Breaking Metrics

CHANGE YOUR THINKING

The Great Inversion of Performance Management

Last week, a curious experiment has started an effect. Like the butterfly effect, we may not appreciate the result just yet but I think it is inevitable. Nathalie Salles-Olivier, a HR leader researching AI-driven HR practices, shared what happened when she asked ChatGPT 4.1 to dream up a performance system for 2045. The AI didn't propose more sophisticated dashboards or real-time analytics. Instead, it conceived something called "LUMA" (Living Understandings of Mutual Agency) and a "Ledger of Unseen Labor"—frameworks designed to capture the invisible aspects of work that typically go unmeasured.

This comes amid a broader transformation in how companies operate with AI. According to Microsoft's 2025 Work Trend Index, 81% of knowledge workers expect AI agents to be extensively integrated into their organizations within 18 months. Meanwhile, tech CEOs from Shopify's Tobi Lütke to Duolingo's Luis von Ahn are declaring their companies "AI-first," signaling a fundamental restructuring of work.

The Curious Case of Stories as Signals

It's a peculiar thing, when you stop to consider it, that we've spent the better part of a century trying to reduce the gloriously complex tapestry of human endeavor to a series of digits that would fit comfortably on a Post-it note.

Even more curious is that we've done this with such unwavering conviction that one might think numbers were handed down from the heavens themselves, possibly engraved on stone tablets alongside various prohibitions about coveting oxen.

The statement "The Ledger doesn't correct for bias with numbers. It reclaims story as signal" is rather like discovering that we've been trying to measure the ocean's majesty with a kitchen measuring cup all along.

Let’s Reclaim Story as Signal

You see, traditional performance systems operate under the charming delusion that bias—that most persistent of human foibles—can be eliminated if we just find the right algorithm, as if bias were a stubborn kitchen stain that would yield to the proper mathematical detergent.

This approach has all the practical efficacy of trying to cure pneumonia by adjusting the thermometer.

Stories. Measuring that is gold mine of understanding real value. Not anecdotes, mind you, which are what managers often substitute for actual insight.

Rather, stories that are structured narratives that capture human contribution in all its messy, contextual glory. Narrative is how we make sense of the world. Our ancestors didn't sit around fires comparing quarterly productivity metrics. They told stories about the hunt, about courage and cowardice, about who could be trusted when things went sideways.

OLD DOGS AND NEW TRICKS

Why Traditional Metrics Fail in the AI Era

In my analysis, we are experiencing what I call "The Great Inversion" in talent economics. In the fully automated firm, you don't hire skills. You compute them. What becomes expensive? Roles where massive compute creates disproportionate returns.

This inversion exposes the limitations of traditional performance metrics designed for a pre-AI world. When AI can instantly generate perfect documentation, flawless code syntax, or error-free transactions, measuring humans by these standards becomes meaningless.

The Business Case for Measuring the Unmeasurable

A CEO of a large enterprise would certainly caution against abandoning metrics entirely.

Yet research suggests frameworks like LUMA might deliver tangible business results.

The research presents compelling evidence regarding the substantial portion of organisational value derived from "non-contractible elements"—aspects of work valuable but impossible to specify in contracts. As much as 20% of a company's economic value is under-measured as non-contractible elements.

A striking 79% of estimated total global intangible asset value, derived from human labour, is not accounted for in company financial reports, highlighting the significant gap between formal accounting measures and actual organisational value. Human capital accounts for approximately 62% of total wealth in studied countries - four times the value of produced capital and 15 times the value of natural capital.

The Sage in AI

In our fervent attempts to be scientific about performance, we've created measurement systems that would make a quantum physicist blush with their pretensions to precision.

Meanwhile, what actually matters—that Janet talked a frustrated client down from the ledge at 7 PM on a Friday, or that she mentored three junior colleagues through their first crisis—remains frustratingly outside our measuring cups.

