaging is crucial for brand trust and identity. Guardrails can enforce AI-generated responses to align with predefined tone and language guidelines, preventing the AI from using unapproved phrases or overly casual language.

3. Avoid unauthorised promotions

Ensuring AI doesn’t offer discounts or promotions outside those approved prevents financial losses and confusion. Guardrails can restrict the AI from creating or offering promotions that aren’t preconfigured, ensuring marketing offers remain accurate and authorised.

4. Minimise legal risks

AI interactions that make false claims or promises could lead to legal issues. Guardrails can help ensure that the AI’s messaging is truthful and accurate, reducing the risk of misleading customers and facing potential legal action.

5. Improve customer experience with contextual relevance

Marketers want AI to deliver personalised experiences without overstepping boundaries. Guardrails can restrict the AI from suggesting irrelevant products or making overly personalised suggestions that may seem invasive, maintaining a balance between personalisation and user comfort.

6. Control response to sensitive topics

Marketers need to ensure that AI doesn’t handle sensitive issues (like complaints or crises) inappropriately. Guardrails can be set to restrict the AI from engaging with certain sensitive topics or directing customers to human agents when complex issues arise, ensuring that the AI doesn’t escalate situations incorrectly.

7. Manage AI behavior and prevent over-automation

While AI is useful, marketers still want a human touch in certain situations. Guardrails can ensure that AI does not take over conversations entirely, automatically escalating to human agents when needed to maintain a personal connection and prevent over-automation.

Firstly, the Einstein Trust Layer mentioned earlier comes with built-in security features that help enforce guidelines for Agentforce agents. These include:

• Data privacy and protection.

• Access control and permissions.

• Logging capabilities that track all AI interactions.

• Bias detection.

• Monitoring for performance and ethical considerations.

• Customisable guardrails for agents.

In short, Salesforce does a lot of the hard work for you.

However, the onus is on marketing leaders to anticipate any brand or business-specific guidelines that are needed. Using natural-language instructions during the setup process, teams can input what the agents can and can’t, and these guidelines will be implemented.

Put time into documenting your Agentforce guardrails first, taking into account:

• Data compliance and regulations - particularly for industries handling sensitive personal data such as finance, real estate, higher education, etc.

• Brand voice and style - ideally, you’ll already have these documented.

• Seasonal promotions and approved discounts - including how and when these should be applied or shared.

• Crucial legal information - such as product claims you can or cannot make.

• Sensitive topics - list words and phrases that would require escalation to a human.

• Personalisation - what boundaries should be set to ensure marketing comms don’t go overboard with personalisation? Look at the data points you allow AI to use.

• Escalating to a human - review previous customer conversations to identify themes and required escalation points.

Once you’re confident your guardrails are solid, you can begin building your agent(s) by following the steps in Agentforce Builder.

Read next: How to Build Salesforce Agentforce Agents for Marketing.

For help setting up Agentforce guardrails, request a call with MarCloud. Our Agentforce support will help lay the foundations for reliable Salesforce return on investment.