Skills Management and Capability Orchestration for Redesigning Work in the AI Era
Directionally Correct Newsletter, The #1 People Analytics Substack
By: Angela Le Mathon & Cole Napper
Description: Why AI-era organizations need capability orchestration, not just skills cataloguing
The Headlines tell the Story
Moderna merges HR and IT to manage 3,000 AI agents alongside 5,300 humans. IBM lays off 8,000 employees, only to rehire just as many for AI automation roles. Goldman Sachs pilots its first autonomous coder. ServiceNow acquires Moveworks to strengthen agentic AI capabilities.
This isn't the future, it's happening now. Organizations worldwide are rapidly deploying AI agents as digital workers, fundamentally reshaping how work gets done. Yet most companies are still thinking about this transformation through the lens of traditional skills management instead of skills and capability orchestration.
With the current technology, your digital teammate is like a brilliant intern who's read the entire internet but takes everything at face value, never gets angry, never asks for vacation, and will confidently give you directions to places that don't exist. Managing these "clever amnesiacs" requires a fundamentally different approach..
Skills Inference Models
Having led skills data infrastructure at GSK, Angela witnessed this disconnect firsthand. Traditional platforms excel at answering "What skills do we have?" and "What skills do we need?" by analyzing employee records and building comprehensive taxonomies for human to opportunity matching. The problem arises when doing this in practice.
Consider two employees with identical skills profiles: Same technical capabilities, same experience levels, same assessment scores. Employee A is given advanced tooling, AI support, and streamlined processes. Employee B works with legacy systems and manual workflows.
Six months later, Employee A delivers significantly more work while Employee B struggles to meet basic targets. Based purely on skills data and performance outcomes, the system recommends more training & coaching for Employee B while Employee A gets promoted.
Was this the intended goal of the decision making process? Probably not.
The honest reality is that most managers couldn't tell you what skills are required for a role/job description, nor could they agree on those assigned by AI models. They need better data. As leaders review candidates from suggested matches, many question or challenge the methodology, leading to credibility concerns about scoring algorithms. In essence, the business couldn’t trust the AI not because it was incorrect but because they didn’t understand what to look for in the first place.
Ask the same managers about tasks, the actual work that needs to be done, and why they do or don't have sufficient talent to accomplish them and these same managers could rattle off responses in seconds, nodding profusely in agreement. This proves that the pain point was always about getting work done and that skills on their own are incomplete.
This growing dichotomy between how we think about work (tasks and capability orchestration) vs. how skills-based SaaS platforms approach it (abstract skills) is why many skills transformations fail and why the agentic era demands a different approach entirely. They don’t have the right data, and they lack context of the internal and external labor market.
Today’s ambitions for agentic AI envision a world where tasks are distributed among humans (47%), agentic AI (22%), hybrid collaboration (30%), an often quoted projection in 2025 (McKinsey 2025). Who’s to say if this timeline is accurate, but in this new reality, bridging the gap between skills and task reality becomes a key competitive advantage.
The agentic era demands that we connect the abstract lack of skills data to the tangible execution of tasks. Combining real skills intelligence based on labor market data and embedding it within a real-time capability orchestration framework.
Industry Leaders Signal SaaS Evolution
Managers struggling to implement some skills-based SaaS tools because these systems have revealed the fundamental business question has changed. Instead of "What skills do our people have?" organizations now need to ask "How do we get work done efficiently across human skills and AI capabilities?" Managers lack the right data to make this happen.
This isn't theoretical. Industry leaders are already signaling this transformation:
Satya Nadella recently predicted that traditional business applications will probably “collapse in the agent era," as AI agents interface directly with databases rather than application layers.
Andrew Ng emphasized that 'building practical applications through agentic workloads' rather than chasing the latest foundational models will be the greatest opportunity for businesses.
Marc Benioff describes moving into "the world of digital labor" where organizations create "digital workers" alongside humans in what he sees as a $12 trillion opportunity.
Even research firms are aligned: Gartner predicts that 33% of enterprise software applications will include agentic AI by 2028, while McKinsey research shows organizations increasingly treating AI agents as workforce members rather than tools – however silly this may seem. Additional, if it’s SaaS that’s evolving, it’s data-as-a-service (DaaS) that is positioning the transition. DaaS is what powers the AI applications of tomorrow, and in partnership with Gen AI wrappers, will be the catalyst of the next wave of skills and task applications.
This creates an immediate strategic choice for skills platform providers: Repurpose your skills data infrastructure to power task and skill orchestration based in real data, or watch organizations build capability optimization directly into their operations without you.