Salles-Olivier’s The Ledger of Unseen Labor, in its oddly artificial wisdom, suggests that perhaps the signal we've been searching for has been there all along, encoded not in numbers but in stories:

It's rather like discovering that after spending billions on increasingly sophisticated weather satellites, the most accurate forecast comes from watching how the neighborhood squirrels prepare for winter. Sometimes the most sophisticated signal processing system is the one we're born with—the human capacity to tell and understand stories. And that, when you think about it, is a rather comforting thought in our algorithm-obsessed age.

The Three Horizons of Performance Evolution

Based on patterns emerging across industries, enterprises appear to be moving through three distinct phases of performance measurement in response to AI:

Horizon 1: Augmentation (2023-2025)

Organisations initially use AI to enhance existing metrics—making traditional KPIs more accurate, frequent, and personalised. This creates incremental improvements but maintains the fundamental measurement paradigm.

Horizon 2: Redefinition (2025-2027)

Companies begin distinguishing between "AI-suited" and "human-suited" tasks, developing dual measurement systems. Technical metrics apply to tasks where precision matters most, while human contributions are measured through what Dwarkesh Patel calls "essence over likeness"—focusing on uniquely human capabilities.

Horizon 3: Integration (2028+)

Organisations adopt integrated frameworks similar to LUMA, where AI doesn't just measure performance but helps reveal previously invisible value. These systems treat technical outcomes and human processes as inherently interconnected rather than separate domains.

The Four Principles of Next-Generation Performance Management

For enterprises navigating this transition, experts recommend four guiding principles:

1. Value Context Over Replication When AI can instantly create perfect deliverables, the value shifts from the deliverable itself to the contextual understanding that shaped it. As leadership researcher Stanley Silverman notes: "You can't optimise what you can't see."

2. Measure Restraint as Much as Action In a world of AI acceleration, knowing when to slow down or redirect becomes as valuable as pushing forward. Providence Healthcare, implementing a human-AI collaboration system, now explicitly recognises medical staff who identify when automation should be overridden.

3. Recognise Integration as Expertise The ability to integrate AI outputs with human judgment becomes a core skill. PIMCO's ChatGWM platform aims to measure not just client outcomes but how effectively advisors integrate AI-generated insights with relationship management.

4. Value Network Effects Over Individual Output For me, as CTO of Eclipse AI, I watch as the markets create teams that are leaderless—a combination between people, frameworks, and real-time data. Performance becomes about contribution to collective intelligence, not individual productivity. As we move towards working closely with AI, agent performance will be evaluated on their output but for humans, we will need to be measured by our value network effects.

The Competitive Advantage of Seeing the Invisible

Adopting these new frameworks will bring significant advantages.

In Microsoft's research, 71% of employees at "Frontier Firms"—those leading in AI integration—report their companies are thriving, compared to 37% globally. What if they are measuring real value. Holistic value of the human too.

They're also more likely to take on additional work (55% vs 20%) and find meaningful opportunities (90% vs 73%).

This isn't about abandoning accountability. It's about finally measuring what actually drives success.

As AI increasingly handles the measurable, the competitive edge for humans—and the companies that employ them—lies in recognising the immeasurable. The organisations that master this transition won't just implement AI effectively; they'll fundamentally redefine how human value is recognised, rewarded, and cultivated.

The question isn't whether this system can work today, the question is: who will have the courage to build it?

▶ The AI in HR Podcast: The podcast specifically explores the intersection of AI and human resources

▶ Digital HR Leaders Podcast: offers deep dives into people analytics and AI adoption measurement 

▶ Article by Nathalie Salles-Olivier: "LUMA" (Living Understandings of Mutual Agency) and a "Ledger of Unseen Labor"

WATCH OUR COMMUNITY CONVERSATIONS

AI For LUNCH Webinar

A brilliant discussion about AI and how it is impacting enterprise, including:

• How human-in-the-loop is going to change, if at all, over time?

• Data cleanliness

• How LLM's work

• AI Agents

• and much much more

Follow us for more