Future of Developers in the Era of AI - Charles Adaikkalam

This talk explores the shifts in software development brought about by advancements in AI. The speaker highlights how the traditional bottlenecks in creating software, such as extensive planning and coding requirements, have been alleviated. Now, even individuals outside traditional technical roles can develop applications, leading to a surge in the creation of software solutions. However, this rapid industrialization of software comes with challenges, particularly regarding code quality and accountability. The speaker emphasizes that developers will need to evolve from merely writing code to becoming system designers responsible for ensuring the integrity and safety of AI-generated outputs. The importance of specifications, orchestration, verification, and security governance is also discussed, along with the role of low-code platforms in this new landscape.

Two years ago, okay, at least just in the recent past, building software was obviously hard. Uh you need to have months of planning, weeks of coding, huge team of engineers and so on. And if you look at the landscape, right, these are the some of the technologies that you need to develop a software or to develop any applications. Uh you need to have this across the board,

across the APIs, across the database, for deployment infrastructure, and all the one to coordinate to make one system work. Um but today, as we all know, you just open uh uh just the chat box, and then you just enter something, and you have a working system available already. So, the question is no longer obviously, can we build software, but the question is what's going to happen to

the industry when stops being the real bottleneck. So, every industry, if you see, has some bottleneck somewhere, okay? And when that bottleneck moves, everything pretty reorganizes around that new bottleneck. For decades or from the time software was there, the bottleneck in software was creation, was simply writing the code, getting it right, making sure nothing breaks, and making sure that we ship without any bugs and so on.

Now, what AI has done, obviously, has It has moved that bottleneck. So, it definitely has implications for everyone, and more specifically, for everyone who's sitting in this room. And it has happened so fast that anyone expected, and what seemed very remarkable 2 years ago is just a new normal today. in putting another words, software is being industrialized. Okay? So, you can see If you want to know

what's going to happen, you can really look at what happened in other industries like power looms, uh automobile, and other industries as well. Everything started with a craft, then the factory, then the assembly line, and then the automation. So, software has already entered into that industrial phase, and once it gets into the industrial phase, it may it ensures that things that are produced becomes very cheap, very

easy to produced, and we start producing more of it. So, here's a recent IDC update that says there'll be more than 1 billion logical applications that will be built by 2028. I think with the pace that goes, I think it could be even 10 billion applications. Who knows? And this even this 1 billion applications are much more than the number of applications that have been built in

the past 40 years. Much more than the number of applications built in the past 40 years. So, this is going to create a problem that nobody probably is seeing. So, when software was hard to build, scarcity was the filter. So, you only have to build only which is absolutely what it was needed. Every feature had to be justified. Every system um has to be deeply understood. Now

that friction is removed, anyone can build softwares. Even marketing managers can build. People who are closest to the problem can build. The line of business can build and so on. Even recently, my marketing manager built a cool white-coded app that uh he brought in all the case studies of our product, and then he showed to us and it was really remarkable. One could search quickly and then

query it and then it will bring up the case studies. I'm able to see a lot of line of business people building applications, cool applications. So, more systems, uh functions and more features, and more automated system is going to be there in production. But, there's going to be a great risk that you all need to understand. As you all know, all code is spaghetti. We don't like

to see our own code, right? In uh few months. Definitely, it was other people's code, and uh because after sometimes that code becomes really Now, with capability of building something just by prompt, so we end up building tons and tons of code that nobody really fully understands, okay? Systems are too large. Systems are simply too generative, and they are all interconnected, and they are moving very fast,

okay? So, that's going to be the defining challenge of the next decade. That would not have encountered such scenario at all. think about the challenge, what it really means in practice, actually. Systems are really too large to be reviewed. They are generated much very fast than for you to audit actually. Okay? Previously, probably a senior developer would have max maybe one or two files a day to

review the code of a junior developer. Okay? And today, you would have tons and tons of AI generated code. Okay? So, what really protects us a developer, the future of developer is not the ability to to write code, but then the ability to understand what's being built Developers must be now accountable for outcomes are going to be from coders to being a system designer. So, this is

