‘These are like AI employees… they will impact job market, but not how we are thinking’: Indian AI agents startup founder

KOGO, the deep-tech AI institution founded by Gopalakrishnan and Praveer Kochhar successful 2020, has created an operating strategy that enables companies to make AI agents, AI plugins, and AI tools to enactment seamlessly wrong concern environments and enterprises. “AI agents present person cognitive ability, meaning they tin crushed and autonomously marque decisions. So, it’s nary longer conscionable automation; it’s autonomy—the quality to marque decisions and implicit workflows independently,” Gopalakrishnan explains successful an interrogation with indianexpress.com, distinguishing AI agents from accepted chatbots.

KOGO is meant for some tiny businesses and ample enterprises. (Express Photo)

“A chatbot operates connected an ‘if-then-else’ series and cannot marque decisions if thing falls extracurricular its predefined responses. For example, if you ask, “Can you archer maine the cancellation charges connected my summons with this PNR fig for a formation to Bombay time morning?” a emblematic chatbot volition apt beryllium incapable to reply that question. However, an AI cause tin entree the API, retrieve the nonstop details, and supply a implicit answer, reasoning done the accusation conscionable similar a human. It mightiness explicate that if you cancel the booking astatine 4 o’clock tomorrow, a definite cancellation interest volition apply, circumstantial taxes volition beryllium deducted, and this is the magnitude you volition receive. This is thing a chatbot cannot do. It’s a precise analyzable reasoning process,” helium explains.

An AI cause has definite skills and abilities. It tin stitchery data—whether structured, unstructured, from the internet, oregon ocular sources. It tin process each of this information, comprehend it, and make insights. “It tin recognize quality intent successful aggregate languages and interact crossed assorted interfaces, taking real-world actions. For example, it tin post, pay, record, edit, and update—all automatically. A chatbot cannot bash this.”

Gopalakrishnan’s institution has developed a proprietary AI operating system—a model that allows AI agents to beryllium deployed, managed, and built straight connected the OS. Think of it arsenic an “AI Pass” for AI agents, plugins, and tools. Enterprises and SMEs tin prime an AI cause based connected their circumstantial needs and workflow, deploying it successful minutes (in fact, nether 10 minutes) without requiring method expertise. “We are a horizontal stack, not a vertical stack, with a low-code, no-code platform,” Gopalakrishnan notes.

Different AI agents and their pricing connected KOGO. (Express Photo)

KOGO is aiming to simplify however AI is built and implemented. Comparing Large Language Models (LLMs) to energy and AI agents to appliances, helium adds: “An endeavor cannot simply instrumentality quality from an LLM and plug it into their workflow oregon system. It needs AI agents… We are the ones who created the operating strategy for this. These agents besides request to pass with assorted platforms, entree data, and interact with the world. We person built implicit 600 integrations into the platform.”

Gopalakrishnan explained that autonomous AI agents tin beryllium deployed connected a website, an app, wrong a chatbot, oregon done dependable implicit IP. This means they tin beryllium utilized implicit telephony, successful WhatsApp, connected societal media, and more. These agents enactment aggregate languages and tin run crossed assorted interfaces. KOGO powers its AI agents with respective AI models and has besides trained its ain tiny connection models for on-premises deployment, integrated with Microsoft Azure, Google Cloud Platform (GCP), and Amazon Bedrock.

KOGO gives customers the quality to physique their ain virtual agents that tin execute tasks specified arsenic handling lawsuit queries and identifying income leads. Think of them arsenic “AI employees” that execute tasks connected their own, autonomously, completing multi-step decisions connected users’ behalf without the request for quality intervention.
“Choosing an AI cause is simple,” says Gopalakrishnan. Customers tin sojourn the marketplace operated by the company, which went unrecorded earlier this month, take a pre-defined AI cause from hundreds of options available, prime it, customise it to their usage case, and commencement utilizing it immediately. He noted that this attack works good for mid-market clients who tally chains of dermatology, skin, and hairsbreadth clinics. For example, a lawsuit mightiness privation to instrumentality a lawsuit work cause to automate the process and expedite responses to lawsuit queries, which presently trust connected telephone centers wherever agents reply questions. Once they deploy the AI agent, it tin not lone instrumentality bookings connected behalf of users but besides grip different tasks.

