How to mitigate risks in your AI-driven marketing campaign

AI implementation is a competitive necessity in today's marketing landscape. However, deploying AI without a risk mitigation strategy can lead to brand damage and regulatory issues. In this guide, we share our experience and outline some critical considerations for marketers seeking to harness AI's power responsibly.  

Production-ready AI use cases in 2024  

AI's impact on marketing ROI is undeniable. L'Oréal's AI-powered app boosted conversion rates by 63% through personalised product recommendations, as reported by the Financial Times earlier this year. Klarna's AI customer service chatbot slashed case resolution time from 11 minutes to under 2, while driving a US$40 million profit improvement in its first year.  

Some brands that used to divide their customer base into just a few segments are now delivering much more targeted messaging with AI-generated content. Others are using AIgenerated videos and images as a novelty.  

In short, there are many proven and effective AI use cases today to suit different brand styles and risk appetite.  

AI/Human gatekeeping  

We believe it’s good practice to implement a two-tiered gatekeeping approach to maintain quality and consistency.  

First, deploy AI agents for initial content screening. Specialised prompting can help make sure that AI posts align with your brand’s standards, and you can even run multiple passes of automated checking to improve reliability.  

Of course, final checks by humans are non-negotiable. But this process can now be reduced to one or two final reviews, saving brands and agencies valuable time to focus on higher-impact efforts.  

Combating Hallucination  

While AI hallucinations (generating incorrect information) remain an unsolved challenge, brands can still take steps to enhance generation reliability. Optimising temperature settings on your Large Language Model (LLM) can help brands find a balance between out-of-control randomness and useful creativity.  

Retrieval-Augmented Generation (RAG) should also be used to ground AI outputs in verified information sources (e.g., style guides, product databases, brochures, etc.), significantly improving content accuracy and reliability.  

Copyright Considerations 

 AI-generated content rights are an evolving field, with ongoing court cases and different legal status in Chinese and Western markets. At this stage, brands can protect themselves by vetting each AI model's and provider’s terms of use.  

For example, Meta allows royalty-free commercial use of their Llama models if a brand has fewer than 700 million monthly active users, and many services allow commercial use and brand-friendly copyright policies with the right subscription tier.  

At the other extreme, LG reserves all property rights over content generated with its EXAONE-3.0 LLM. Therefore, always consult experts and scrutinise terms of use to ensure compliance and protect your brand's interests when selecting AI tools.  

In short, the question for marketers today isn't whether to implement AI, but how to do so effectively and responsibly. There is no “best” approach because each use case must balance innovation with risk management based on each brand’s needs. Nevertheless, the future of marketing is clearly AI-driven - ensure your brand is prepared to lead rather than follow. 

This article is sponsored by Pontac.