Over the past year, the buzz in technology has undeniably been centred around the rise of AI. And Human Resources is not exempt from this AI wave.

“The question we need to address is whether AI in HR is a fleeting trend or a genuine problem-solving force at scale,” asked Chaitanya Peddi, Co-Founder and Product Head – Darwinbox, addressing an audience of HR leaders at the People Matters TechHR Conference 2023.

“We find ourselves at a pivotal moment,” he added.  This pivotal moment is driven by three primary forces.

First, there's the relentless rise in technology, which is evolving at an unprecedented speed. "The progress we've witnessed in the past 10 to 12 months alone surpasses the advancements of the past decade in both software and hardware," he said.

Secondly, there's an abundance of data. Thanks to digitalisation, organisations have digitised substantial volumes of data, providing a fertile ground for AI models to train and operate.

Lastly, there's a growing acceptance of AI. Scepticism has given way to enthusiasm, with employees increasingly embracing AI in the workplace. A Microsoft survey among employees revealed a strong consensus that AI could enhance various aspects of work, from administrative tasks to creative endeavours and data analysis.

Here are some key takeaways from the keynote on how HR could transform by leveraging AI.

• Addressing Skill Gaps:

To achieve the goal of finding the right person for the right job, you need to intelligently match skills proxies. For candidates, this involves their resumes or CVs, while internal employees have talent profiles. Matching these intelligently with the job descriptions yields impressive outcomes.

Skills and roles are intricately connected, with proficiency levels determined by job descriptions. Visualizing this relationship leads to a complex yet beautiful mesh that represents your skills graph.

For instance, when matching candidates to jobs, you can effectively stack rank thousands of applications to focus on the most relevant ones. Similarly, matching talent profiles to job descriptions within your organization can inform your internal talent marketplace, helping employees discover internal opportunities.

Furthermore, matching talent profiles to CVs reveals whether a candidate aligns with the success profile for a role. At its core, this process relies on precise and effective search and matching. Analysing this data through analytics helps identify the precise skills that are lacking, allowing for targeted learning interventions.

• Contextual Intelligence:

The second category of AI  is "Contextual Intelligence", which involves visual analytics and complements traditional dashboards. For example, consider an engagement dashboard that seems positive at first glance, with a high NPS (Net Promoter Score) and engagement drivers.

However, a closer examination reveals a critical cluster: female employees aged 25 to 30 in a specific location and department who feel disengaged due to work-life balance issues. Identifying such clusters and taking targeted action is challenging through manual analysis but can be achieved through segmentation and recognising critical clusters.

Additionally, the AI capabilities extend to text summarisation and sentiment analysis. For instance, in a peer-to-peer feedback scenario, AI can analyse feedback data throughout the year, generating sentiment scores. By correlating these scores with achievement percentages, actionable feedback and skill improvement suggestions can be provided to employees.

These summarisation capabilities have practical applications in various contexts, from condensing interview feedback at the job level to providing insights in performance management scenarios, where both goal-level and overall feedback are crucial.

• Conversational AI:

In the final part of the keynote, Peddi delved into the evolution of conversational AI, particularly the significant advancements made possible by large language models (LLMs) like GPT-3.

Previously, building a policy bot required manual input of hundreds of frequently asked questions (FAQs). However, LLMs have drastically reduced the need for extensive manual training. Now, you can simply upload your policy documents, and within minutes, a bot is ready to respond intelligently.

Yet, one crucial challenge remains: maintaining context in responses. Consider a global organisation with employees in various countries, each subject to different policies. Providing contextually relevant responses, especially when dealing with diverse employee segments, is necessary. This requires implementing guardrails to ensure the bot responds appropriately to each query.

However, it's crucial to maintain clear boundaries and avoid open-ended generative AI responses that may yield unpredictable results. Ensuring that the AI responds within the trained parameters and relevance to the user is the key to success.