Adl Consulting is an advisory, public affairs and institutional communication firm specialized in lobbying activities, advocacy and change management. Since 2012 we have been fostering data-based decision-making and promoting Digital Lobbying methodology within the field.
The “Digital Lobbying Blog” is a digital space where our Team makes available to all sector professionals insights on emerging national and international issues, innovations and best practices.
What role can Generative Artificial Intelligence (AI) play in Public Affairs? After identifying a range of tasks and responsibilities of public relations professionals, we sought to understand how they might be supported, facilitated, or automated using Generative AI tools.
Artificial Intelligence (AI) is all the talk. Across various sectors, artificial intelligence raises crucial questions about improving work processes and potentially replacing professionals.
The impact is set to be profound: according to OpenAI, the parent company of ChatGPT (Eloundou, T., Manning, S., Mishkin, P. and Rock, D., 2023), AI will partially or completely affect 80% of the US workforce, while according to Goldman Sachs (Hatzius, J., Briggs, J., Kodnani, D., Pierdomenico, G., 2023), over 300 million jobs in the United States are at risk of automation. Specifically, the administrative, legal, communication and public relations sectors are predicted to experience the most disruption. Additionally, while previous technological revolutions have primarily automated manual jobs, AI is poised to predominantly affect knowledge-based professions.
Within this context, AI may be perceived solely as a threat to employment, especially in the public affairs field, which is our primary focus. However, this report aims to offer a fresh and new perspective. Indeed, at ADL Consulting, technology is the linchpin of daily operations as we highly value innovation for public affairs. Particularly, we conceive knowledge management platforms or AI for digital lobbying not as mere artificial replacements of human beings, but rather as powerful tools to enhance, support and help professionals in their decision-making processes. To sum up, AI is a support for augmented intelligence.
Drawing on this background, we face the AI challenge with mindful enthusiasm, seeking to understand how Generative AI systems, like ChatGPT, can facilitate and streamline the daily work of public affairs.
As stated in the article, we acknowledge that AI is a broad field and that Generative AI represents just one aspect of it. However, we focus on Generative AI because of user-friendliness, widespread availability and our target audience of public affairs professionals. Future research and analysis will explore other AI domains not covered in this report.
The report is organized as follows: after a theoretical introduction of Generative AI, we lay out a range of practical use cases of AI, including policy paper writing, image creation and text analysis. Finally, we conclude with a brief evaluation of future scenarios.
In recent months, there has been fervent discussion on news sites and social networks about ChatGPT and the “bizarre” yet plausible images generated through artificial intelligence. As previously noted, these tools only represent a fraction of the broad spectrum of artificial intelligence, specifically referring to what is known as “generative” artificial intelligence.
Generative AI is a subset of AI, particularly within Deep Learning and Machine Learning technologies, facilitates the creation of diverse content types, including texts, images, music, graphs, slides, and more. Generative AI systems are trained on vast amounts of images or texts and employ techniques like Convolutional Neural Networks (CNN), Natural Language Processing (NLP) systems for images, or transformer models for texts (like GPT-4) to generate new content from training examples.
Beyond the numerous benefits in terms of content creation, it is however crucial to mention the existing risks. Specifically, the datasets training AI systems can contain incorrect information, and sensitive data, or replicate human biases; these flaws are then mirrored by AI systems in content generation. In other words, the quality of the AI systems’ output depends on the quality of the input data used for training.
In the public affairs realm, it is particularly important to keep these potential risks in mind, acknowledging that content produced by AI systems may not be accurate or, even worse, may contain false information and hence could be deliberately used to generate fake news.
Moreover, regarding ChatGPT—which will be important for our analysis given its popularity and ease of use—it is noteworthy that the tool currently deploys two different versions powered by two models: GPT 3.5 and 4, with remarkable differences in their datasets and computational power. For our analysis, we have adopted ChatGPT Plus, which allows for the use of GPT-4, along with a set of plug-ins with a wide range of features.
Lastly, to avoid any saving of the entered data for potential chatbot training and, therefore, to ensure data privacy, we have used ChatGPT with the “Chat history & training” feature disabled. This feature was globally enabled by OpenAI in April 2023, following a ban imposed by the Privacy Guarantor in Italy.
With reference to Generative AI, we selected a set of tasks carried out by public affairs professionals that AI can enhance, facilitate and support. For each of these uses, we have then identified suitable AI tools and conducted test cases, which are detailed below.
