Skip to content

Playbook: From content to capability: Scaling people with AI

, , ,

AI is often sold to Learning & Development professionals as a tool to create more content, faster. But what if its real value is in helping us create less content and more capability? In L&D, we're constantly asked to build new courses, design varied content and build learning pathways. But often, this isn't the right solution. The real challenge is to solve performance problems and enable people in the flow of their work.

This playbook, born from a co-creation session with the L&D Leaders Community, offers a practical framework for using AI to scale your people, not just your content library. We'll break down two real-world examples of AI tools that automate time-consuming processes and deliver personalized development at scale, helping you shift from being a content creator to a strategic performance partner.

By following this playbook, you'll learn how to identify high-impact AI use cases, partner with technical teams, and build solutions that solve real business problems.


Playbook at-a-glance

This guide will walk you through a strategic approach to implementing AI in your L&D processes.

The Steps:

  1. Identify high-impact processes: Find the operational tasks that touch everyone in the organization.

  2. Build a "smart" interface (not another course): Automate the process directly within the workflow.

  3. Create an AI Performance Coach: Deliver personalized development plans at scale.

  4. Fine-tune and mitigate risks: Address AI "hallucinations" and set usage limits.

Prerequisites:

  • Access to organizational data (e.g., job descriptions, performance feedback, competency frameworks).

  • A mandate from leadership to explore and implement AI solutions.

  • A collaborative partnership with a technical team (or access to AI development tools).

Deep dive: 4 steps to scaling capability with AI

Step 1: Identify high-impact, "touch-everyone" processes

The Goal: To find the best starting point for your AI initiatives. Instead of chasing dozens of small ideas, focus on processes that are universal, operational, and have the potential to impact the entire organization.

The Action:

  • Look for universal pain points: What are the recurring, time-consuming tasks that every employee and manager has to complete?

  • Analyze your L&D requests: Are you constantly being asked to create training for the same things year after year?

  • Prioritize based on reach: Ask yourself: "What process touches the majority of our organization at least once a year?". In our case, we identified two: annual goal-setting and creating yearly development plans.

[Voice from the Room] "We looked at it from this perspective: What can we do that is very operational but still has huge value for our people and organization?"

Pro-tip: Start with operational processes that L&D is often asked to support, but which aren't strictly "learning" issues. Goal-setting is a classic example. It's a business process that L&D is tasked with "fixing" through training, making it a perfect candidate for AI automation.

Step 2: Build a "smart" interface (not another course)

The Goal: To make the desired action easier to perform directly within the user's workflow, eliminating the need for separate training.

The Action:

  • Take the "SMART goals" example: For decades, L&D has run workshops on writing SMART goals. The impact is often limited and requires constant reinforcement].

  • Automate the "how”: Instead of teaching the theory of SMART goals, we built an AI assistant directly into our goal-setting system. This enables an employee to write a rough draft of their goal, click "Write with AI," and the tool refines it into a measurable, specific, and time-bound objective.

  • Provide organizational context: It’s imperative that you provide the right context regardless of which model you use. The AI was trained on our company's job architecture, descriptions and overarching strategic goals. This ensures the "smartened" goals are not just well-written but also aligned with the individual's role and the company's direction.

Pro-tip: The key is to embed the AI where the work happens. By making the interface smart, you remove the friction and cognitive load from the employee. You're not asking them to remember a framework from a workshop; you're giving them a tool that does it for them, instantly.

Step 3: Create an AI Performance Coach for Personalized Development

The Goal: To automate the creation of customized learning plans for every employee, moving beyond manual Training Needs Analysis (TNA) and generic course catalogs.

The Action:

  1. Build a multi-agent system: We developed an "AI Performance Coach" using four distinct AI agents to form a crew which work together seamlessly.

    1. Agent 1: Summarize Performance Data: This agent gathers and summarizes an individual's performance data, including self-assessments, manager feedback, and 360-degree reviews.

