Algorithmic Allure: The Rise of AI and the Future of Media Analysis

Artificial Intelligence is here, and in theory, it should scare me. As a writer and communications professional, the idea that there are systems in this world with capabilities to analyze vast amounts of data on any topic, write press releases, op-eds, and other basic copy tailored to any audience, in any style, tone, or language, and offer valuable media monitoring and analysis, all before I’ve had the chance to finish typing this sentence, is well, daunting.

But also exciting, in a certain “watching the 2024 presidential election through my fingers because I can’t look away” kind of way.

Experts on AI have identified three general categories when it comes to a person’s perspective on its use: bloomers, gloomers, and doomers. In The Threshold: Leading in the Age of AI, Dr. Nick Chatrath describes bloomers as “people who believe that AI adoption will follow a similar path as previous industrial revolutions,” “gloomers see the encroachment of AI into creative domains as the first step of a robotic invasion that will soon wipe out a wide swathe of jobs,” and doomers are downright apocalyptic about the future of AI, stockpiling guns, gold, and gas masks in doomsday bunkers in anticipation of the great machine rebellion a la The Matrix.[1] But unlike the movies, Keanu Reeves and his floor-length leather jacket aren’t going to save us. Hopefully, it’s because they won’t have to.

If developed, regulated, and utilized correctly, future machines won’t become sentient, intelligent operators using our bodies for energy. Instead, AI’s power will be harnessed for the benefit of humanity, preventing the industry and its capabilities from spiraling out of control.

Welcome to AI (It’s Been Waiting for You) (Alexa’s Version)

When I began my investigation into the value of AI, I would have classified myself as a moderate gloomer, but as of this writing, I can confidently say I am a cautious bloomer.

In truth, I used AI throughout the entire process of compiling this case study. If I was going to write about AI, I at least had to experience AI for myself.[2] From transcribing notes to data analysis to summarizing PDFs, there was a tool for almost everything. With AI’s help, I was able to extract the key themes, insights, and recommendations I needed much faster than without. Not every question yielded perfect results, and it took some trial and error, but through my interviews, research, and practice, I uncovered the true depths of AI’s capabilities.[3]

Ultimately, I decided to turn to AI for the same reason my colleagues did when conducting their first-of-its-kind AI-powered media analysis: to test its capabilities. When presented with a significantly larger sample of data than any traditional method would allow, could AI prove to make life easier for future client services, or would the end result be a load of gibberish?

Harnessing AI for Media Analysis

Tackling some of the biggest issues due to a changing climate, Environmental Defense Fund (EDF) sought to amplify coverage of its farm finance and agricultural carbon markets work in high-profile, mainstream, English-language media outlets and influential financial sector trade publications. To help achieve this goal, Signal Group worked closely with EDF to conduct a comprehensive survey of the media landscape surrounding: 1) climate change, 2) the agriculture sector, and 3) financial solutions, carbon markets, and financial risks – the intersection of all three being the main point of interest.

Traditional media analysis and monitoring approaches lacked the capability to survey such a large scope and analyze trends that would help increase EDF's share of news coverage on its climate-smart agriculture work. So, to tackle this obstacle, Signal proposed a groundbreaking approach – AI.

Instead of shying away from the challenges that can arise when employing an unfamiliar product, EDF was drawn to the innovative approach. April Ann Opatik Murray, senior communications specialist for climate-smart agriculture at EDF, noted that “AI is a very powerful tool,” for conducting media analysis, and “when used in the right setting, is a perfect example of using technology to its maximum potential instead of getting used by it.”

Embracing the possibilities, Signal leaned into AI’s potential, capitalizing on its capabilities to enhance efficiency and provide insightful data that could lead to strategic media recommendations tailored to the client’s needs.

From Data Deluge to Strategic Insights

To effectively capture the media landscape, Signal partnered with financial data science consultancy firm, Conlan Scientific. A total of 34,563 articles from 21 leading publications were analyzed via custom web-scraping[4] tools, an essential step as media websites are often built and organized differently, emphasized Chris Conlan, CEO of Conlan Scientific. Finding topic-specific articles from myriad publications required not only investigation into how each website functioned, but also trust that the news organizations at hand had properly defined what coverage was and was not related to the client’s requested topics.

Enter OpenAI’s[5] API,[6] a revolutionary platform with the power to transform the way we work, learn, and consume information. With the data extracted, Signal began processing the articles through GPT.[7]

To distill the data, Signal worked with EDF, data scientists, and analysts to create 20 custom-built questions to ask GPT of every article, seeking only those explicitly related to climate, finance, or agriculture published within the last two years. The resulting dataset was 17,011 articles – almost 50% of the initial articles identified in the scrape were scrapped.

