Leadership in AI: It's Not Just About Buying Tools, It's Redesigning Work Portfolios
The discussion during HBR Executive Live between Adi Ignatius and Ryan Roslansky, CEO of LinkedIn, sheds light on a pressing issue for companies: the real labor market, backed by quantifiable data. Under Roslansky’s stewardship, LinkedIn grew from $7 billion to $17 billion in annual revenue and surpassed 1 billion members, leveraging investments in AI, smarter hiring tools, skill matching, and video. This data matters more for revealing the underlying dynamics than for corporate pride: when a platform sees the entire market, it detects how work is reconfiguring faster than anyone else.
Roslansky argues that AI is a net positive addition to employment, supported by two figures that contradict the simplistic narrative of mass replacement: 1.3 million new roles related to AI (including data annotators) and over 600,000 new jobs in data centers. Meanwhile, LinkedIn has observed a more than 25% change in skill requirements per role in recent years, projecting a 70% change by 2030.
This set of numbers compels a new understanding of leadership. AI isn’t “coming”; it’s already redistributing budgets, talent, and internal power. The operational query for executives shifts from what tool to purchase to how to redesign the work portfolio without underfunding the revenue engine.
The Real Story Behind "Getting Ahead": Work is Fragmenting into Tasks
The most valuable takeaway from the conversation isn’t an aspirational recommendation list. It’s the framing: work is no longer a “position” but a collection of tasks with varying degrees of automation. In practice, this shifts the management focus. If a role involves repetitive tasks, part of that time can be liberated. If the market simultaneously demands new capabilities, that time is either reassigned or lost. The difference between productivity and chaos lies in organizational design.
LinkedIn’s data indicates both job creation and reassignment. On one hand, new jobs emerge around AI: not only data scientists but also annotators, implementation profiles, and, materially, physical work in infrastructure with data centers. On the other hand, while there’s pressure on entry-level positions, Roslansky attributes this decline to macroeconomic factors like interest rates, not to AI. This matters too: if the diagnosis is wrong, the talent plan becomes mere propaganda.
For a leader, “getting ahead” in this context means mastering three levers, not one. First, decompose critical functions into tasks and map what is being automated today. Second, reassemble the role with higher value tasks that indeed require human judgment. Third, recompose the incentive system so that learning becomes part of the job rather than a marginal activity.
Here, bureaucracy often wins by inertia. Many organizations respond by creating AI committees, endless policies, and dashboards that measure adoption as if it were capital expenditure. But real adoption is reflected in flows: saved time, reduced delivery cycles, enabled capacity for sales, operations, or product. If leadership does not make work moves explicitly, AI becomes just another layer of complexity.
The LinkedIn Case: Monetizing AI Isn't Magic, It's a Capital Allocation System
LinkedIn’s growth under Roslansky suggests a more interesting pattern than just “they used AI”. The company turned data and product into an asset that enhances labor demand and supply matching. Improved matching—if better—adds perceived value, retention, and willingness to pay for recruitment solutions and associated services. Here, AI isn’t an isolated project: it’s part of the engine.
In terms of portfolio, I separate it into four fronts that every company must govern, even if not named as such: (1) the current revenue engine; (2) operational efficiency; (3) idea incubation; (4) transformation to scale the new. LinkedIn seems to have executed across all four: they improved the engine with better recommendations and matching, used AI to make hiring more “intelligent”, pushed new formats like video, and—most importantly—reinforced a market thesis: the static resume is worth less than the dynamic evidence of skills.
This last point carries a strategic subtext. If skill requirements have already changed more than 25% and are projected to hit 70% by 2030, the advantage will not come from “having talent” but rather from recycling it quickly. A platform like LinkedIn benefits from this friction: when the market shifts, everyone updates profiles, seeks signals, validates capabilities, and hires faster. For a traditional enterprise, this same friction translates into costs: turnover, prolonged vacancies, hiring mistakes, and diminished productivity.
The executive takeaway is uncomfortable: it’s not sufficient to train people in AI tools. We must redefine how crucial skills are determined, who certifies them internally, and how leaders who restructure roles are rewarded without losing performance.
The C-Level Blind Spot: Measuring Reinvention with Mature Business KPIs
The greatest risk I see isn’t technological; it’s governance. Most corporations attempt to introduce AI into structures designed for stability, not for learning. Immediate returns are demanded from initiatives that, by definition, start with uncertainty. Teams are pressured to promise savings before understanding the process. Decisions are centralized “to control risks,” which stifles velocity.
LinkedIn’s data on the accelerated skill shift makes the annual talent planning based on positions obsolete. If the content of work changes, the control model must shift. This requires separating two rhythms within the same company.
In the revenue engine, discipline: protect margins, ensure quality, avoid service degradation. In exploration, different rules: learning objectives, short cycles, and real autonomy to redesign processes without seeking permission from five committees. When these worlds are mixed, the usual occurs: the company declares a “transformation” but ends up doing incremental optimization.
Roslansky also advocates a skills-based hiring logic. Beyond cultural discussions, operationally, this redesigns filters. If the market stops rewarding linear paths and “career paths,” as he claims, a company that continues recruiting with rigid requirements imposes talent scarcity on itself. Moreover, it loses profile diversity for purely mechanical reasons: it confuses signals with credentials.
In leadership, this translates into concrete decisions. Budget: how much capital and time are allocated to redesigning key tasks. Incentives: what goals are set for functional leaders to deliver results while rebuilding capabilities. Rhythm: how often critical skills are reviewed and people reassigned. If none of that changes, AI enters through software licenses and exits through operational frustration.
The Sustainable Advantage: Protecting the Core While Building Change Capacity
The HBR conversation does not announce alliances or a product plan; it acts as a market signal. LinkedIn positions itself as a labor market thermometer and, with that, pushes an agenda: AI literacy alongside irreplaceable human skills. Roslansky mentions a set of “five human skills” that AI cannot replace, though they are not outlined in the available excerpts. Even without the list, the crucial point for a leader is that these human capabilities are not “declared”: they are designed in practice.
If an organization automates tasks and does not reconsider where human judgment applies, people do not develop discernment; they disengage. If a team incorporates AI and does not alter the decision flow, output accelerates but accountability becomes diffuse. If creativity and collaboration are requested but only execution against budget is measured, compliance rather than adaptation is achieved.
The job creation in AI and data centers further cement an economic reality: spending is shifting towards infrastructure and deployment. It’s not merely a model change; it alters supply chains, energy utilization, operations, and maintenance. For non-tech companies, this translates into reliance on vendors and pressure to develop implementation and operational profiles, not just “strategy.”
Effective leadership in AI appears when the portfolio is explicit: the core is defended with efficiency and commercial focus, while a section of the organization operates with enough autonomy to redesign work, validate new practices, and scale what works without being trapped by mature business KPIs.
Closing: The New Standard is Governing Two Speeds Without Destroying Either
The data presented by LinkedIn describes a market where jobs are created, but skills become perishable at a rate that most companies are unprepared to navigate. Leadership that “gets ahead” is the one that transforms roles into tasks, reassesses talent with discipline, and funds exploration with learning metrics, while protecting the revenue engine with controls suitable for mature businesses. Viability depends on maintaining current profitability without stifling the ability to rebuild skills before 2030.
