Retail’s using genAI 2X faster than finance! But why? Discover the coding strategy differences in the retail vs. finance sectors. #GenAICoding #RetailTech #Fintech
Explanation in video
Who’s Using AI to Write Code, and How? A Tale of Two Industries!
Hey everyone, John here! Today, we’re diving into something super interesting: how different types of businesses are starting to use a new kind of Artificial Intelligence, called Generative AI (or genAI for short), to help them create computer programs and software. It turns out that not everyone is jumping in the same way. We’re going to look at two big players: retail companies (think your favorite online stores) and finance companies (like banks and investment firms).
A recent report from an AI security company called Apiiro shed some light on this, and the differences are quite fascinating!
Retail’s Fast Lane vs. Finance’s Cautious Path
Imagine two race cars. One is zipping ahead, eager to cross the finish line, while the other is taking its time, carefully checking every part of the car before speeding up. That’s a bit like how retail and finance are approaching genAI for coding.
The report found that retail companies are moving much faster. They’re putting genAI tools into their live, customer-facing systems more than twice as often as finance companies. It seems like retail is saying, “Let’s get this cool new AI working for our customers NOW!”
Lila: “John, what do you mean by ‘customer-facing systems’ and ‘genAI tools into their live systems’?”
John: “Great questions, Lila! ‘Customer-facing systems’ are any parts of a business that customers interact with directly – like the website you shop on, or the app you use. And when we say retail is putting ‘genAI tools into their live systems,’ it means they’re using AI to power features that real customers are using in real-time. For example, an AI might be helping to suggest products to you as you browse an online store. It’s not just an experiment happening behind the scenes; it’s out there, working!”
The study looked at tons of digital folders where companies keep their computer code – these are called repositories. It found that about 61% of retail’s AI code folders show lots of activity, meaning programmers are constantly working on them, testing them, and getting them ready for customers. For finance companies, that number is much lower, around 22%. This suggests that many finance teams are still in the early stages, experimenting more slowly and carefully, often in separate, isolated teams.
Lila: “So, a ‘repository’ is like a big shared document folder, but for computer code?”
John: “Exactly, Lila! It’s a central place where developers store all the code for a project, keep track of changes, and collaborate with each other. It’s super important for building any kind of software.”
Why the Different Speeds? It’s All About the Goals (and Rules!)
So, why is retail racing ahead while finance is more measured?
- Retail’s Focus: Retail businesses are often using genAI for things that directly touch the customer and can quickly make more money. Think about:
- Recommendation engines: AI that suggests products you might like.
- Automated support: AI chatbots that answer your questions.
These features can immediately improve a customer’s experience and hopefully lead to more sales. The feedback is quick, and the impact on revenue can be seen directly.
- Finance’s Focus: Financial institutions, on the other hand, have a lot more rules and regulations to follow. They handle very sensitive information – people’s money and personal data! So, they have to be extra careful. Their genAI projects are often more focused on internal systems, things that employees use, rather than systems that customers interact with directly.
Lila: “What are ‘recommendation engines,’ John? And what does ‘regulatory scrutiny’ mean for banks?”
John: “Good one, Lila! A ‘recommendation engine’ is like a smart shopping assistant. When you’re on a website like Amazon or Netflix, and it says ‘You might also like…’ or ‘Because you watched…’, that’s usually a recommendation engine suggesting things based on what it knows about your preferences. As for ‘regulatory scrutiny,’ that means banks and financial companies are under a very watchful eye from the government and other official bodies. There are tons of strict rules (regulations) they must follow to protect people’s money and data, and to ensure fairness and stability in the financial system. So, they can’t just try out new tech without being super cautious.”
Who Started Tinkering First?
Here’s a surprising twist: even though retail is putting genAI into action faster, financial companies have actually been experimenting with it for longer! The average age of an AI code project in a finance company is about 688 days (almost two years). For retail, it’s younger, around 453 days (roughly a year and a quarter).
One expert, Jason Andersen, found this interesting. He actually thought retail would have been even slower to start. He said that the finance industry usually understands data really well and tends to be quick to explore new technologies. The fact that their projects are about two years old makes sense because that’s when many of these powerful genAI tools first started appearing.
Lila: “So, if finance started earlier, why isn’t it ‘ahead’ in using these AI tools with customers?”
John: “That’s a great point, Lila. ‘Starting earlier’ here means they began experimenting and learning about the technology sooner. But ‘being ahead’ can mean different things. In this case, retail is ‘ahead’ in terms of actually launching AI-powered features for their customers to use. Finance is taking a more cautious, longer-term approach, likely due to those strict rules and the sensitive nature of their business we talked about.”
What Kinds of AI Helpers Are They Building?
The report also looked at what these industries are trying to achieve with genAI.
Retail is often using genAI to:
- Create personalized product recommendations.
- Automate customer support (like with smart chatbots).
- Offer promotions tailored to individual shoppers.
These systems often need to use real-time, specific information about the user, which means they might need access to sensitive data.
Finance, on the other hand, is using genAI more for:
- Pilot programs: Small-scale tests to see how well something works before a bigger rollout.
- Internal assistants for their employees.
- Training AI models in ways that don’t directly involve live customer data, perhaps by using data with personal details removed.
The expert, Mr. Andersen, put it nicely: Retail IT tends to ask, “How can I use this AI to make us more efficient or increase our profits?” Finance, however, is often looking at genAI for innovation, asking, “How can this AI help us create entirely new products or services?” They have more resources (money!) and their industry is often about taking calculated risks, so they can be more experimental in their explorations.
