Artificial Intelligence for SMEs in the HR function - an Introduction
Context
Excitement over developments in Generative Artificial Intelligence (GAI) has exploded after the announcement of ChatGPT by OpenAI in Nov 2022. Both enterprises and individuals across the world have started adopting it for enhancing their productivity.
Various surveys have shown increasing use of GAI by enterprises across all sectors of the economy. Unfortunately large enterprises are much ahead of the smaller ones in GAI adoption, leading to fears of an AI divide. Mostly smaller businesses are constrained by a lack of awareness and resources. Thankfully rapid development in low cost and open source resources related to GAI have made it possible for small businesses also, to start exploring the use of GAI.? To make my point I will show how a small business can deploy low cost and open source GAI models in the HR function. But first it will be useful to get a sense of what is GAI and how enterprises can benefit from it.
An introduction to GAI
GAI is a sub field of AI. GAI is mainly based on Transformer models. Transformer models were first proposed by Google researchers in the famous ‘Attention is all you need’ paper in 2017. The field has made rapid progress since then. Transformer models can be developed to deal with text, images and audio data, but recent advances have seen the emergence of multimodal models that can deal with multiple forms of data at the same time. The term generative implies that the user can prompt the model with an input and the model will then generate a response. If you have not tried using a GAI model till now, you can easily use the free Gemini model from Google at this link. https://gemini.google.com/app?hl=en-IN. If you type the prompt ‘What is the capital of India ?’ Gemini will respond saying ‘The capital of India is New Delhi. It's a fascinating city that's both historic and modern, and it's home to the Indian government.’
GAI models first have to be trained on extensive data and then they can be deployed on the cloud for inference and used by users through the web. Training large GAI models is very expensive, with costs running into millions of dollars, so only large companies are able to undertake this task. In turn inference costs can also be high. For example ChatGPT costs 20 $ per month for individual users. However, as I said earlier, SMEs now have the option of using low cost and open source resources.
How can an enterprise use GAI in the HR function ?
Now that we have seen what a GAI model is and what it does, we will explore how an enterprise can use it in the HR function. Individual users can use a GAI model to generate text, images and also sound for fun or for helping them in their work. However, an enterprise needs to solve specific business issues. So valid use cases need to identified and explored. How should a business get started on this journey.
Businesses are organized in terms of various functions like finance, HR, operations etc. While an enterprise wide strategy for use of GAI can and should be explored, it is often easier for the top management to encourage functions to start by exploring use of GAI in their respective domains. In this article I will throw light on this by presenting a very simple example of use of GAI by the HR department in the recruitment process.
Why start with a simple example. Can’t we be more ambitious ?. It is true that GAI is being used extensively in automation of almost all processes related to HR. However it is a journey and it is best for a company to start with simple use cases to gain confidence and build over time.
Simple example of use of GAI by the HR department in the recruitment process.
The recruitment process comprises of following basic steps.
-??????? Create a Job Description (JD)
-??????? Source Curriculum Vitae (CVs) of candidates against the JD
-??????? Shortlist CVs for interview
-??????? Conduct interviews
-??????? Issue appointment letters to selected candidates.
The ownership of the recruitment process lies with the HR department. In may companies especially in the IT sector there is a lot of churn and HR is under pressure throughout the year on recruitments. So they would welcome any tools that can help speed up their work.
In our example we will consider how GAI can help in the third step which is shortlisting CVs for interview. This step is generally a joint exercise by the HR recruitment specialist along with the department manager that raised the recruitment request. So let us understand the use case better.
In the second step of the recruitment process the JD is publicly shared on various job portals and candidates apply by mailing their CV to the company. Against any JD several hundred CVs are received and sifting through this deluge of CVs to select relevant ones is a tedious and time consuming task. So how can GAI help in this ?
In our simple example we will use the summarization functionality of GAI models, more specifically large Language Models (LLMs) to summarize the hundreds of CVs that we received, so that sifting through them is easier. We can do the screening in three steps.
-??????? Feed all the CVs to an LLM for summarization.
领英推荐
-??????? Go through all the summarized CVs to select the interesting ones.
-??????? Take a closer look at the selected interesting CVs for final shortlisting for interview.
Let us see how well an LLM can summarize a CV. I asked Gemini to summarize the CV of a software engineer XYZ and here is the output it provided me.
XYZ is a software engineer with over 4 years of experience. He graduated from IIT Kanpur in 2018 with a Bachelors in Mathematics and Scientific Computing. He is currently working at ABC as a Senior Software Developer.
XYZ has experience working on a variety of projects, including:
XYZ is skilled in a variety of programming languages and technologies, including JavaScript, Golang, Python, NodeJS, React, MySQL, MongoDB, Redis, AWS, Digital Ocean, and Docker.?
Not bad. Gemini summarized a 455 word CV in 85 words. We can control how concise we want the output to be. This is a good proof of concept but this process ?is not suitable for process automation is an enterprise context. Why ?. This is what we did.
-??????? We copy pasted from the CV PDF to Gemini.
-??????? We asked Gemini to summarize.
-??????? We copy pasted the summary for later review.
What we need is a simple app running in the cloud to which we can load all the CVs as a batch and get the output back as a set of summary files. Fortunately we can easily get this done at a very affordable cost. For illustration you can see my code for a simple implementation of the LLM part of the app here.
I used the recent Command R+ LLM, developed by Cohere, a company cofounded by Aidan Gomez , one of the co-authors of the Google ‘Attention is all you need’ paper mentioned earlier.
Conclusion
Samuel Colt invented the revolver in 1836 and he coined the slogan ‘Be Afraid of no man, no matter what size.?When trouble threatens, call on me, For I shall equalize’
In today’s time the wonderful Open Source community is providing the equalizer to SMEs in the form of free GAI models. It is for us to take advantage of this equalizer. Hopefully this brief article will motivate you to explore possibilities. Having helped several SMEs to benefit from GAI, I can say with confidence that almost every SME will be able to identify use cases and implement affordable solutions. Many professionals in the field including me can help you in this exciting journey. My coordinates are as follows.
-??????? Mobile/WhatsApp +91 9910995649
-??????? Mail ID [email protected]
-??????? GitHub https://github.com/sudhir2016
-??????? Hugging Face https://huggingface.co/sudhir2016
Dean (R&D) at Swami Keshwanand Institute of Technology, Management and Gramothan
7 个月Sudhir. Wow what in brief you have explained a use case for HR in IT sector where HR really struggles in speedy filling of posts.