There are many trends in 2024 that will have an impact in 2025, but most are simply continuations of what we’ve seen over the last five years such as:

• AI and Automation: Has been around for years
• Remote and Hybrid Work: Has been a strong focus for at least the last 4 years
• Employee Experience: Has been a focus for almost a decade
• Upskilling and Reskilling: Was always a focus
• Diversity, Equity, and Inclusion: Has been a focus for the last half decade
• Mental Health and Well-being: Like remote work has been a focus since COVID

There is nothing new in the above, so what is the one big, truly new trend we think hasn’t been seen before? What is it that wasn’t generally available in 2024 that we think will bloom in 2025? 

Enter the multimodal, conversational AI job interview agent.

The Job Interview Agent

For years companies have worked on the dream of a virtual, fully automated, human like candidate interviewing agent and one sided versions do exist. For example HireVue has a real candidate access their platform and answer a number of questions on a video interview after which it analyses the video using AI to provide a summary to hiring managers. However, this is not the same thing as a conversational AI job interview agent, who you can see, who reacts to your responses and adapts the dialogue, can be interrupted and interrupt you in real time. An agent that can even assess work you may show it when sharing your screen.

Achieving this has been a significant challenge - until now.

A number of recent breakthroughs are making a real time, conversational AI agent possible. These breakthroughs include more capable, smaller large language models that deliver high quality outputs with less than 10 billion parameters (many of the powerful original models such as ChatGPT had over 100 billion parameters), specialised AI processors (so called LPUs - language processing units created by companies like Groq) and experience with optimising and running LLM pipelines with tools like Langchain.

Underneath the hood companies are moving away from just large language models and are leveraging multimodal generative AI tools which integrate diverse data types—such as text, images, audio, and video—enabling them to perform complex tasks across various input formats. By combining these modalities, such technologies enhance comprehension, context-awareness, and accuracy, making them versatile for complex interactions with candidates during an interview. At the end of the interview, this AI agent can then transcribe notes from a dynamic and unpredictable interview, create an assessment and summary of the discussion. An incredibly powerful technology.

How do we ensure they tell the truth?

As is well known, many large language models can hallucinate causing major issues for interviewing candidates, especially if they start making up non-sensical questions or responses. So why are these conversational AI agents not falling victim to these issues?

The power of many of these conversational bots is due to is called retrieval augmented generation (RAG). RAG is a technique that enhances the accuracy and reliability of AI models by allowing them to access and reference external information sources. Instead of relying solely on their training data, RAG-powered models can retrieve relevant facts from a knowledge base to provide more accurate and informative responses. This makes them more adaptable and capable of handling a wider range of queries, especially those requiring up-to-date or domain-specific information or ensuring that accurate answers without hallucinations are given.

In the HR case what it means is that when a candidate (or even a newly onboarded employee that is trying to orient themself in finding corporate documents, policies or relevant procedures) interacts with a conversational AI agent, instead of the conversational AI agent just responding with what its large language model would provide on its own, it takes that input question and turns it into a query of a company’s database to ensure that the responses it returns are based on a ground truth of information that the company has validated and provided as a basis for correct answers.

Conclusion

These developments are fascinating as it is not just employers that can deploy these tools but candidates too. During 2024 social media lit up when HeyGen released a video showing an AI conversation job interview agent interviewing another conversational AI agent, that was sent on the behalf of an employee. This was just a demo, however real tools such as those created by the company FairGo.AI, can now do this and it is a very powerful technology.

On the flip side, it may portend a future where we will see an AI arms race between candidates and employers (even without employers using these tools, it is highly likely we will see job seekers using such tools). We have seen this already with keyword stuffing in CVs but it will clearly become more sophisticated, leading to questions such as do non-AI savvy candidates still have a chance? How will they differentiate themselves when more AI literate job seekers can apply for 1000 job applications at once with AI-tailored CVs for each job and an AI agent carrying out the interviews for them? Will these new technologies exacerbate the current gap between the technically literate and the illiterate? 

The only thing that is certain is that these technologies will keep evolving and, short of strict regulations, will have an enormous impact on the way we hire candidates in the future.