Lila: “What’s a ‘pilot program,’ John?”
John: “Imagine you have a great new idea for a school play, Lila. Before you put on the full performance for everyone, you might do a small ‘pilot’ show for just a few friends or teachers to get their feedback and fix any problems. A ‘pilot program’ in business is very similar. It’s a small-scale test run of a new product, service, or system to see how well it works before the company invests in launching it fully.”
The AI Toolkit: Different Tools for Different Jobs
Just like a carpenter and a plumber use different tools, retail and finance companies are also picking different genAI tools.
Finance teams are using a wider variety of genAI tools. The report mentions names like OpenAI Client, LangChain, and LiteLLM. This wide range suggests they’re actively experimenting with different types of AI and for many different purposes.
Retail teams, however, have focused on a smaller, more specific set of tools, like OpenAI Python SDKs and LiteLLM. These tools are well-suited for the customer-facing things they’re building, like those recommendation systems.
Lila: “Whoa, those names sound very technical! ‘OpenAI Client,’ ‘LangChain,’ ‘LiteLLM,’ ‘OpenAI Python SDKs’… Can you simplify what those are?”
John: “You bet, Lila! Think of them as different brands and types of power tools for builders, but in this case, the ‘builders’ are software developers, and they’re ‘building’ AI applications.
- OpenAI is a famous AI research company. ‘OpenAI Client’ and ‘OpenAI Python SDKs’ are basically toolkits and connectors that programmers use to easily access and use OpenAI’s powerful AI models (like the ones that can understand and generate text or code). An ‘SDK’ stands for Software Development Kit – it’s a bundle of tools that makes a developer’s life easier.
- LangChain and LiteLLM are also helpful tools. LangChain helps developers link different AI components together to build more complex applications. LiteLLM helps developers use various different AI models more easily.
So, finance is trying out lots of different power tools, while retail has found a few favorite power tools that get their specific jobs done quickly.”
Fewer Tools, Faster Progress?
Retail’s choice to use fewer tools seems to be helping them get things up and running faster. When you have fewer tools, there are fewer connections to worry about, and it’s easier to create systems that work reliably and can be easily repeated.
Finance, with its broader toolkit, gains flexibility to try many things. However, this can also create more potential weak spots for security and make everything more complicated to manage and oversee.
One consultant, Maman Ibraham, put it bluntly: “Having 20 genAI tools doesn’t make you innovative. It makes you ungovernable.”
Lila: “What does ‘ungovernable’ mean here, John? Like a rebellious teenager?”
John: “Haha, something like that, Lila! In this tech context, ‘ungovernable’ means it becomes extremely difficult to manage, control, and keep secure. Imagine trying to keep an eye on 20 different software tools, each with its own updates, potential security holes, and ways of working. It can become chaotic and very hard to ensure everything is running smoothly and safely. It’s like trying to manage a classroom with 20 kids all doing completely different things versus just a few focused groups – much harder to keep track of!”
Staying Safe: Custom Advice for Each Industry
Because these industries are using genAI so differently, the company Apiiro (the one that did the report) recommends different safety strategies for each.
For Retail:
- Start by carefully mapping out where all your data is and how it flows (this is called data mapping).
- Conduct audits to check who has access to what data.
- Use tools that check the computer code for problems very early on, even before it’s run.
For Finance:
- Prioritize finding any hidden sensitive information, like passwords or secret keys, that might be accidentally left in the code (this is called secrets detection).
- Make sure all the bits and pieces of software they use are kept up-to-date and secure (this is like good ‘hygiene’ for their software).
- Review any old AI projects that aren’t being actively worked on anymore and decide if they should be updated, fixed up, or just shut down.
Lila: “Okay, two more, John! What exactly is ‘data mapping’ and ‘secrets detection’?”
John: “Excellent questions to wrap up with, Lila!
- ‘Data mapping’ is like creating a detailed map for your company’s information. It shows where all the important data (like customer names, purchase histories, etc.) is stored, how it moves from one system to another, and who uses it. It helps companies understand their data better and protect it.
- ‘Secrets detection’ is a security practice. ‘Secrets’ in the tech world refer to things like passwords, API keys (special codes that let programs talk to each other), or encryption keys. These are super sensitive. Secrets detection involves using special tools or processes to scan computer code and other places to find any of these ‘secrets’ that might have been accidentally left exposed where bad actors could find and misuse them. It’s like making sure you haven’t left your house keys under the doormat for everyone to see!
My Thoughts on All This
John: It’s really fascinating to see how the same powerful technology – generative AI – is being adopted so differently based on the unique pressures and goals of each industry. Retail’s rush to implement customer-facing features makes sense given their need to compete and innovate quickly in the customer experience. Finance’s caution is equally understandable due to the heavy regulations and the critical importance of security and trust in their world. It’s a good reminder that technology is just a tool, and how it’s used really depends on the user!
Lila: From my beginner’s perspective, this was super helpful! It’s clearer now that AI isn’t a one-size-fits-all thing. The idea that online shops want to use AI to show me cool stuff faster, while banks are more careful because of all the rules and our money, really makes sense. It’s less like magic and more like different companies using new tools in ways that work best for them. Still a lot to learn, but it’s exciting!
This article is based on the following original source, summarized from the author’s perspective:
Retail versus finance: How genAI coding strategies
diverge