Beyond ChatGPT: How Talent Teams Can Use AI to Get More Out of Their Data
We've all heard it before: data is the new gold. Sure, but only if you can use it. Yes, some parts of organizations have gotten pretty good at capturing data and using it for decision-making. Marketing is a prime example of a function that's evolved into a data-driven powerhouse.
However, the same can't be said for talent teams, despite years of overblown hype and costly, often painful, software rollouts. These modern tools — HCMs, LMSs, LXPs, and other people analytics tools — while useful, still leave a lot of valuable data woefully unanalyzed. The data these tools gather is often easy pickings, neglecting a wealth of underutilized, unstructured, and qualitative data.
While many talent folks have at least experimented with AI Chatbots like ChatGPT and some are using them to generate content in their daily work, few have taken the next step of creating a co-pilot to help analyze their wealth of data.
We've been helping talent teams tap into their data by building AI tools and extracting valuable insights from data that would've otherwise sat idle. The shift we're seeing–talent teams are starting to build up their own data analytics muscle, enabling them to become more strategic and deliver faster results.
What we’ve seen
There are several software vendors offering valuable tools in a narrowly defined scope. Still, their clients end up wanting more. Companies are obsessed with data collection and hoarding but aren't getting insights out of it. Mission-critical data is analyzed manually by non-experts, opening it to errors, bias, and flawed methodology. To solve for this, many enterprises pile on more data without a solid plan for analysis or impact improvement. The good news is: talent teams can build their data analytics capabilities using AI tools. We’ll start by outlining the current state of affairs before we dive deeper into what steps you can take to up your data game.
The underutilized data you already have:
Historical hiring, promotion, attrition data
Engagement surveys
Performance evaluations, goal-setting, 360s, etc
Job descriptions
Exit interviews
Post-program feedback
Knowledge management documents
The insights you’re missing:
Organization-wide skill gaps
Future hiring needs
Impact and improvement opportunities for learning programs
Quality of goal setting and feedback
Factors that drive engagement or disengagement
What the organization truly expects from the talent team
Predictors of promotion or attrition
Competency models/skills profiles based on hard facts
X factors for each role
So, how do we tap into this potential? We've seen a four-step approach work well that non-experts can experiment with themselves:
Step 1: Data Preparation and Aggregation
Before you can utilize AI data analysis, you need to gather and prepare your data.
Identify Relevant Data Sources: Such as historical hiring, promotion, attrition data, engagement surveys, performance evaluations, job descriptions, exit interviews, post-program feedback, and knowledge management documents.
Consolidate Data: Merge this information into a unified data set. This step might require some manual efforts, such as converting different data formats into a common format, reconciling data from different systems, and cleaning the data.
Anonymize Data: Ensure all sensitive information is appropriately anonymized to respect company policy, privacy laws and ethical guidelines.
Convert to Vector Format: Convert this data into a vector format that can be ingested by Pinecone or a similar service.
Step 2: Load Data into Vector Database
Add your data to a vector database, effectively creating a form of long-term memory for your AI.
Choose Your Tools: select a Vector DB provider like Pinecone and your preferred cloud provider like AWS that meets org security requirements.
Create a Vector Database: Using Pinecone or a similar service, create a vector database.
Load Data: Add your vectorized data to the database. This step is the equivalent of storing your data for the AI as a form of long-term memory.
Step 3: Model Testing and Validation
The next step is to use this data to train and validate an AI model on the datasets you collected.
Choose an Appropriate Model: Use an AI model that's capable of interacting with vector databases.
Testing the Model: Connect your model to your vector database. The model should learn to use the long-term memory provided by the database to better understand the data and generate more accurate insights.
Model Validation: Validate the model's performance by comparing its predictions or insights with known outcomes. Refine the tool as necessary based on these results.
Step 4: Insight Generation and Interpretation
The part we’ve all been waiting for. Once the model is tested and validated, it can be used to generate insights.
Generate Insights: Use the model to analyze the data and generate insights related to your focus areas. These could include identifying skill gaps, predicting future skills needs, assessing the impact of learning programs, evaluating the quality of goal setting, and identifying factors influencing engagement and attrition.
Interpret the Results: AI-generated insights need to be interpreted and contextualized by human experts. Interpretation should consider the context of the data, the objectives of the analysis, and the practical implications of the insights.
Implement Changes: Apply the insights into actionable strategies. This could involve changes in hiring practices, modifications in learning programs, refinement of goal-setting processes, or interventions to improve engagement.
The result
The shift we're seeing is one of empowerment. Leading talent teams are no longer just consumers of data; they're becoming insight-driven strategists. They're using AI-enabled analytics to weave valuable narratives, predict future trends, unearth hidden variables, and develop actionable strategies.
The path is clear for enterprise talent teams—they must extract more value from their existing data collection process, build up their teams’ capabilities and get better results, faster.
The real question is when will you embark on this journey?
