Bengaluru: There is a strong talent pipeline in India, and this energy of innovation is amazing here. In an interaction with Uma Kannan of DH, SAP Business Technology Platform President and Member of the Extended Board Michael Ameling said it’s an exciting time, more than ever, to be in the software industry. He also mentioned that there is huge opportunity for AI in India, and that companies and providers are taking this adoption seriously here. Edited excerpts:
How are you looking at the AI revolution taking place in India? Where does the country stand?
Many innovations and developments are done here in India. A large portion of the AI delivery comes out of the country. This starts with foundational work, meaning prerequisites. You need to have a database that is AI-ready. You need to have certain infrastructure that gives you access to underlying models. Then, innovations are built on top of it. Many of these agents, which we have delivered and announced at our key events, are delivered here in India. Secondly, the country contributes heavily to the distribution of AI.
Every company is now trying to adopt AI. Will AI replace coders’ jobs at the beginning level, since it’s automating some tasks traditionally done by entry-level coders?
We had this discussion when AI came a few years ago. What AI did best was creating a test case. Testing code. You ask them (freshers) first to learn code. When I went to school, in the initial years, I was not able to use a calculator because I needed to understand the mechanics. Later we used it. Similarly, you need to understand the mechanics, you need to study what engineering means. You (developers) show me how you use AI to develop, instead of showing me how you can do it without AI.
I have been asked this question as well: on the one hand, there’s this opinion, we have a shortage of developers, and on the other, AI is replacing developers. It can’t be true at the same time. My recommendation to freshers is that you need to understand the mechanics and how AI works. It helps you improve, and provides suggestions. With this knowledge, you need to have a dialogue, you need to ask, is this right? You need to think with them and don’t take everything for granted. It means developers have a bright future, because they will shape the future of our software industry with AI. Even if you think about these freshers, they would have a faster journey to get into technology, the onboarding and upskilling will be faster. Tasks will get complex, but they will focus on different things. Instead of writing code, they will write agents who write code. The purpose is different, but you still need engineers and developers to do this. This is where it’s moving. We strongly believe in this. Otherwise, we wouldn’t invest much in developers. It’s an exciting time more than ever to be in the software industry, as how we work will be different.
When you talk about AI innovation, where is it exactly heading? Any particular area you are seeing growth, like health, retail, among your customers?
Agents are the key because you will see functionality moving into agentic behaviour. That’s the reason why we are focused on agents. These are real agents, autonomous working artifacts. What is even more exciting is the agent-to-agent communication. These agents will talk to other agents. The next agent will talk to a third-party agent. We can bring finance, procurement, supply-chain, and manufacturing. We can bring this knowledge together with agents and solve complex business problems. That’s where it’s moving. Agents are bringing more value in combining them together. We at SAP, since we cover the entire value-chain, have a huge opportunity doing this.
Currently, two agents can communicate. In future, are there possibilities that many more agents can communicate across domains/verticals? Is there any limitation right now in terms of agents?
Currently, yes, there are limitations. Firstly, if you look at MCP (Model Context Protocol) servers, you hear this: MCP servers are the answer for everything, which is not the truth. There are limits. Large language models struggle as soon as it comes to more than 20 MCP tools. In agent-to-agent communications, there are limits today because this entire orchestration is not easy. For example, when you have different models, you want to change them on the fly. We have this prompt optimiser where you can have an abstract layer to resist. This is solved, but this is evolving at tremendous speed. What we are envisioning is that Joule (AI co-pilot) will become our agent orchestrator.
The first assistant we delivered is a combination of agents. In our support, you can classify support tickets by complexity. The first natural thing is that you have an agent who answers simple queries. Then you can imagine there is a more complex question. This is a more expensive and time-consuming task. You can deliver multiple agents who solve complex problems. You’d be trained in this model.
Our support is working with a set of 20 agents. Depending on the need, what is incoming, helping us serve and support these tickets. This has a tremendous advantage, because you’re more responsive to the customer. The quality of tickets is increasing, and it saves cost. These are the use cases we need to look at. Currently, we are focusing on delivering in these two domains, these agents. We are bringing them together, because if you’re about the contingent workforce and you want to do planning, you need your finance data and you have finance agents working in the background. This is happening; we have agents across different domains. They get data from different sources, from finance to procurement to manufacturing. We combine this. Do we have complete agents that cover the entire value-chain? This is not the case yet, but it’s moving in that direction.