There is currently a wide range of AI adoption by communicators from those who have not begun to explore use cases to those who have launched a fully functional AI model trained on internal data. While companies must explore ways to implement AI and remain competitive, the path to a systematic integration is not clear-cut for all companies as privacy, lack of structured internal data, hallucinations, and the need to upskill remain barriers to usage.

The following are some key insights from our recent roundtable, What’s the Latest on AI in Communications?, held in partnership with Takeda Pharmaceuticals, where we discussed adoption, challenges to integration, measurement, and other use cases, and leading teams through this transformation.

Trusted Insights for What’s Ahead™

• Leaders should explore what AI can and cannot do to inform their strategy, innovation, and day-to-day work. Currently deployed use cases for Gen AI include aiding employees with tasks such as content creation, supporting customer service associates and chatbots trained on past conversation transcripts, measurement of reputation and sentiment, key message pull-through, and competitive analysis.
• While AI can enhance work, it does not replace people. Upskilling talent will play a crucial role in supporting AI integration.Prompt engineering will be a key skill to seek and develop to leverage AI. Sandbox environments can be conducive to learning and experimenting with AI. Additionally, establish clear expectations with employees about compliant and safe AI systems, including protecting proprietary data.
• The key roadblocks for AI integration are suitable AI tools and AI skills on the team. Regarding tools, there is a lack of structured internal data to train AI models. Regarding AI skills, making employees comfortable using AI and upskilling them, also by experimenting with use cases, is key to a productive AI adoption. Additionally, effective change management is necessary to navigate the fear of job changes and entrenched workflows.

AI use cases in communications have the potential to significantly improve productivity and capabilities

AI’s potential to improve individual or team productivity varies drastically based on the type of work and how the team deploys AI. Some roles will benefit more from AI, particularly those focused on content creation such as developing briefs, coding, and brand positioning (scoring content for brand voice and compliance), or those with a lot of menial and process-heavy tasks. As such there is currently a lot of variance in estimates of AI’s impact on productivity. We heard estimates on improvement in productivity from 30% to 500% in the roundtable, which may differ based on the specific task, function, and industry.

Following are some of the current AI use cases that were discussed during the roundtable:

• Internal Gen AI tools can serve as junior assistants for ideations, creating email drafts, etc.
• Improving customer service – AI-powered chatbots have seen a lot of success, providing quick and effective customer service with the potential for features such as translation. Such chatbots can be trained on company-specific information and past customer service chat and call transcripts. In addition, transcripts of past customer interactions can serve to create AI-based support for customer service associates tailored to the topic of a specific customer interaction.  
• Measurement and analytics – AI use cases in measurement have the potential to be revolutionary, particularly for bulk analysis of large amounts of text information and cutting through the noise quickly, which can be applied to news and social media monitoring. Communicators can use AI to build better metrics for tracking reputation and sentiment analysis. For example, large commerce websites currently use AI to extract themes from thousands of written reviews about a product.
• Key message pull-through emerged as a metric that can be particularly enhanced using AI. Prompting AI to analyze content and pull key messages from drafts of press releases and other content can help communicators see what stands out, compare their messaging to competitors, and gauge whether they are reaching their audience with the intended message. For instance, AI models can be used to rewrite a press release to see if the intended message is coming across or if it didn’t land in the expected way. Additionally, internal AI models can be used to automatically flag content that deviates from company guidelines or to create content that aligns with the company’s core philosophies and values.
• Competitive analysis – In addition to synthesizing large amounts of internal data, AI models can be used to monitor and distill information about competitors and the market efficiently. For example, it can be used to extract key information from a competitor’s website such as their product offerings and messaging, as well as track conversations on certain topics with less manual effort.

In addition to exploring potential use cases, understanding what AI cannot do is just as important. An example of what AI cannot do well for communicators is strategy. It won’t be able to identify the business problem or how to address it but it can provide valuable tools to tackle the challenge and measure progress.

Talent strategy will play a crucial role in supporting AI integration

Investing in upskilling employees and hiring the right talent considering the opportunities and potential that AI can bring is important to stay competitive. Change management will be crucial as AI brings unprecedented change to the way we work and the overall talent market.

• Learning to use AI and experimenting with it are key to finding the right use cases, improving productivity, and ensuring employees are equipped to leverage AI for their work. Sandbox environments for people to experiment can be a powerful learning tool. Some leaders are asking teams to think about ways that AI could help them before embarking on new projects which can prime employees to consider it a part of their regular tool kit, and not just a novelty.
• Crafting prompts to get the desired output from the AI model requires the user to have a strong command of logical writing, rhetoric, and context.  Prompts for Gen AI models are more complex, specifying context and the question in a natural language format, as opposed to the keyword search prompts used for searches that people are used to. Prompt engineering, which may sound complicated but is simply this process of crafting prompts to input into AI models, will be a necessary skill across functions as Gen AI becomes more integrated into workplaces.
• Establish ground rules for AI usage. To reduce risk for the company, there should be zero tolerance for non-sanctioned AI systems and guidelines to ensure that everyone uses only approved platforms that have been investigated for compliance with company policy. Some teams are experimenting with de-identified information while finding the right tools to integrate. In the long run, ensuring employees have access to the AI tools they need can help reduce the risk of them using unauthorized systems.

Primary roadblocks for AI integration include structured data, resources, and change management

While most communicators agree that integrating with AI will enhance their productivity, for many companies, there is a lot of work to be done before they can transition to an AI-powered workplace. A small fraction of companies have integrated AI. Following are some of the common barriers companies are facing:

• Clean, structured internal data are crucial for many AI use cases, particularly for creating AI models tailored to the company’s needs. Companies are training AI models on their internal data and creating chatbots similar to ChatGPT to sidestep common issues with the public models including privacy concerns, hallucinations, and violating copyright law. Overall, this creates a far more efficient tool for usage within the company since it requires less rework to align with company-specific information. The quality of data is key regardless of how competent the model is – garbage in, garbage out. Companies that invested in improving their internal data collection and storage during the age of big data over the last decade or so will be at an advantage.
• Resource allocation – AI integration can be a resource-heavy process requiring experimentation, money, and upskilling. Larger companies have an advantage in terms of resources they can allocate to invest in customized AI integration. Yet, there are also off-the-shelf AI applications that can be tailored to some extent to companies’ needs. Currently, around 10% of companies are scaling one or more Gen AI applications across functions, while 40% are piloting AI projects to test its value and 50% have not yet taken action.
• Change management – Fear of job displacement, resistance to change, and entrenched workflows can hinder experimentation and adoption of AI. Additionally, misunderstandings by leaders around what AI can and cannot accomplish can be a roadblock to integration and finding the right use cases. Along with investing in AI, companies need to invest in AI learning and effective change management to avoid pushback to the rapid pace of innovation we are experiencing.