Dedicated to ensuring maximum data quality, it took considerable refinement to develop the 20 most optimal questions that would result in insights that could be translated into actionable recommendations. The final questions were aimed at filling in details surrounding the various aspects of each article, such as who wrote it, where it was published, and what subject matter it would classify the article as (climate change, finance, agriculture, etc.).

By utilizing GPT, Signal efficiently cleaned and organized the remaining 17,011 articles to create a comprehensive media resource for the last two years, which was then analyzed to identify valuable trends and key insights. The experienced team of media relations experts assisted the data analysts in interpreting the findings and translating them into actionable recommendations that would aid in EDF’s goal.

Armed with the combined insights from GPT and expert analysts, Signal provided EDF with a clear roadmap for navigating the media landscape. The thorough media strategy encompassed not only EDF’s current needs but also offered invaluable insights for shaping future media relations initiatives, highlighting the necessity of combining AI analysis with human expertise.

And just like that…Signal’s inaugural AI-assisted media analysis was complete.

Cultivating Continual Collaboration

After mandatory celebrations[8], the would’ve, could’ve, should’ve’s began to reveal themselves. We learned that it’s not as simple as just gathering data and running it through AI. Navigating the intricacies of AI analysis is akin to playing the Game of Thrones[9] – it requires strategic planning, adaptability, and collaboration among allies,[10] e.g., data scientists, analysts, media relations experts, and clients.

Future AI-powered media analyses would benefit from a well-defined workflow with frequent interactions between an engaged network of allies/stakeholders to enhance efficiency and effectiveness. Ever-evolving AI technology requires users to grow alongside it, continually adjusting and adapting to improve its potential. This constant transformation means all stakeholders are essential to include in every aspect of the project lifecycle and must be equipped with a basic understanding of the content, no matter their role.

Question formation is crucial in directing AI analysis and ensuring it delivers relevant insights that precipitate valuable client recommendations. Due to the enormity of data collected through AI analysis, a multi-layered approach that allows for follow-up questions is needed to ensure insights will lead to elevated strategic media recommendations. For this project, clients enforced the parameters to ensure their goals and objectives were met; data scientists provided technical support on what questions would or would not receive sufficient responses; and analysts helped refine the questions and output. But for Signal to produce an all-encompassing media analysis with meaningful recommendations, the questions needed to provide a better understanding of the article’s purpose, credibility, and value.

Given the involvement of the media relations team in interpreting the findings, the next AI-powered media analysis should increase collaboration from the beginning to aid in the development of questions. Communications professionals have a deeper understanding of the media landscape, and their insights can help craft targeted questions that delve into specific media outlets, content types, or coverage angles. Refined methodologies, such as pre-defined question templates, will facilitate communication, leading to smoother execution and a more comprehensive analysis.

Despite challenges in question formation, leveraging the team’s collective experiences and knowledge throughout the process would prove to be a determining factor for a successful outcome. EDF’s Opatik Murray asserted that the ability to work together as a team and craft analysis questions for the dataset was one of the more appealing parts about working with Signal, one that no other organization identified or presented to EDF.

“As a team overall, we were really impressed with the amount of communication and considerations that were taken throughout this analysis,” said Opatik Murray. “Each person on the Signal team was identified and played a key role in the execution — which I think also led to a successful product and trust between all parties.”

This key differentiator can be maximized for future AI-powered media analyses by involving communications experts earlier in the process, including a wider range of analysis questions that allow for follow-up, and remaining flexible in this unending web of change. Creating a standardized, replicable process for future AI-centric projects will also open the door for Signal to explore new workable tools, such as searchable databases for storing extracted articles, which can enhance efficiency and usability for clients and analysts.

For all the lessons learned, EDF stated that “other organizations facing similar challenges” will receive “high-quality end products” by working with Signal.

“The results of the analysis were exactly what EDF was looking for,” shared Opatik Murray. “EDF is looking forward to utilizing the recommendations for specific media coverage with specific individuals to better target and understand their coverage of our three topic areas.”

Human-AI Partnership

Employing both manual experience-driven and machine-driven analysis was critical in understanding and vetting the dataset to provide a usable product that successfully met the client’s needs, especially given the novelty of using AI to conduct a large-scale text analysis. While AI is powerful, human oversight remains crucial for guiding AI-powered tools and interpreting data.