going to be the new developer stack, if you see uh if you see one thing that is missing is writing code actually from the stack. obviously, the first layer in this is going to be the specification. Uh it's not going to be just the prompting. You just need to uh learn the latest uh things like writing your skills files, writing your empty files, and what considering that

what boundary that you want to have. Writing the spec is going to be more important than writing your code, obviously. Now, because the spec is the new contract, any vague instructions or any vague spec that you're going to give is going to really produce a very vague system. And next is obviously the orchestration. With a lot of agentic thing coming up, you're going to run all now

a lot of agents in parallel. That's going to be a a planning agent, that's going to be a design agent, that's going to be a lot of agents that's going to run in parallel. So, there has to be some governing mechanism to manage all those agents that's running in parallel. So, that someone is going to be you. And there are going to be another two layers which

are going to be even harder than uh ever going to be. That's what I just told you, right? The first is going to be the verification sister system. So, what code has it produced really? Is it really compliant with the regulations? Is it uh can it handle the load up up yes up I mean an unexpected load that could happen during the midnight? Because AI today it

generates very confidently even when it's wrong actually, okay? So, the developer who can really look at the output, who can understand really systematically and know whether it is safe to ship or not is going to be the most valuable person among the the engineering team. And finally the security and the governance. There are already very significant security problems, security incidents that's happening, that's being reported, and that's

because mostly because of the AI committed um AI assisted committed code. AI assisted code has to be really audited. So, we have one rule at Zoho that any code that's going to be the the the responsibility lies with the developer who does it. So, no code can just go into production just like that. Every single line of code has to be reviewed very clearly. And a lot

of um uh regulation acts also that are coming up like the EU AI Act that's going to be fully enforced in 2026 as that question, who is going to be responsible for the code that's generated is going to lie to the So, the more AI code means the more ownership that you have to own actually, maybe in hindsight we may think, "Oh, AI code generation it's going

to reduce my ownership." No, it's not actually. It's going to only expand your ownership because you have to verify secure governance and maintain the code there are fundamentally two approaches or or different ways where a assisted development works. The first one you know, right? Obviously the cursor kind of tools, the white coding cursor or co-pilot on the cloud code. There are tons of pro code assisted um

tools that is available in the market. And the code that is generated can be probabilistic in nature. That's it's very important that you verify the output because the output varies actually. And there is another thing that I want to introduce that is the low-code platform which is again AI assisted. A low-code platform like Zoho Creator and Microsoft Power Apps. So these low-code platforms by design require very

very less number of code. So that's that's I mean low-code something that is not just started yesterday or just sometime back. It's been there for a very long time. And low-code works by abstracting that spaghetti code into artifacts like you have heavily abstracted things and then as a developer you need to write only very less number of code. So by design since you need to write very

less number of code, you need to just maintain very less number of code. So so that is easier and safer. So and because a lot of things are abstracted there are a lot of predictable outcomes that's there in low-code platforms. So when do you need to choose between these two? Obviously these both would coexist actually. And if your team has the capacity to govern all the code

that is generated, if you have a large team, and if you need maximum flexibility, obviously, then pro coding is the right approach. But if you need speed and at the same time you want some predictability, then low code definitely could be the way actually. And um if you know if you want to know more about low code, there is a session that's happening at 12:10 in Hall

E. Uh my colleague Vignesh is going to take more about low code you can know more about uh the low code platform over there. to summarize every line that AI generates, right? You are going to commit and that responsibility lies with you actually. The ownership, the liability, the security, the long-term behavior, everything lies with you. Just the fact that AI has generated doesn't mean that you cannot

absolve of the ownership of the code that you generate. Obviously, you are the owner of that code. And as a developer um like I told, you no longer have to generate the code, but you are going to be the system designer who is going to verify the outcome. as a ending note, I would say each developer is going to be the guardian of the digital trust. So,

AI really is going to generate a lot lot lot more than that you can imagine and that responsibility is going to lie with you and not as a go coder, but as a guardian of digital trust. Thank you. >> [music]