Despite being aware that the range of public affairs activities is significantly wider, the tasks we have identified are the most recurrent and time-consuming ones. Thus, support in their execution or possible automation represents significant opportunities. Finally, reducing the effort on these activities allows for reallocating resources to other more qualitative activities requiring emotional intelligence, sensitivity and specific understanding from professionals.
Practical Uses of Generative AI
For this task, we have used Bing, the chatbot developed by Microsoft and powered by GPT-4, the most advanced version of OpenAI’s language model. As a “Generative Pre-trained Transformative”, GPT, once prompted by the user, can generate textual answers drawing on the massive amount of data it has been trained on. In this regard, text generation with a wide array of styles is the main application of the chatbot. Bing was chosen over ChatGPT, OpenAI’s chatbot, because of its browsing capabilities that ChatGPT currently lacks. Additionally, Bing allows the use of the GPT-4 model free of charge.
For this specific case, the envisaged scenario is the drafting of a policy paper on biometric surveillance AI technologies for a tech-sector company.
As detailed in the following video, we provided the chatbot with a well-detailed prompt so that AI could draft a plausible outline for the policy paper. Then, to receive the text of the policy brief, we entered a set of prompts such as “Now write the section on [title]. The tone must be concise and structured in an easy-to-understand manner” where in [title] we have entered the titles provided by Bing for the outline in response to the first prompt. With this method, it is possible to obtain specific and detailed texts for each of the paragraphs of the structure outlined by Bing.
The final result is quite satisfying: it identifies the key points on the analyzed topic and, by using the web browsing function, also provides updates on the current state of regulation. However, it is also possible to detect a certain vagueness and generality in the arguments developed, especially in the policy proposal section. In this vein, we imagine that the texts produced by Bing for similar tasks can be adopted by public affairs professionals as a baseline in the writing process, particularly in the informative sections. Professionals are then expected to fine-tune the final result.
Generative AI can significantly support and streamline the activities of reading and analyzing articles. To this end, we used ChatGPT with the plugin “Link Reader”, easily installable in the Plus version. This plugin enables text reading and carries out diverse tasks, including keyword extraction and executive summaries.
In our case, as shown in the figure below, we asked the chatbot to read the article Who Is Going to Regulate AI? and to provide a summary and a list of key concepts relevant to a public affairs consulting firm focusing on the digital sector.
Although the provided summary is excellent and consistent with the article, the bullet points are rather general. Such use can significantly help a public affairs professional, but it is worth pointing out that a basic knowledge of the topic of the articles and, above all, the ability to formulate the right questions through prompting – i.e. relevant to what is of interest and appropriate to the articles under examination – are essential for proper use of the IA tool. Ultimately, this plug-in is a very useful tool for reducing reading time and speeding up the summary and revision tasks that are often of major importance.
For this purpose, generative AI can be of great utility through the plugin “AI Pdf”, installed in ChatGPT. This plugin enables access to PDFs stored on Google Drive or other repositories and can perform summarizations, analyses, and various other processing tasks.
In this specific case, having uploaded the policy paper on biometric surveillance, we asked the plugin to provide a summary and, then, draft a press release. Below are the results.
The outcome is quite positive: the summary effectively grasps the essential points, while the press release provides an excellent basis for subsequent and necessary fine-tuning by a professional. We envisage this use as particularly useful as a support in reading and summarizing extensive documents for research, monitoring and policy analysis.
Content creation, whether in the form of texts or images, for social media is one of the predominant applications of Generative AI. Specifically, AI can effectively enhance copywriting, adjusting the tone, the style and the target audience for different platforms. This use can be of critical importance both for consulting firms in their in-house activities and for supporting clients. In this regard, Copy.ai stands out as an optimal tool, designed to craft compelling social media content.
In our test, we prompted the tool to generate two distinct texts for two different posts on LinkedIn and Instagram aimed at promoting the policy report on biometric mass surveillance. In particular, we wanted the tool to tailor tones and styles to different target audiences of the two platforms.
The texts provided is good, answering all the points of the detailed prompts and, most importantly, adequately adjusting the tone to the different platforms of use, i.e. LinkedIn and Instagram. However, one can also detect an excessive sticking to the prompt and, thus, a lack of ‘creative’ capacity. In general, though, Copy.ai can be quite useful for providing source texts for further processing by communication professionals or even of great use to public affairs professionals less familiar with social media formats.
