AI is not Valyrian. It’s the new English.

If you are counting HR leaders who are actively planning or already deploying GenAI, a Gartner report tells you that number went up from 19 per cent in June 2023 to 61 per cent by January 2025. Not far away- The Hackett Group’s assessment pointed out that in 2024, 66 per cent of HR teams tested Gen AI but only 16 per cent prioritised it. 89 per cent seem to be scaling fast now to close a 12 per cent efficiency gap and meet rising workforce demands. Korn Ferry’s 2025 CHRO Survey revealed that 42 per cent of CHROs are prioritising investments in AI for HR, and yet only five per cent of HR teams feel fully prepared to implement it effectively.

Something that echoes in the Littler’s European Employer Survey report 2025, when it showed that 20 per cent of respondents (C-Suite folks, HR leaders, etc.) are not at all prepared for compliance with the EU AI Act. This report also warned that employers deploying AI technology will face a slew of compliance obligations for ‘high-risk’ systems when the Act goes into effect next year.

On one hand, AI brings augmentation, time savings, cost savings and intelligent automation for HR folks. On the other hand, citizen AI tools are adding new forms of fraud, and AI-driven errors and bias are muddying the HR waters. How to wield AI then? Not with fear for sure, stresses Hari Bayireddi, President, COO & Co-founder, Phenom (a player in the realm of applied AI that works towards helping organisations hire faster, develop better and retain longer). But also not with audacious jumps. He delves more into how to hire AI right – as he presents a blueprint of HR’s strides in the AI world in this interview.

Our detection engines can help spot frauds and double-payroll instances. HR is based on EQ. We provide IQ to this EQ.

What’s Phenom’s space- are you another HCM player in a market dominated by big HCM players like Workday, SAP, Zoho, etc.? What can AI change in this space?

No. Ours is a new category. We are on top of HCM systems. We are trying to bring integration and context into the field of AI in the talent space. With automation engine, decision engine, orchestration engine and automation engine- we are spanning the AI adoption part – and deeply. Phenom’s applied AI is different from generic AI because it is about leveraging HR-specific data models that understand how talent decisions connect to organisational outcomes.

AI is bigger than the Internet. But it will be as pivotal and game-changing as the Internet was. It is moving too fast. There will be ups and downs. Speed creates security and quality issues. That’s normal with any technology-related change. So we need a proper approach and guardrails for AI.

What does ‘applied’ exactly mean?

It is a combination of intelligence and automation, and very specific to the enterprise in question. General AI is very universal, but enterprises need something that works on proprietary data. That’s very hard for foundational models to manage. That also needs a lot of guardrails. Every specific use case and workflow calls for different datasets. Our ontology helps in bringing that search specificity. We bring that rigour and context. For instance, our detection engines can help spot fraud and double-payroll instances. HR is based on EQ. We provide IQ to this EQ.

Is it easy for verticalisation to work in a space like HR Tech that has been horizontal for a long time?

It is happening fast as we see it now. Horizontal work takes time. We are solving not just the effort part but also the value and impact aspects here. Unlike SaaS, Agentic AI has to be very clear and tangible on outcomes. That’s really new with AI. Customers have to see value here and be able to maximise revenues with technology. Traditional talent approaches have created persistent gaps where strategic priorities do not translate into hiring plans, workforce data rarely turns into coordinated action, and promising talent remains stuck in silos. We need a new way to address HR’s opportunities and issues.

Would agentic disrupt SaaS terrain in a big way?

Most AI adoption has been on the customer side for now, with barely five per cent on the B2B side. That’s understandable given the need for multiple steps that this space entails. There is security, there is compliance, there is workforce training, etc., to be put in place. Most LLM data is public. This space needs expertise with proprietary data. That’s why SLMs are relevant now. But this space also needs integration to drive decisions- and with a good grip on the function in question. Along with apt guardrails and guidance. That’s what we are trying to do. With just-in-time intelligence through unstructured data, which software tools do not always collect?

But there is also the flip side of AI-driven frauds, wrong dismissals and a new wave of bias, especially with citizen-AI tools. Your observation on that?

Yes, all of these are real challenges. That’s where we are trying to work on summaries of biases and alerts while also fine-tuning models and SLMs. I would like to illustrate one of our recent announcements here- what we call an Industry’s first fraud detection AI agent that safeguards hiring integrity by identifying potential candidate deception without slowing decision-making. The agent analyses candidate responses, verifies identities, detects resume falsification, and flags AI-assisted interview responses in real time. It also helps with post-interview analyses, compiling fraud indicators for interview team review.

How much would the fears of compliance shape AI adoption in HR? Especially with the slow pace and lags seen in various surveys recently, and in the wake of the EU AI Act?

Compliance is a key part. It comes not before but later. Users can only comply with known factors and can adapt to the unknown ones. This space is evolving. But those who are committed well will move fast.

Our Unified Orchestration Engine is an adaptive orchestration layer that keeps complex HR workflows running smoothly, handling exceptions in real time. The engine uses decision intelligence and simulations to identify operational bottlenecks, create alternate pathways when rules don’t fit, and ensure actions are policy-compliant and explainable, keeping a human in the loop.

What else stops adoption- presently?

Fear of job losses. And it’s a valid concern. It is complicated, but it can be solved with the right approach. With AI and automation, new jobs would be created, but which ones – we don’t know yet. That ambiguity is real. But India has proven earlier how learning English and mastering software helped us shine on the global IT map. AI is like English. We have to learn it to win the future.

Look at how AI changed the space for radiologists. Their volume actually increased because AI could take care of a lot of time-consuming tasks in that job.

Will this space also run into the Elevator Guy problem- how automation still needs humans – sometimes, just for psychological reasons?

Let’s draw some parallels to various stages of automation in a car. It can vary from being completely manual to driver assistance to semi-autonomous to a self-driving one. Every job would be at a different stage of need for and readiness for automation. Even self-driving car companies would have drivers working in the back-end running the whole system, like monitoring cars. So the jobs will change- the people will stay, and it will be on a case-by-case basis. Look at how AI changed the space for radiologists. Their volume actually increased because AI could take care of a lot of time-consuming tasks in that job. Similarly, when the word processor came, the typewriter went away. The typist did too. But typing survived. Now everyone types on their devices. There are three gaps that AI has not covered – the communication gap, context gap and accountability gap- which is where humans will stay relevant now. But the transition will not be easy.

How should India move on this transition?

The next four or five years would be about figuring it out and working out various concerns. These years would be about foundational challenges. These are very important years. Once that is fixed, the next five years will become about known challenges. India should embrace AI as the new language, covering as much population as possible. It will chart a new growth booster path for the country.

pratimah@cybermedia.co.in