Since the generative AI boom began, advertising holding companies and independent agencies have been racing to develop their own platforms. Now, other agencies are entering the fray with their own AI platforms focused on media buying, sales, strategy, and other operations.
Gale, owned by Stagwell Inc., has a new platform called Alchemy.AI. This evolved from his previous Alchemy platform for media agencies, which also had its own internally built machine learning models. While many agencies are debuting new generative AI products built with large-scale language models (LLMs), Gale founder and CEO Brad Sims says agencies must take all the appropriate safeguards. He said he would like to confirm that he is taking the necessary steps. For example, he said Alchemy.AI was developed to be ISO compliant and also adheres to strict data standards in Europe and California.
“This is one of the reasons why some might say we've been late to the AI conversation,” said Sims, who founded Gale in 2014. Our customers will expect. If we were sitting here a year ago, I'm not sure either of us would have had any confidence in these LLMs of Enterprise Security. ”
As part of Stagwell's agency network, Gail will also be able to leverage the holding company's various AI investments. For example, the company could benefit from Stagwell's 2023 deal with Google Cloud, which focuses on marketing-related solutions for generative AI. Other agency holding companies, including Publicis Groupe, WPP and IPG, have also debuted new AI platforms and partnerships with various companies in recent months.
Trained on both first- and third-party data, Alchemy.AI also includes capabilities for both sales strategy and media planning. You can create custom audiences, analyze the results of past advertising campaigns, develop strategies for future campaigns, and provide insights into customer loyalty. For example, you can compare sales data for loyal and non-loyal customers, suggest areas for reinvestment, and analyze customer acquisition and retention rates. The platform also has ways to leverage his CRM data from external partners, brand-owned platforms, and transactional data to help develop campaigns across various platforms. Other features include tools for data visualization, business development, project management, human resources operations, and other operations.
To develop Alchemy.AI, the agency has entered into corporate agreements with several leading AI companies, including Google, OpenAI, and Anthropic, to provide API access to AI models that power chatbots such as Gemini, ChatGPT, and Claude. did. Gale also abstracted Alchemy.AI's backend and frontend, making it easier to switch between AI models and cloud providers. This approach not only allows multiple users within any given AI chat, but also makes it easy to switch between AI models. This is especially useful when LLMs perform differently or when companies have data in different locations.
Alchemy.AI also has scenario planning capabilities aimed at both media planners and non-media buyers. While other companies have their own scenario planning tools, Sims says Gale's platform allows for a more agnostic approach to eliminate bias when recommending media spend. He argued that it would be.
Asked about the amount of historical data needed to train Alchemy.AI, Sims said clients use at least two years of performance data when optimizing owned media, and ideally for paid media. We mentioned that it would be ideal to use three years of performance data. According to Simms, the process of onboarding clients to her Alchemy.AI also takes a lot of time and effort. He said it took him four weeks for his first few clients, but now the process is taking him three to five days.
Having a more objective perspective allows advertisers to better assess performance across different advertising platforms. This may be attractive as more advertisers are concerned about the opaque nature of the AI offered by platforms such as Google's P-MAX and Meta's Advantage Plus.
“I'm not saying any media company does it intentionally, but it's a question we get asked all the time by our clients,” Sims said. “How independent and agnostic is this recommendation?…If you’re in a client environment and the media plan is there and created, you can actually go in there and [ask] What if the client had 2 million more? ”
In addition to agencies building new platforms, outside experts say AI will reshape ad budgets. And while some expect the current AI hype to subside, analysts expect AI to continue to reshape advertising budgets. Macquarie senior media technology analyst Tim Nolen told Yahoo Finance last week that ad tech companies such as AppLovin and The Trade Desk could also benefit from the change.
Integrating multiple AI models into a single platform is something more companies are pursuing. In addition to tooling enhancements, companies also have a stop-gap solution in case problems occur with their large-scale language models. New AI and privacy laws, including the newly approved AI law in Europe, are also leading companies to ensure they strengthen both their safeguards and capabilities.
Nitish Mittal, partner at IT research and consulting firm Everest Group, said procuring and building an AI platform also requires a different approach than other technology stacks. He added that the very nature of LLM and data will lead to a “new era of multi-LLM.”
“AI is not just a small thing,” Mittal told Digiday. “Everything goes from hardware to software, data to services, and we have to stitch them together. We need almost new specialized vehicles to do this.”
Improving LLM data privacy practices also requires new approaches to how companies collect, store, and use data. This has led companies to integrate privacy by design when building new tools. In an online panel discussion last week after the AI Act was approved by the European Parliament, IBM's Chief Privacy Trust Officer Christina Montgomery said companies will document what data they use to train their AI models and how they obtain it. He said that is important.
“So how do you create user documentation that helps explain and provide clients and others how to use those algorithms,” she said. “So how do you monitor AI throughout its lifecycle as it changes over time?”
Prompts and Products — AI News and Announcements
- The European Parliament passed the AI Act, a comprehensive law with guardrails for how companies develop and deploy AI systems.
- An agency's new AI-powered commercial for Under Armor has sparked new controversy over generative AI tools trained on previously created brand assets.
- Prescient AI, which uses machine learning models to improve media measurement for e-commerce brands, announced it has raised $10 million in additional funding.
- A Wall Street Journal interview with OpenAI's CTO left many wondering what the startup's Sora model was trained on.
- OpenAI announced new deals with European publishers including Le Monde and Prisa.
- Two US senators have introduced a new bipartisan bill that would create new guidelines on AI ahead of the US election.
- The Aspen Institute announced it will host an event focused on AI and elections on March 28th. Speakers include former US Secretary of State Hillary Rodham Clinton, Google co-founder Eric Schmidt, and various other government officials from the US and European Union.
Other articles across DIGIDAY