Among the uses of AI in recent months, image generation is certainly one of the most popular. Tools such as Midjourney, DALL-E and Stable Diffusion allow image generation from a simple textual description entered by the user. In public affairs, these images can be used as covers or illustrations of policy papers, in newsletters and social media. The potential is impressive: anyone can produce images or graphics to enrich their work. For our analysis, we choose Midjourney 5.2, a tool that is accessible via Discord paying a monthly fee.
We asked Midjourney to generate a graphic serving as a cover for the policy paper on mass biometric surveillance mentioned above. Here, the prompt text has to be particularly accurate, indicating the subject, style, colours and angle of the desired image.
Midjourney generates images of high quality and responds well to the given prompt. Generally speaking, the use of a tool such as Midjourney can be helpful if the prompts accurately describe the desired image and technical characteristics. However, the state-of-the-art, Midjourney and other image-generating software allow a low level of customisation, so they may not be suitable for those seeking clear and customized graphic style. Furthermore, it is essential to point out that a good command of prompting techniques is vital to achieving a satisfying level of images.
Creating effective, concise and graphically appealing slides is yet another essential task for public affairs professionals. Here again, AI can assist, as it enables slide generation from a simple text or even a simple prompt on the topic. Among the various interesting tools, we chose Gamma‘s ‘Text to Slides’ function.
Specifically, having uploaded the entire text of the policy brief on biometric mass surveillance, we asked for summary slides. Gamma autonomously retrieves the relevant text snippets, organizes the information and creates the layout of the presentation. Afterwards, the slides can be edited either manually or with the help of an artificial intelligence chatbot. Below is an explanatory video of slide generation with Gamma.
Although in the beta version, the tool produces a promising result in terms of extracted information and graphic organization. However, it overlooks some important elements and creates some graphics inappropriate for the purposes they should serve. Nevertheless, they provide a good initial organization of the information and good visuals, facilitating and speeding up the initial processing and organization phase for a public affairs professional, who is still required to contribute significantly for a satisfactory outcome. As of now, we imagine the use of similar tools as support. However, it’s important to note that rapid technological development will soon take this functionality to a very high level so that a few slight customisation changes to the generated slides will be enough. Incidentally, such functions are likely to be integrated into traditional slide editing software such as PowerPoint within a matter of months.
Nowadays, writing texts in English is a priority for public affairs professionals to keep up with the rising interconnectedness of services and global dynamics. However, as English is not the mother tongue for many, it can be of utter importance to resort to traduction or paraphrasing tools. In this regard, QuillBot is a precious resource, allowing for text paraphrasing with clarity and conciseness.
In this case, we asked Quillbot to paraphrase and enhance a text written for an international client.
The paraphrased text is appropriate as it improves parts of the original text that may not sound natural. All in all, this tool can significantly help public affairs professionals, speeding up the writing process and improving the quality for better fluency in the English language. However, it is worth noting that the provided text is already of good quality.
The article offers an essential but meticulously reasoned review of the current applications of generative AI in executing crucial tasks within the public affairs sector. Having identified several tasks that AI can support, streamline or automate, we have selected easily accessible online tools and conducted a series of rational simulations.
As of the state-of-the-art, it is not easy to determine the impact of generative AI on the public affairs sector. There are, thus, several factors to consider, such as the rapid development of these tools, increasing investments, as well as and the extended adaptation periods required by complex organizations—such as institutions or large companies—to new processes and substantial technological shifts. Just a year ago, no one could think of delegating the writing of a policy paper to an AI chatbot and obtaining a satisfactory result. Now, it is perhaps even more complex to understand whether and when it could become the norm.
Yet, as proved in this article, the opportunities of AI tools are relevant, making it increasingly critical for public affairs professionals to keep up with their rapid developments.
As a result, it will be crucial to rethink both the specialized training of professionals as well as the processes within the public affairs sector. We recognize that the leading role and expertise of professionals will be pivotal to fostering a valuable interaction with artificial intelligence. Thus, we do not envision a human-machine competition but, rather, a collaboration enhancing the quality and value of human work, as witnessed in previous technological revolutions.
Lorenzo Ancona, Giulia Geneletti
Marialessandra Carro, Valentina Cino, Pietro Quercia
Communication & Digital Lobbying Strategist presso Adl Consulting
Policy Consultant at Adl Consulting