By David Livermore and Okthalia Adams

The speed and efficiency provided by ChatGPT is a blessing and a curse. It can help managers draft emails, talking points, and even lengthy proposals in a fraction of the time it takes to do it alone. But it’s easy to end up with content that is not quite right for the intended audience. For example, starting an email with an abrupt request may work in some contexts but be off-putting in others. A German engineering firm might expect a detailed, formal proposal while a Silicon Valley startup may expect something that is brief, conversational, and creative. Is AI smart enough to calibrate for these differences?

While 65 percent of organizations are now using generative AI, many managers lack the skills to use it effectively, increasing the risk that AI-generated communications may alienate customers and employees. IBM executive Rob Thomas says, “AI won’t replace managers. Managers who use AI will replace managers who don’t.” We propose an additional layer: Managers who use AI with cultural intelligence will outperform everyone else. Cultural intelligence (CQ)—the ability to relate and work effectively in culturally diverse situations—enables managers to prompt AI for content that not only avoids being offensive, but actively builds trust, strengthens connections, and drives success in an increasingly diverse business environment. Drawing from our research and experience with companies around the world, we offer three best practices for writing prompts that yield more culturally intelligent, effective responses.

1. Nudge the bot to consider the cultural context

AI defaults to the dominant cultural values embedded in its training data, often resulting in generic responses aligned with Western norms. To avoid this, managers can explicitly include cultural context in their prompts. For example, if a manager asks AI to draft a communication about a “return to office” policy, the output may default to a Western approach that announces the need to be back in the office while still emphasizing flexibility and individual choice. However, in other cultures, there may be a preference for clear, explicit guidance on in-office days, with less emphasis on individual autonomy and more explicit directions on what is expected. By instructing AI to consider the range of cultural expectations around work location and team dynamics, managers can produce communications that are better aligned with diverse values, leading to a smoother, less disruptive transition.

GlobalTech (pseudonym), a US-based company with teams and offices across the world, has been offering AI training to all of their managers. The training covers everything from using AI to draft communications and tackle complex problems to simulating interactions with hypothetical customers. It also teaches managers how to specify the cultural contexts for which they need content, including writing detailed personas that represent specific cultural segments.

In one session, GlobalTech managers compared the AI-generated responses when they requested help drafting a professional development plan for two different personas: a “young professional in Tokyo who is an avid gamer” and a “mid-career manager in Berlin who runs marathons.” By providing these nuanced personas, the AI gained a better understanding of the values, expectations, and work dynamics at play, resulting in more personalized and contextually relevant responses. Managers then developed a list of these personas to use as needed, copying them into prompts when relevant.

It’s also crucial to localize prompts. While humans instinctively understand that drafting an effective proposal for a small company in northern England differs from doing so for a global bank in London, AI lacks this local context unless it’s explicitly included in the prompt.

Here’s what a prompt might look like with and without cultural intelligence (CQ):

• Without CQ: “Draft a proposal outlining our services for a global bank in London.”

• With CQ: “Draft a proposal outlining our services for a global bank in London. Consider that this is a large, multinational corporation with a formal communication style. Use formal, professional language and emphasize scalability, innovation, and industry expertise.”

By nudging AI to account for cultural and regional differences upfront, managers can generate communications that better align with the expectations of diverse, global stakeholders.

2. Tell the bot what style and tone to use

AI can churn out responses quickly, but the tone and style may be inappropriate for the real-world audience. People focus as much or even more on how a message is communicated than on the content itself. This is especially true in high-context cultures like Japan and Korea, where much of the communication’s meaning is derived from what is implied between the lines.

GlobalTech divided its marketing managers into two groups. One group wrote their own talking points to announce a product launch to customers in Brazil, India, and Germany. The other group used Microsoft’s Copilot with a prompt like: “Give me talking points to announce a new product to our customers in Brazil, India, and Germany. The tone should be warm and collaborative for Brazil, formal and detailed for Germany, and emphasize innovation for the Indian market.”

A cultural expert conducted a blind review of the responses from both groups. They found that the AI generated talking points were better attuned to the cultural differences than those created by the managers alone. When prompted thoughtfully, AI can deliver culturally intelligent responses. By explicitly specifying expectations around style and tone, managers can guide AI to produce content that aligns more closely with the cultural and emotional norms of the target audience.

Here’s what a prompt might look like with and without cultural intelligence (CQ):

• Without CQ: “Review the tone of the following email, which addresses a difficult topic with an employee.”

• With CQ: “Review the tone of the following email, which addresses a difficult topic with an employee. The employee comes from a culture where saving face is important, so ensure the tone is constructive, avoids unnecessary direct criticism, and focuses on solutions rather than personal shortcomings, while still being clear.”

3. Call out the bot’s bias

Even when a manager includes culture and context in a prompt, the AI tool may still default to biases. For instance, a manager might ask, “How should I involve my team in Mexico in our decision about remote work?” AI might default to stereotypes, suggesting that people in Mexico don’t expect their manager to involve them in this kind of decision or asserting that they don’t want to work from home. While these norms may have been more common historically, they may not apply to the specific team members in question, especially if they’re from the younger generation or work in a more cosmopolitan environment like Mexico City.

A more culturally intelligent prompt is, “How should I involve my team in Mexico in the decision we need to make about remote work? Consider that their Mexican culture may influence how they want to be involved. But I also have team members in Mexico who may not reflect the stereotypes often associated with Mexican work culture.” This ensures that AI provides advice tailored to the diverse makeup for the team, avoiding potential missteps in how you lead.

GlobalTech trains its managers to recognize and challenge potential biases in AI outputs. For example, when an AI-generated performance review for a woman returning from family leave included phrases like “has returned to full productivity despite balancing work and parenthood,” it subtly reinforced gendered assumptions. Managers are taught to challenge biased responses by prompting AI with a follow-up like, “How would this language change if the employee were male?” AI can offer a more effective draft when reminded to avoid bias.

Here’s what a prompt might look like with and without cultural intelligence (CQ):

• Without Bias Check: “Draft a performance review for an employee returning from maternity leave who needs to work on XYZ.”

• With Bias Check: “Draft a performance review for an employee returning from family leave who needs to work on XYZ. Avoid language that implies balancing parenting and productivity challenges.”

Managers can also use AI to check their own biases. After drafting a communication, a manager can ask AI to evaluate the content for potential biases and assumptions. As a result, AI and managers can co-create content that is more inclusive, respectful, and effective.

Can AI be Culturally Intelligent?
Generative AI has enormous potential for leaders to streamline communication and improve operational efficiency. Yet, its true power lies in its ability to adapt and resonate with diverse audiences–if guided appropriately. To unlock this potential, AI tools must be trained and prompted with cultural intelligence. Managers who apply cultural intelligence when interacting with AI don’t just create more effective, inclusive content—they help shape the AI’s ability to generate more inclusive responses in the future. While AI won’t replace managers, those who wield it with cultural intelligence can lead more effectively, build deeper connections, and drive lasting success in an increasingly globalized world.

Okthalia Adams is an MBA candidate at Boston University. Okthalia has extensive experience as an AI product development leader and is working with me as a graduate research assistant.

COMING SOON! An innovative AI tool designed to coach you through all your cultural challenges.