Generative AI has quickly become a transformative technology in various industries, such as logistics, customer service, and data management – all crucial facets of the retail sector. Many major retailers are turning their attention to this growing technology to revolutionise their business. However, retailers seeking to implement generative AI should understand its potential successes and its risks and regulation.

In this article, we explore what exactly generative AI is, the capabilities of generative AI, the regulatory landscape surrounding it, and the next steps retailers can take to implement AI, supported by resources from our Partners RSM UK and RPC.

 

What is generative AI?

Generative AI is artificial intelligence with the ability to create unique content in response to a human prompt.

One of the defining features that sets generative AI apart is its ability to read these prompts and respond in an accessible way that is easy for humans to understand. For example, ChatGPT is a large language model (LLM) that can interpret normal human speech patterns and respond conversationally (including adapting the language it uses according to further prompts), and programs like OpenAI’s DALL-E and new iterations of Photoshop allow users to edit images by inputting a written prompt.

This ability to “generate” gives AI incredible use cases across industries – for example, a language model with the ability to converse naturally can drive technology like chatbots, while prompt-based image generation has the potential to speed up processes for the huge volume of online content necessary for e-commerce and social media marketing, both for experienced photo editors and beginners.

However, with this technology comes risks: if AI has the power to do all of this, then is there a point in having dedicated members of your team doing these roles as well? What is the point of your customer service staff, your photo editors, and your writers?

These roles continue to be vital for your business, for the key reason that AI can only respond to a human-given prompt. AI alone cannot think outside of the box, fact-check itself or assess the quality of its work - this still falls to your talented humans. Instead, generative AI functions as a tool for your team to improve the productivity of their work.

 

Use cases for generative AI in retail

In his webinar on generative AI, Ben Bilsland from RSM UK highlighted key areas where generative AI can be implemented in retail. These include:

• Text generation – from short sentences to entire articles
• Image generation – high-quality visual content
• Chatbots – making it simple to engage with customers online
• Code generation – create and explain functional code
• Summarisation – instantly sum up a document or concept
• Plugins – integrations with other software to make it more powerful

To hear Ben expand more on the capabilities of generative AI, watch the full webinar on demand.

Generative AI for sustainability

As well as its use in text generation, code generation, and many more business functions, recent developments in generative AI have shown its potential benefit for helping retails achieve Net Zero.

With less than 30% of retailers feeling confident about their sustainability journey, the role AI has to play could be huge - analysing data, navigating complex reporting requirements, and more.

To learn more about the potential of generative AI to support retail sustainability, register for our webinar on the 31st October 2023 at 11 AM - registration is free for BRC members and non-members.

Register now.

Which roles in retail will benefit most from AI?

With more and more retailers investing in artificial intelligence, specific areas have been highlighted as benefiting most from the implementation of AI, such as:

• HR
• Customer service
• Marketing
• Content writing
• Supply chain management
• Finance

We’re still in the early stages of understanding this technology, and there are inherent risks around being the first to market when harnessing new innovations. For example, unleashing an LLM in your online chatbot comes with its own set of risks - do you need to tell consumers they are interacting with AI? Should consumers know how you are processing the data they are inputting into the LLM to train it to become more effective?

Upskilling teams so they are aware of the risks surrounding generative AI will be key to the success of implementation.

What are some of the risks of using generative AI for retail?

In our recent event, Setting Digital Transformation Goals, Tania Williams and Helen Armstrong of RPC expanded on the potential risks when implementing generative AI in retail. Below, they expand on the key takeaways from the session:

While generative AI can boost your team’s efficiency, it comes with a unique list of potential pitfalls. To implement AI properly, retailers should be aware of the risks and limitations of AI.

IP infringement

Generative AI large language models use pre-written content on the Internet to formulate their responses (although ChatGPT currently uses the Internet up to September 2021, which comes with its own host of problems).

This means that while the model can generate its own responses to prompts, it is debatable who the original information belongs to. It is also currently unclear whether training AI on the basis of this 'data scraped' publicly available information falls within any statutory copyright exceptions. This leaves a user potentially open to the risk of a claim in copyright infringement (or even plagiarism if the chat model copies directly from another source that it deems the most optimal response).

For this reason, language models such as ChatGPT are better used to provide scaffolds for content rather than writing the content itself. Not only does this speed up the process of creating search-engine-optimised content, but it also creates higher-quality content too, written by and for humans.

Bias in AI

While the experts behind generative AI language models do their best to prevent this, bias from an underlying data set can become entrenched in an AI algorithm. This creates the danger that, in reality, an apparently neutral decision-maker can have bias built in. For example, when used for shortlisting job candidates, gender stereotypes relating to the role may be applied. 

Not only are humans crucial in ensuring that the data used to train AI is itself free of bias, but also in programming generative AI to avoid these responses and properly auditing the responses to ensure that bias output is removed.

AI hallucinations

In addition, generative AI can produce completely inaccurate information/responses, presented with complete confidence. This is shown in RSM UK’s webinar on Generative AI, where the AI produces an incorrect response to the prompt for a “haiku” – watch the full webinar recording here.

While this is an innocent case, AI hallucinations have the potential to be incredibly damaging if not noticed: for example, providing non-existent sources. To use generative AI as a tool to receive information, retailers should double-check the sources it uses and quality-check its responses, something the AI cannot do alone.

AI regulation

Over the coming years, important AI regulations will be coming into place. As the UK is positioning itself on the forefront of AI, describing itself as taking a “pro-innovation approach”, retailers have the space to experiment with and benefit from AI, but the government has recently suggested that it will implement more guardrails around AI. While retailers have the space to experiment with and benefit from AI, it’s likely that more stringent regulation will be put in place.

Many are anticipating that AI and the UK’s role in its innovation will be a key part of the October Summit 2023.

RPC have created a useful one-pager resource on AI regulation – download it for free here.

 

Possible concerns around generative AI

In his webinar, Ben Bilsland shared that many people are concerned about being “replaced” by generative AI, especially in light of OpenAI’s prediction that 18% of the workforce will see 50% of their work affected.

However, this does not have to be negative. By “affected”, this could mean increased work output, higher productivity, and even higher quality of life. Further progress in generative AI can be of significant benefit in fields such as medicine or law, and could give your team time back to do what they love. Ben encouraged viewers of his webinar to think of generative AI as a tool with which workers can 10x their work, not be replaced.

This does not mean that AI shouldn’t be used with caution. This just means that generative AI has the power, in the coming months and years, to revolutionise the way we work for the better if used correctly. That means having talented humans at the helm.

 

Conclusion

While it is crucial to be aware of the unique risks of using generative AI, the possibilities and benefits are limitless. Whether it’s helping you edit product photos, add copy to your website pages, respond to customers or organise your company’s data, generative AI can unlock new horizons for retailers that implement it correctly.

With many major retailers like Adidas, Levi’s, Coca-Cola and more investing into generative AI, the future of retail seems to sit in the promise of this digital landscape. The most important takeaway is that, while this technology has immense promise, it cannot be used in isolation: your team and your talent is the most important part in using generative AI to its fullest.

And on the legal and commercial side, RPC has shared some key areas of thought for retailers considering generative AI for their business:

For more information on generative AI and its regulations, download RPC’s one-pager or watch the full webinar on Generative AI for Retail – and if you’re interested in supporting the BRC’s work around digital transformation in the form of a partnership, contact us.