Generative AI in HR: Part 1.5 - Choosing the Right Path

In our previous installment, "GEN AI in HR Part 1" we embarked on a journey to explore the realm of Generative AI, the Gen AI Value Chain, and offered insights for HR professionals to delve into this transformative technology. We also introduced a meta model for pinpointing Gen AI use cases across the employee lifecycle and discussed strategies for engaging with vendors.

Now, in Part 1.5, we revisit Gen AI harnessing the power of visuals to simplify complex concepts. We shine a spotlight on the stark contrast between traditional computing and AI capabilities, unravel the three distinct categories of AI platforms in HR, and equip you with essential questions for making informed vendor decisions.

Parts 1 and 1.5 serve as a foundation, providing insights into the realm of Gen AI. They offer a framework for identifying use cases across the employee lifecycle, distinguishing between various types of AI platforms, and establishing effective vendor engagement strategies.

However, it's important to recognize that the true value of Gen AI lies in its integration within a broader context. To fully harness its potential, we must align its utilization with a Digital HR Strategy that forms an integral part of a larger People Strategy. This holistic approach ensures that our HR initiatives are in perfect harmony with the overarching Business Objectives.

I've already shared Part 4 of GEN AI in HR , which delves into the practical aspects of operationalizing AI and ML within organizations and well aware that the missing link between Parts 1 and 1.5 and Part 4 is a deeper exploration of the Digital HR strategy. I aim to address this crucial component as part of our ongoing learning journey.

The Generative AI Revolution

Generative AI technology is evolving at breakneck speed. The frequent release cycles, the surge of startups, and its seamless integration into existing software applications are astounding.

Generative AI holds the potential to create value across various facets of the employee lifecycle. The initial economic and technical barriers to entry may not be as daunting as they seem, while the consequences of inaction could mean falling behind competitors in the talent race.

Gen AI in HR 1.5 (this piece) draws inspiration from Josh Bersin's insights on AI in HR and distills key reference content for sound decision-making in HR tech adoption and vendor partnerships. To embark on this journey, it's crucial to differentiate between platforms with AI embellishments and those built from the ground up for AI. We will help you determine the right system for your context and guide you through the due diligence process when selecting vendors or partners.

Traditional vs. AI Platforms: Unveiling the Divide

In the simplest terms, AI represents an entirely new paradigm of computing capability compared to traditional systems. It's akin to being offered a Bento Box versus having access to a fully stocked kitchen, allowing you to craft diverse menus and savor unique experiences.

Traditional Platform vs. AI Platform

The basic idea of AI is simple; in a traditional software system, a software engineer codes an algorithm (or workflow) and users then use the software to automate work. As people use the system, they produce data. This data, the output or results of some transaction processing (a payroll run, a candidate applying for a job, an employee registering for a course, etc.), is stored. Thus, the system creates data as more and more people use it.

An AI platform, by contrast, starts with the data. Think about an AI platform as a massive array of data that fuels the AI “models” (complex mathematical algorithms) to learn, understand, classify, predict, and behave based on the data. Then, various applications use these models.

Platform Differences between Traditional vs. AI Systems

Traditional systems and AI (Artificial Intelligence) systems differ significantly in terms of their underlying principles, capabilities, and applications.

Here are some key differences between these two types of systems:

1. Data-Driven vs. Rule-Based:Traditional Systems: Traditional systems typically rely on predefined rules and algorithms. They follow a deterministic approach, where inputs are processed based on explicit instructions. AI Systems: AI systems, on the other hand, are data-driven and learn from data. They use machine learning algorithms to identify patterns and make predictions or decisions based on the information they have been trained on.
2. Complexity of Tasks:Traditional Systems: Traditional systems are suitable for relatively simple and rule-based tasks, such as basic calculations, data storage, and simple decision making AI Systems: AI systems excel in handling complex tasks that involve recognizing patterns in large datasets, natural language processing, image and speech recognition, and playing strategic games.
3. Problem SolvingTraditional Systems: Traditional systems are good at solving problems with well-defined rules and structured data.AI Systems: AI systems can handle unstructured data and tackle more ambiguous problems where human-like reasoning or pattern recognition is required.
4. Decision-Making:Traditional Systems: Traditional systems follow predetermined decision-making rules and do not have the capability to make decisions beyond what they were explicitly programmed for AI Systems: AI systems can make decisions based on probabilistic models and can adapt to uncertain or changing conditions.
5. Human Interaction:Traditional Systems: Traditional systems generally require humans to provide explicit instructions and do not have the ability to understand or respond to natural language.AI Systems: AI systems can understand and generate natural language, enabling more natural and interactive communication with users.
6. Error Handling:Traditional Systems: Traditional systems may fail or produce incorrect results if they encounter unexpected data or conditions that deviate from predefined rules. AI Systems: AI systems can be designed to handle some level of uncertainty and make probabilistic decisions, which can improve error handling in certain situations.
7. Adaptability: One other aspect of second-generation systems is their ability to adapt. Vendors that build on an AI core can regularly replace models with new algorithms, adapt their LLMs, and literally “upgrade” their systems continuously. Traditional first-generation HCM platforms (Workday, Oracle, SAP) find this difficult and may not be able to upgrade their systems very quickly.

