Welcome
Scott McDonald — CEO & President, the ARF
Scott McDonald CEO & President of the ARF opened the virtual town hall by covering the current state of AI capabilities and how it’s been experimented with in the industry. AI is employed in the production of advertising, the analysis of advertising placements, in improving SEO, optimizing CRM copy to avoid spam filters and putting copy in compliance with general advertising effectiveness frameworks, such as AIDA or PAS. There have been other uses too, like speeding up certain repetitive creative processes. However, as much as AI offers new capabilities it also creates novel concerns.
Key takeaways:
- At the Creative Effectiveness Conference in October 2022, AI programs like DALL-E were shown to increase the rapidity of prototyping and stimulating creative processes that once took a considerable amount of time to complete.
- AI might be on the “inflated expectations” section of the famous Gartner Hype Cycle. This speculation was made last October, interestingly, at the same time ChatGPT came out.
- As reported in News You Can Use, the ARF asked ChatGPT, “How do you make effective advertising?” While nothing revolutionary popped up, it still gave a plausible and concise summary of best practices.
- A recent JAR article covered they many ways deepfakes could affect the advertising industry.
Anastasia Leng — Founder & CEO, CreativeX
Anastasia Leng from CreativeX challenged us to change our mindset on AI. Rather than thinking about what AI can do, consider what specific business problems AI might be well suited to solve. While creative is responsible for 49% of business growth, creative decisions are not often data driven. AI can extract data from each piece of content, which lends insights that can help achieve creative stability, allow for a granular understanding of the content produced in terms of things like representation, accessibility and sustainability, and help achieve organizational intelligence where AI is used to answer specific questions about an organization’s content.
Key takeaways:
- Creative data is our ability to extract information from a creative asset.
- On the highest level, creative analysis achieves organizational intelligence, extracting creative data to answer how specific creative decisions impact marketing effectiveness.
- Micro-creative data is defined as things like objects or colors, recognizing the basic things contained in an image. AI can take this information and tell whether for instance, an ad is in line with brand principles, if it has a purpose-based message or if it’s representative of a brand’s audience.
- Artificial intelligence can also be used to measure brand consistency. AI can look at distinctive brand assets, see how often they are applied and even measure how optimizing brand consistency affects brand lift and brand perception.
- AI can answer such questions as, “What percentage of our content features people of color?” Or “How often are women shown in progressive roles?” In the absence of AI, we do not possess the tools to answer such questions at scale.
Kyle Monson — Founding Partner, Codeword
Kyle Monson of Codeword said that while the media has hyped up automation and AI early on in 2023, few outlets delved into exactly how these tools can be used in an agency setting. To address practical concerns, such as how to integrate AI tools into actual creative teams, how an art copy team should use AI, what kinds of work AI should do and how clients think about AI being employed on their projects, Codeword began the year by announcing a little experiment. They had “hired” the world’s first AI interns for their creative team. This experiment yielded several insights and offers glimpses into how AI will shape the creative future of the industry.
Key takeaways:
- The first AI intern was Aiko (they/them) who was a mix of Midjourney, DALL-E2 and Stable Diffusion. Aiko supported the design team in typical intern tasks like mood boards, all-hands decks, brand identity audits, rough mocks and social visuals.
- The second intern was Aiden (they/them) who is ChatGPT (soon to be Bard). Aiden supported the editorial team in intern-type tasks, such as tone and voice analysis, news crawls and first drafts of manifestos.
- These AI need lots of oversight and prompts must be specific to receive quality work.
- These tools at present may be best suited for tasks creatives don’t enjoy doing or find draining.
- Most brands are likely going to give personas to these tools to try and humanize them. This raises questions about representation, culture and bias. Their solution was to let the AI make up as many details of its own persona as possible.
- How to avoid AI replacing entry-level talent is an issue they are still working on.
- Ensure that clients are onboard, and the business model is safe by framing it as staying cutting-edge. In the coming years, clients will likely expect agencies to have experience with these creative tools. Those cost savings and efficiencies will then be passed back to them.
- Things like billable hours, staffing models, what scopes and contracts look like and others still must be worked out.
Patrick Moriarty — SVP & Analytics Practice Leader, North America, Kantar
Patrick Moriarty from Kantar talked about how large organizations might approach AI, the use of which will soon become a requirement to stay competitive in the market. Where a large organization has the advantage is leveraging deep and profound data assets that have been accumulated over time. The variations that exist throughout them are great for generating insights. Not only can AI help face creative challenges but media ones as well. Such algorithms can be used in the planning stages with creative briefing and with measurement once in market. To be successful though, there must be true collaboration among developers, engineers, data scientists and creative analysts. Beyond collaboration, oversight, guidance and standards must be employed for AI to work properly.
Key takeaways:
- Dx AI solutions gather insights from “the wild” about what consumers are saying about different products and categories, which are used in creative development.
- AI is also used to expand the set of testable alternatives, mitigate bias and support integrated creative and planning strategies.
- AI’s always-on capability can generate insights that help make marketing copy more effective, increase the effectiveness of campaign execution and optimize the mix.
- Copy performance predictions can be made quickly using the LinkAI algorithm and Kantar’s vast data assets on copy testing.
- AI solutions are also used to analyze and integrate data from Numerator, Kantar Profiles, media plans, clients’ CRM systems, to provide a causal view of campaign performance.
- AI can analyze data across the entire ecosystem, from search to social, e-retail and reviews to digital performance analytics, and make sense of it, a feat that would take more people than budgets allow.
- The data science and data development spaces are highly competitive and will likely remain so. AI can help unlock value in this space at scale while competing in a resource constrained environment.
Panel Discussion
Anastasia Leng — Founder & CEO, CreativeX
Kyle Monson — Founding Partner, Codeword
Patrick Moriarty — SVP & Analytics Practice Leader, North America, Kantar
Moderator: Scott McDonald — CEO & President, the ARF
Scott opened the discussion saying the uses of AI seem to be around speeding up processes or relieving drudgery. He asked one year out, what would be the biggest contribution of AI to the advertising process? Kyle Monson of Codeword said in the creative field, AI is used to speed up campaign mood boards and other brand identity explorations. Since these tasks are pure research, generative tools save weeks of time. Anastasia Leng of CreativeX said that AI will become more involved in repetitive processes like tagging and reviewing ad labels, where human concentration lags over time. Patrick Moriarty from Kantar said standardization is key.
How can we account for inaccuracies or misinformation in training sets? Human oversight is essential. Several quality assurance processes can be put into place. For instance, there are tools like Two Plus One Reviews, where any data label goes through a system of checks and both reviewers must agree before it goes into the machine for training. Another consideration, to get a better handle on this nascent technology, organizations should have their data scientists use whatever time AI has saved them to further discover their potential.
Key takeaways:
- Use AI for what it is better than humans at doing and humans for what they are better than AI at doing.
- Soon, robotic elements of programming will help alleviate the shortage of code developers and programmers.
- In terms of inputs for generative tools, users are currently downstream on how the data is analyzed and collected, unlike conversation monitoring tools or sentiment analysis tools.
- Superbots causing havoc and AI respondents to surveys are just some of the challenges the industry will face.
- Even though it is important to good advertising, to date we cannot get AI to recognize the presence of humor.
- Good examples of the ethical use of AI include disclosure and good quality control practices.