Understanding Human Machine Collaboration
The shift from skills management to full capability orchestration reflects a fundamental change in how work gets done. Traditional human-to-human collaboration relies on shared context, intuition, and accumulated wisdom. Human-AI collaboration requires a different approach entirely.
AI agents operate as stochastic simulations of people in probabilistic environments, hallucinating with jagged intelligence, prone to anterograde amnesia and surprisingly gullibility. Said another way, your digital teammate is like a brilliant PhD intern who’s read the entire internet, takes everything at face value, never gets angry, never asks for vacation, but also never learns from their mistakes and will confidently give you directions to places that don't exist.
The old model: Human assigns tasks, machine executes, process finishes; assumed clear boundaries between human judgment and machine execution. The emerging model requires continuous human-AI loops where strategic decision-making, autonomous reasoning, and creative problem-solving flow seamlessly between capabilities that complement but differ fundamentally.
This situation creates an interesting problem for organizational culture. As companies introduce a cast of eager savants with known cognitive issues (i.e., agents), does culture become more human because the agents pick up all the undesirable work, or does it evolve into something else? Our focus is squarely on the lift agents can make via automation and productivity gains – not wholesome replacement of “human” tasks.
This perspective creates a powerful opportunity: When you have tasks mapped to roles, generative AI can accurately identify the skills needed to execute that work if they have the right data. This task-plus-skills combination enables real-time talent supply and demand analysis, powering workforce analytics and strategic workforce planning with unprecedented precision.
Four Requirements for Capability Orchestration
To operate in this paradigm, organizations must evolve from static skills inference to dynamic capability orchestration through systems that can:
1. Split Tasks Between Humans and AI Agents
Rather than cataloging what people can do, organizations need systems that determine optimal task allocation that determine what tasks should be agentic versus which require human problem solving skills. This means evaluating every role for AI exposure and systematically resizing jobs based on capability optimization across hybrid teams.
2. Create Dynamic Capability Blueprints
Instead of traditional skills inventories, organizations need living blueprints labor data that continuously adapt as AI capabilities expand and human roles evolve. These blueprints map real-time relationships between skills, tasks, and outcomes. Outcomes-focus is key, and has been lacking from too many internal initiatives in this space until now.
3. Validate Skills Against Real Outcomes
Legacy skills inference SaaS tools rely on historical data (i.e., resumes, performance reviews, project descriptions) to predict capabilities. It’s not enough. Agentic systems correlate actual task performance with inferred skills, creating feedback loops that improve both AI task allocation and human skill development.
4. Orchestrate Continuous Optimization
Advanced organizations can build systems where AI agents manage the capability optimization process itself, validating data integrity, reviewing logical consistency of skill-task mappings, and assigning confidence thresholds for downstream talent decisions.
The result: Organizations that assign work to humans or machines based on performance optimization, using skills data for career matching and for real-time task allocation across autonomous teams.
When AI agents execute most technical tasks, human value shifts from what you know to what you can accomplish. Curiosity, persistence, creative problem-solving, and the desire to tackle complex challenges become the scarce resources that determine career trajectory.
In this model, someone with moderate technical skills but high engagement often contributes more value than someone with advanced credentials but limited willingness to adapt. The capability blueprint doesn't just map skills to tasks, it identifies humans who actively want to solve problems that machines cannot.
What's Next
AI workforce transformation is the new hot field in HR. The shift to capability orchestration is more than a tech upgrade — it’s a redesign of work for the AI era. As AI agents join the workforce, leaders must master real-time orchestration of human expertise and machine capabilities.
AI can be brilliant and confidently wrong, which demands oversight, context, and validation. The payoff? Humans are freed from routine tasks, and value shifts from simple recal to a productivity and outcome focus.
Skills intelligence, using DaaS and AI, become the foundation for capability systems that align human motivation with work that truly needs it, while AI handles the rest. The organizations that embrace this shift will define the agentic workforce era not by replacing human choice, but by amplifying it.
Angela Le Mathon is the Chief AI Officer at People Alkemie, a boutique advisory firm that serves as Fractional CAIO to purpose-led organizations. She previously led skills data infrastructure at GSK and is currently completing the Chief AI Officer program at University of Chicago Booth.
Cole Napper is…well Cole Napper. “People Analytics: Using Data Driven HR and Gen AI as a Business Asset” comes out Aug 26th!
I hope you like this article. If so, I have a few more articles coming out soon. Stay tuned. If you are interested in learning more directly from me, please connect with me on LinkedIn.
Cole’s recent articles
The Global Wage Rate Is Coming: How We All Get Paid the Same
Who Needs a “Human in the Loop” When AI Gives Itself Feedback
Beyond Prediction: Exploiting Organizational Events for Causal Inference in People Analytics
For access to all of Cole’s previous articles, go here.