    2. Agent 2: Visualize Competencies: This agent takes the summary and creates a radar chart, showing the employee's strengths and development areas against the organization's core leadership principles.

    3. Agent 3: Verify Information: This agent cross-checks the data to ensure accuracy and reduce errors before the final plan is generated.

    4. Agent 4: Generate the Development Plan: Based on the verified data and identified gaps, this agent creates a 6-month development plan. It recommends specific resources from our learning libraries (like LinkedIn Learning) and can even suggest other interventions like coaching or mentorship.

[Voice from the Room] "The general chat about doing TNAs has actually reduced a lot because we have a customized program already... No one in the organization is without a learning plan for the year."

Pro-tip: Connect the AI to your existing learning ecosystem. The AI coach is most powerful when it can recommend content you already have, whether it's in your LMS, a third-party library like Coursera, or internal coaching programs. This ensures the recommendations are immediately actionable.

Step 4: Fine-tune and mitigate risks

The Goal: To address the practical challenges of implementing AI, such as cost management and model accuracy.

The Action:

  • Set usage limits: AI tools that use APIs, like OpenAI, have associated costs. We quickly found that high usage led to significant financial exposure. The solution was to implement a reasonable usage limit per person to manage the investment effectively.

  • Acknowledge and reduce "hallucinations": AI models can sometimes generate incorrect or nonsensical information. We found our model was about 90% accurate, but that 10% error rate is still significant.

  • Fine-tune the model: By continuously training the AI with more specific organizational context and data, we were able to minimize the error rate. We also built in verification steps (like the third agent) to catch errors before they reach the user.

Pro-tip: Be transparent about the AI's limitations. When launching, communicate that the AI is a "coach" or an "assistant," not an infallible judge. Encourage users to review the AI's output critically and provide feedback, which can then be used to further improve the model.

Troubleshooting: Common pitfalls & fixes

  1. Pitfall: The AI becomes a "box-ticking" exercise, where employees complete courses without applying the knowledge.

    • The Fix: Integrate the competencies into the entire employee lifecycle—from performance reviews to promotion criteria. If demonstrating a skill is tied to career progression, people are more motivated to truly learn and apply it.

  2. Pitfall: The AI only recommends formal courses (the 10% of the 70-20-10 model).

    • The Fix: Program the AI to recommend a variety of interventions. Based on the data it receives, it can suggest coaching, mentorship, or even connecting with "culture champions" who embody specific skills within the organization.

  3. Pitfall: You don't have the "perfect" data to start.

    • The Fix: Start with what you have. The quality of the AI's output depends on the quality of the input data. Begin with a single, reliable data source, like manager feedback, and expand from there. The key is to start experimenting and build momentum.

Implementation Checklist

  •  Identify a universal process: Pinpoint a recurring, organization-wide task that L&D is asked to support (e.g., goal setting, performance reviews).

  • Form a partnership: Secure collaboration with your IT or a dedicated tech team.

  • Start with one pain point: Choose one specific problem to solve, like improving goal quality or automating TNA.

  • Build a prototype: Create a simple version of your AI tool to test the concept.

  • Gather contextual data: Feed the AI with job descriptions, competency frameworks, and strategic goals.

  • Embed the tool in the workflow: Integrate the AI assistant directly into the software people already use.

  • Connect to your learning library: Link the AI to your LMS, LinkedIn Learning, or other content sources.

  • Test and fine-tune: Run experiments to identify and correct inaccuracies or "hallucinations."

  • Monitor costs and set limits: Keep an eye on API usage and implement caps if necessary.

  • Communicate transparently: Launch the tool with clear instructions and manage user expectations about its capabilities and limitations.

This article was written by Saqib Riaz, Manager, Talent Management & Performance & AI Academy Lead at Jazz, is an HR and organizational development leader who specializes in talent management, workforce culture, and building AI capability within businesses.

Enjoyed this perspective? Saqib is part of our global community of L&D leaders who are actively driving change through collaborative learning. If you want to learn with peers like Saqib, apply to join our community.

Related Articles