The key to success lies in balancing the innovative potential of AI with the nuanced understanding and strategic insight that only human expertise can provide. During a panel at the 2023 World Economic Forum's Growth Summit, Richard Baldwin, an economist and professor at the Geneva Graduate Institute in Switzerland, assured the audience there was nothing to worry about. "AI won't take your job," he said. "It's somebody using AI that will take your job." 

And he’s right. AI is not a substitute for humans, but those who don’t learn to adapt will be left behind.

GPT’s ability to process large datasets efficiently and at a lower cost than traditional methods marks a significant advantage in handling big data for media analysis, but it cannot be stressed enough – the tool is only as capable as the person who wields it. AI, while able to add great value to work, is not infallible. No matter how much information it has access to, which is seemingly all of it, humans are famously very fallible,[11] and as long as humans remain fallible, so will AI.

A recent study conducted by Harvard Business School, The Wharton School of the University of Pennsylvania, MIT, and Boston Consulting Group (BCG) on “the effects of AI on knowledge worker productivity and quality” found that strategy consultants who used GPT-4 to help them complete a set of tasks were “significantly more productive and produced significantly higher quality results” than those who did not have access to GPT-4. However, the consultants using AI to complete tasks outside its competency, such as business problem solving, were “less likely to produce correct solutions” compared to those without.

Perhaps unsurprisingly, participants tended to take GPT-4’s incorrect output at face value, even those “who were warned about the possibility of wrong answers.” One of the more head-scratching results of the study found that participants exhibited paradoxical behavior when it came to trusting generative AI. In areas where AI could actually add substantial value, such as creative product innovation, participants were found to be less trusting, despite trusting it too much in areas where AI lacks competence. The study’s saddest revelation found that the technology's uniform output can reduce the diversity of thought within groups, potentially hindering innovation.

And therein lies the dilemma with integrating AI into the workplace – without proper coaching, we’ll start to see swaths of highly skilled workers turning to AI for assistance, operating under the incorrect assumption that these programs can do no wrong, resulting in work that leaves companies highly vulnerable and susceptible to risk.

Even I began to get sucked into AI’s web of complacency. My darkest moment came when I found myself wishing I could skip the bothersome step of having to put my wants into words by allowing the machine to implant itself in my mind.[12] Once I snapped back to reality, I remembered the considerable ethical and logistical concerns that come with implementing AI into an organization’s workflow, which cannot be left unacknowledged. From privacy concerns to data sharing to accessibility limitations, generative artificial intelligence has the ability to make or break us. I choose to believe it will make us, propelling Signal to new heights for decades to come.

Because despite its very real challenges, AI offers endless possibilities. Universal adoption is far from becoming a reality[13] as the technology continues to evolve almost daily, requiring ongoing adjustments and a little bit of risk. Yet, there is still a growing sense of optimism about AI’s ability to revolutionize how we operate as a society.

By jumping into the AI field early, Signal can invest more time in developing a standardized process for future projects, training its team on AI capabilities, and building a marketable product to enhance data exploration and analysis for clients, data experts, and comms professionals alike.

Beyond the Algorithm

Public affairs and strategic communications teams need to start accepting that AI is not going away. Instead of waiting on the porch all night in a prom dress for a date that’s never going to show, communications professionals must take control of their fate and begin broadening their skills to include data analysis and technical fluency. Doing so will pave the way for an unrivaled level of service in modern public affairs.

The U.S. Department of Commerce observed that generative AI can “empower users to discover and quickly derive insights from public data without specialized expertise or knowledge,” making the case for Signal to offer this valuable service to a wider range of organizations. Through thoughtful and careful implementation, Signal can become a leader in the field of AI-powered communication solutions, and as noted by EDF, continue to “exceed expectations” for clients.

To do so would first require investment in building internal expertise.[14] The real power of AI centers around knowing the right tool to work with, learning how to harness its capabilities, and understanding its basic operational guidelines to produce effective results. Without that foundational knowledge, lack of experience will lead to suboptimal results and wasted potential.

However, let me be clear: the answer is not to abandon employees without this foundational knowledge in favor of new AI talent. Instead, the focus should be on upskilling existing talent, leading to a ‘best of both worlds’ solution – one in which[15] communications expert Hannah Montana and developer/data scientist/analyst Miley Cyrus can leverage their complementary skills through mutual understanding to provide indispensable media analysis and strategy for clients.

An overly technical focus neglects the art of analysis and contextualization, but by merging AI’s powers with the complex, critical-thinking skills that only humans can provide, the prospects are high.

