The middle tier of the knowledge workforce is the layer between frontline execution and senior leadership. It’s the part of the org chart most exposed to AI. PwC is calling the result an hourglass, and most organizations still aren’t measuring it.
PwC’s hourglass framing describes what can happen when AI agents take on work that midlevel knowledge workers have traditionally owned: coordination, analysis, summarization, reporting, routing, and the many decisions that sit between senior strategy and frontline execution. As that layer becomes more automated, organizations face a structural question a tool deployment plan can’t answer: what do midlevel employees do now, and how does the company create a path from junior to senior roles when the midlevel training ground changes?
BCG’s analysis puts numbers on the scale of this transition. It found that 43% of U.S. jobs exceed the 40% automation threshold where role and organizational redesign becomes a stronger business case. That doesn’t mean those jobs disappear. It means the work inside them changes enough that the org chart, career path, and skills strategy need to change with it.
That connects workforce structure directly to the AI skills gap: leaders need to know which people, roles, and tasks are changing, not just which tools are available.
Scheduling, routing, status reporting, cross-functional coordination, and data summarization are all categories of midlevel work where autonomous agents are already being deployed. Gartner’s projection that 40% of enterprise applications will include task-specific agents by the end of 2026 reflects the same shift. The governance infrastructure required to manage those agents is covered in our post on AI agent governance in the enterprise.
As midlevel work becomes agentic, the performance premium grows for employees who can direct and orchestrate AI effectively. They take on higher-complexity work while agents handle the routine layer. Employees who don’t develop AI proficiency may find fewer career development opportunities as the work that used to build those skills shifts to agents.
When the midlevel work that traditionally trained junior employees for senior roles is handled by agents, the organization needs new mechanisms for developing that capability. This is what makes the hourglass model disruptive rather than merely structurally interesting: the pipeline that produces senior talent depends on midlevel experience, and that experience is changing.
Changing the work inside a role changes more than the org chart. Employees have to understand where their judgment remains valuable, how they will develop new capability, and what career progression looks like when agents take on familiar coordination and analysis tasks. A redesign plan that addresses only task allocation leaves those questions unanswered.
The July 13, 2026 workforce-readiness analysis describes a trust gap: employees may use AI without trusting its outputs, the adoption process, or their own ability to use it well. That distinction matters as the midlevel training ground changes. Compliance with a tool mandate does not establish readiness for a redesigned role.
Fear of becoming obsolete, or FOBO, concerns skills losing value faster than people can adapt. An employee may not expect replacement tomorrow but still worry that their experience and judgment are becoming less relevant. As career paths compress, leaders should explain which capabilities remain important and how employees will have opportunities to build them.
The linked July 13 analysis cites EY research reporting concern about AI among 71% of employees, and Pew research reporting 52% worried about AI’s future workplace impact and 33% feeling overwhelmed. It also summarizes ManpowerGroup’s 2026 Global Talent Barometer as reporting regular AI use rising 13 percentage points to 45% while confidence in using technology fell 18% over the previous year. These are separate source-reported findings from different research populations, not a combined estimate for your workforce or proof that automation caused declining confidence.
The workforce-readiness analysis also cites research in Humanities and Social Sciences Communications finding an association between organizational AI adoption and employee depression, with psychological safety playing a mediating role. This is not proof that AI causes depression or a diagnosis of employees. It is a reason to assess the support and learning conditions around change rather than assuming higher usage means a healthy transition.
Psychological safety means people can ask questions, acknowledge knowledge gaps, and learn without fear of punishment. When roles are changing, employees need room to challenge an AI result, report a failure, or request help. Without that room, utilization may rise while verification, confidence, and effective judgment remain weak.
Treat declining confidence, disengagement, policy workarounds, and avoidable rework as risks to investigate alongside structural change, not inevitable consequences of AI. Task telemetry cannot tell leaders how employees feel; readiness assessment requires employee feedback and appropriate safeguards.
Workforce planning needs three connected views: how tasks are changing, how capability is developing, and whether employees have the confidence and support to adapt.
Which tasks in each role are being done by AI, which are still being done by humans, and how is that ratio changing quarter over quarter? Our Workflow Intelligence platform captures this at the workflow level continuously, so leaders can see where roles are actually changing instead of relying on assumptions from job descriptions.
Are junior employees developing AI fluency fast enough to reach senior-level capability as the midlevel training path compresses? AI Fluency measurement gives leaders that data at the individual, team, and function level, making it possible to see where capability investment needs to accelerate.
Use appropriately governed employee feedback to understand trust in AI outputs, confidence in verification, clarity about changing roles, access to relevant training, and comfort asking for help. Review trends by role, team, and level where the group size supports responsible reporting. Do not infer mental health or individual anxiety from usage patterns.
Read those signals alongside proficiency and workflow outcomes. High utilization with weak proficiency suggests a need for support; low confidence may reflect genuine quality problems rather than resistance. Investigate the cause before changing performance expectations or restricting career opportunities.
WEF’s 170 million new roles projection can make AI sound like a simple job-creation story. It isn’t. The same report projects 92 million displaced jobs and says nearly 40% of skills required on the job will change by 2030. That’s a workforce redesign problem, not a hiring forecast.
PwC and BCG point to the same practical requirement: companies need to redesign work, career paths, and capability building as agents spread. Layering AI onto old structures may create productivity bumps, but it won’t show leaders whether the midlevel path is still developing future senior talent.
The structural decision and the readiness decision belong in the same workforce plan. An organization can automate coordination without building the human capability needed to govern exceptions or develop future senior talent. That is not a completed redesign.
The hourglass workforce is a structural shift in which human talent becomes more concentrated at junior execution and senior strategy levels while the middle tier shrinks. PwC ties it to agents taking on more midlevel coordination, analysis, and reporting work. Midlevel employees do not disappear overnight; the work that develops them starts to move.
Track task distribution at the role level: which tasks are shifting to AI, which remain human, and how that ratio changes over time. Organizations that can’t answer this at the team level don’t yet have the measurement infrastructure to see the structural shift as it happens.
Build AI fluency measurement into workforce planning. Identify which roles are most exposed to task automation. Create accelerated development paths for junior employees who need to build senior-level capability as the midlevel training ground changes. Our CHRO guide to AI monitoring covers this strategic planning dimension in more detail.
Not directly and not immediately. But it’s automating a significant portion of the work middle management has traditionally owned. The question is whether organizations redesign those roles proactively, creating new forms of human value at the midlevel, or react after the structural shift is already visible in performance and retention data.
Employees may worry that their skills or career path are losing value as tasks move to agents. That can coexist with high tool usage. Ask about confidence, role clarity, and learning support rather than treating utilization as proof that people are ready.
No. Proficiency and usage can identify capability or adoption patterns, but they do not establish how employees feel. Use appropriately governed feedback to assess learning conditions, and do not diagnose anxiety or mental health from tool telemetry.
Explain how AI changes responsibilities and progression, provide continuous task-specific training, and create safe ways to ask questions and challenge outputs. Pair that support with evidence about proficiency, quality, and whether new development paths prepare people for higher-complexity work.
Larridin’s AI Fluency and Workflow Intelligence capabilities identify where AI is already changing task distribution, which roles are most affected, and where capability investment creates the most strategic leverage.
Book a discovery call to see your workforce structure data.