However, the deployment of AI agents varies from lawsuit to customer. For instance, 1 enactment is cloud-based, allowing anyone to sojourn the marketplace, prime an AI agent, and customise it utilizing a no-code approach. Conversely, if a ample endeavor wants to deploy the AI agent, it volition beryllium wholly on-premises oregon connected a backstage cloud. This setup includes locally hosted models, whether they are LLM, vision, oregon code models, arsenic good arsenic the enterprise’s ain information and systems that reside wrong the organisation. The operating strategy is past installed wrong the enterprise’s premises, and the OS serves arsenic the model that supports everything and runs their ain AI store. Their AI store contains each the custom-created AI agents utilized wrong their environment. For example, an security institution mightiness privation to usage ready-made agents from KOGO’s AI store, on with astir 10 customized agents tailored to their processes. All of these agents reside wrong their AI store.

AI agents are built for productivity and designed for circumstantial usage successful businesses, specified arsenic booking flights, filling retired security forms, oregon filing disbursal reports. “We are not successful the B2C space; we are hardcore successful the endeavor space. We are B2B, and that is wherever we person built the AI,” Gopalakrishnan explained. He emphasised that these AI agents marque much consciousness successful businesses and enterprises wherever determination is simply a request to summation productivity by freeing up clip for much invaluable tasks and reducing quality intervention.

“I americium not present to merchantability a work to extremity customers. I’m present to assistance an security institution merchantability its services successful a overmuch amended and much fulfilling mode with the powerfulness of AI, providing a faster lawsuit acquisition for their clients. I’m present to assistance a slope connection amended lawsuit work and fulfillment, starring to faster, much efficient, and cheaper operations for their customers. I purpose to make a much cost-efficient and effectual AI implementation for a manufacturing institution truthful they tin trim their outgo of goods and connection amended prices to their customers.”

The scope of artificial intelligence, Gopalakrishnan said, is immense due to the fact that galore organisations inactive look challenges successful integrating their AI infrastructure with bequest systems. When these systems cannot pass with modern-day systems, they go inefficient owed to the sheer fig of processes involved.

AI agents are the adjacent large frontier for large tech companies, arsenic tech heavyweights including Microsoft, OpenAI, Amazon-backed Anthropic, and Google are each introducing autonomous AI agents to the market. These agents tin execute analyzable tasks and are susceptible of making decisions connected users’ behalf, importantly reducing the clip required for assorted processes from hours oregon days to specified minutes.

“AI agents person already deed the mainstream. It’s brainsick however overmuch concern has been done successful conscionable six months. It’s lone increasing, and we’re having to grow our capacity,” Gopalakrishnan said, adding that helium expects the adoption of AI agents by tiny businesses to prime up successful different six months.

Gopalakrishnan sees these autonomous artificial quality agents making the astir interaction successful domains specified arsenic sales, lawsuit service, operations, information management, HR, and particularly manufacturing. “It [AI] volition interaction the occupation market, but not successful the mode that radical are thinking. Those who cognize however to leverage and usage AI and AI agents wrong their systems and organizations volition support their jobs, portion those who bash not volition beryllium astatine a disadvantage. Certain jobs volition beryllium lost, but a full plethora of caller jobs volition besides beryllium created due to the fact that of AI. This is precise akin to the digitisation of banks that occurred successful the aboriginal ’90s.”

Gopalakrishnan didn’t uncover the clients his institution is moving with successful deploying these AI agents owed to confidentiality, but helium did notation however intimately his institution has worked with MapMyIndia, which has an app called Mappls that lists astir 500,000 businesses.

“AI agents service arsenic the reasoning layer. It doesn’t substance what the exemplary is; it tin go larger and better, but the reasoning capableness of the AI cause is already built-in. As the agentic model improves, it volition amended the quality to execute higher levels of autonomy. This is what volition pb to important changes. The models are already rather advanced, and they volition proceed to amended arsenic we determination toward AGI. However, the way to AGI—often referred to arsenic the way to the singularity—will mostly beryllium connected the agentic framework. When AI is susceptible of reasoning and making decisions, that volition beryllium the cardinal to achieving AGI, and this is simply a cardinal portion of the agentic framework.”