In summary, traditional systems are rule-based and deterministic, while AI systems are data-driven and capable of learning from data to perform complex tasks.

Three Categories of AI Solutions

Three Categories of AI Solutions

Emerging AI

Emerging AI is “added on.” These solutions are from vendors primarily using analytics and data management to deliver reports, dashboards, and some predictive models. While these solutions are not powered by large language models (LLMs), do not necessarily use neural networks or generative AI, but they do use advanced analytics to give you insights about your company.

These are all important features, but they are what we call “add ons” in the sense that they don’t change the underlying architecture of the platform but instead they add bolt-on AI features.

e.g. this may be a payroll system like ADP that provides HR managers with recommendations for pay adjustments or a report generated by a human capital management (HCM) platform that points out diversity or gender pay problems.

First Generation

First-generation AI is “built in.” These solutions are from vendors that use AI models and various levels of machine learning (ML), predictive analytics, and candidate or employee matching. They typically have AI engineers, large data sets to analyze, and “single use” ML models that often provide intelligent recommendations to users.

These platforms are “first generation” as they do not use neural networks or leverage external labor market data. These systems use natural language processing, but they do not have the power of GPT-4 or deep neural networks.

e.g. many learning experience platforms (LXPs) or HCM systems recommend courses to individuals based on their job role, activities, or specified skills. Talent marketplaces and recruitment tools have been “inferring skills” for years, creating recommended job matches, career paths, or even mentors.

Second Generation

Second-generation platforms are “built on AI.” These are “next generation” systems that are built for AI from the ground up. Second-generation AI systems can perform deep learning, natural language processing, and LLMs in their core platform, extending functionality to build models that accommodate thousands to tens of thousands of data elements. They also let the vendor build better and more important models, which leverage open source LLMs and new algorithms and extend previous models.

Most second-generation AI systems were built by people with PhDs or serious academic backgrounds in AI, mathematics, or computer science. I/O psychologists are not trained in large data systems or AI processing, so look for a vendor with strong AI credentials. Often, they will share their patents, PhD theses, or other experiences they had building systems that are free from bias.

e.g. recruiting is one of the most powerful applications of AI today, because it demands a massive amount of data and lets vendors build a variety of models to identify who is a good fit for a role by looking at experiences, skills, job history, and more.

Why Are Second Generation Systems Important

In the world of HR, you need far more than the HCM data within your company. If you want a powerful model to identify a perfect salesperson or high-performing engineer you’d want to look at all the salespeople or engineers in the world! That is what second-generation AI systems do, they are built for such computations.

These systems amass hundreds of millions of employee profiles, often anonymized, and often scaling into the billions. And again, because of their architecture, they are designed to integrate hundreds of other data sources: your company’s performance management data, learning data, and actual work data performed by each employee.

Comparative information

Your Vendor Roadmap: Essential Questions

As you navigate the HR tech landscape, here are crucial questions to ask vendors to understand their AI capabilities:

• AI Technologies: What AI technologies are you leveraging, and how do they differ from mere statistical tools?
• Training Data: What datasets fuel your models, and how does your system train itself?
• Data Scale: How large is your data set in terms of profiles or objects?
• Bias Mitigation: How do you ensure your models are free from bias?
• Industry Specifics: Have you tested your data and models in industry-specific contexts?
• Explainability: What provisions do you offer for the explainability of your AI?
• Bias Research: Can you share research on bias reduction in your system?
• Feedback Loop: How does your system utilize human feedback for continuous improvement?
• Customer Insights: Can we connect with your AI users to gauge its performance?
• Skills Validation (for skills vendors): How do you maintain your skills taxonomy?
• Safety Assurance (for recruiting vendors): How do you validate the safety of your recommendations?
• AI Team Expertise: What is the background and experience of your AI team?
• LLMs Selection: Why have you chosen specific LLMs for your system?
• Regulatory Compliance: How does your system comply with AI regulations in <your> jurisdiction?
• Self Awareness - 5 reason to not "buy" your solution and 5 reasons to "buy"

In conclusion, as you embark on the Gen AI journey in HR, equip yourself with knowledge, ask the right questions, and choose partners who can harness the transformative power of AI to drive your HR strategies forward. Gen AI is not just about technology; it's about shaping the future of HR and ensuring your organization thrives in a competitive world. Stay tuned for the next instalment, where we delve into the critical realm of Digital HR strategy. It's all a learning exercise, and together, we'll unlock the potential of Gen AI in HR.

Resources for your learning journey:

This is a great book to get into the details of AI

This a great short course to understand Generative AI

Understanding AI in HR - Josh Bersin

Open LLM Leaderboard aims to track, rank and evaluate open LLMs and chatbots.

Three generations of AI

What every CEO must know about Gen AI

Foundational Models

========================================================

Puneet Sachdev globally experienced and award-winning People, Culture & Change leader who has held leadership roles in the US,?Australia, UK and India in large corporate organizations (e.g. GE, United Airlines), management consulting (e.g. Accenture company) and in technology ventures (e.g.?Fintech ,?Digital Healthcare ?etc). He has a proven track record in coaching HIPO female leaders from all over the world and also helps organizations reduce their leadership gender gap.