The Path Forward

If I learned one thing from this report, there is a definite market need for AI-powered media analysis services. After successfully achieving EDF’s goals through AI, Signal now holds a huge competitive advantage, and the process is only going to get better, faster, and ultimately, smarter, allowing for more time to be spent on strategy and growth.

Regardless of this potential, the impact of AI on Signal’s operations, products, services, and underlying business model depends entirely on what happens next. Adopting AI into an organization’s workflow requires a collective shift in culture and mindset, starting from the top. AI doomers and gloomers may be slow to embrace integration, and there will be triumphs and tribulations along the way, but embracing the unknown and being unafraid to experiment can offer more value to clients in the near and long term.

With some improvements and overall refinement, there is enormous potential for Signal to provide additional AI-powered services, layered with indispensable human expertise, for an array of organizations. Through effective communication, responsible integration, and human-machine collaboration, Signal can have a huge impact on the future of media analysis, cementing its role as a leader in public affairs and strategic communications for the next 20 years and beyond.

Check out the TL;DR version of this report and download the PDF on signaldc.com.

Alexa Velickovich is a Manager at Signal Group with an extensive background as a writer, content creator, and communications professional. Connect with her on LinkedIn.

[1] Did you know that in a 2016 interview with the New Yorker, Sam Altman, former CEO of OpenAI, said he has “guns, gold, potassium iodide, antibiotics, batteries, water, gas masks from the Israeli Defense Force, and a big patch of land in Big Sur (he) can fly to” in case of an ‘AI apocalypse’? Fun!

[2] The first time I used AI, I asked Gemini, Google’s AI chatbot, what it thought my boss meant when he asked me to write a “Vanity Fair-style” case study on the use of AI in conducting media analyses. You wouldn’t believe my face when the machine spit out an almost word-for-word response to the direction I was given: combine key insights from the project’s execution with an engaging narrative to create a unique and informative case study that captures the excitement and potential of utilizing AI in media analysis. It also told me to humanize the story and inject my own wit and humor, so here we are. As my father and the New Radicals would say, “you get what you give,” in this case meaning, I took the advice to heart and really ran with it.

[3] For example, in speaking with Chris Conlan, the CEO of the financial data science consultancy firm Signal collaborated with, I learned you should only ask one question of GPT at a time, a tip that immediately increased the value of its insights.

[4] Web scraping, or web data extraction is data scraping used for extracting data from websites.

[5] OpenAI is an artificial intelligence research organization dedicated to advancing the field of artificial intelligence (AI) and making AI technologies more accessible. OpenAI has developed and released several products aimed at this mission, including, GPT, OpenAI DALL-E, and more.

[6] API (Application Programming Interface) is a set of rules, protocols, and tools that allows different software applications to communicate with each other. The OpenAI API is a platform that allows developers to access and leverage artificial intelligence models developed by OpenAI.

[7] GPT (Generative Pre-trained Transformer) models are trained on vast amounts of text data and can generate human-like text, translate languages, answer questions, summarize text, and perform various other natural language processing tasks.

[8] Everyone came into The Office and stood in a conference room, decorated with half-filled gray and brown balloons, eating an undecorated ice cream cake, while listening to a broken radio. 

[9] You win or you die.

[10] Without all the politics, murder, nudity, and backstabbing (hopefully).  

[11] To err is human, or so I’ve been told. I’ve never done anything wrong in my 30 years on Earth, but I’m sure I will one day!

[12] I felt like the finance reporter for New York Magazine who got conned out of $50,000 when the realization hit me – I’m supposed to be a champion for financial safety (or in my case, authentic creativity, free-thinking, and having a job in general), responsible for providing others advice on how to be smart with their money, so why did I fall for one of the most ubiquitous scams of our generation?

[13] The first computer was invented in the 19th century, but personal computers didn’t become a mass-market consumer product until the late 1970s.

[14] The ubiquitous influence of computers by the early 1990s necessitated the development of educational resources for users to learn its functionality. Let’s not forget that, prior to the age of toddlers learning how to use an iPad before learning how to speak in complete sentences, computer classes were a regular curriculum in schools across the country. We learned how to type, navigate the internet, and conduct research responsibly. Computer classes still exist today, but instead of equipping users with a basic understanding of responsible computer usage, they’re focused on teaching kids to code and learn programming language. It’s not out of the realm of possibility to assume that AI will follow a similar path to the computer for widespread integration.

[15] Stay with me here, folks.