The same data, two inverse causes

26 min read

Based on «Global Workforce Hopes and Fears Survey 2026 | PwC» · October 2026

There is a sentence in PwC's Global Workforce Hopes and Fears Survey 2026 that no one is quoting, and it is the harshest of all: «twice as many workers think their bargaining power has fallen over the last three years than think it has risen» [1]. It is worth treating it as a market fact rather than a state of mind. Bargaining power is not a mood; it is a relative position in a relationship with two parties. If twice as many workers believe they have lost it compared to those who believe they have gained it, the question the survey does not ask is elementary: with whom, exactly, is one negotiating today. The sample warrants the question — 49,364 responses, 48 countries and regions, 29 sectors, collected in May and June 2026 [1]. It is not an opinion poll; it is a large-scale photograph of a relationship that has become unbalanced.

The answer begins to appear in another document, written for another reader and with another intent. The 2026 High End Independent Talent Report, by Heidrick & Struggles, authored by Sunny Ackerman, describes the upper side of that same relationship: «In a world where disruption is the baseline, independent talent are moving to the center of how critical work gets done» [3]. Requests for CEOs and chairs rose 38% year over year; interim C-suite leadership grew 151% since 2021 and a further 14% in the last year; within that category, interim COOs increased 250%, CMOs 100%, CEOs 24%, CFOs 14% — and interim CFOs now account for 51% of all interim leadership requests [3]. It is worth saying clearly what this is and what it is not: an internal series of orders passing through a single executive search firm, solid as a trend, fragile as an estimate of the global market. Even so, the direction is unambiguous. The employer side has also become mobile.

PwC reads the workforce divergence through a technological dividing line. The «front-runners» are 14% of the total: 51% use generative AI daily, 75% trust top management, and 29% say it is very or extremely likely they will change employers within the next twelve months [1]. The «engine room» is 56%: 11% daily use, 33% trust in the top, half with no confidence in their own job security, the majority having never used these tools [1]. The diagnosis follows naturally: those with access see opportunity, those without see threat; therefore, work needs to be redesigned around the technology instead of injecting it into old operating models. Let us accept the diagnosis provisionally — and then submit it to cross-examination.

What the anchor report measures, it measures well. Adoption rose: 64% of workers say they have used AI at work in the last twelve months, ten percentage points above the previous year, and daily generative AI users went from 14% to 22% [1]. What did not keep pace was everything else. Trust in top management fell seven percentage points year on year, with less than half of the global workforce now trusting its leaders [1]. And only 34% can pay their bills and still have money left over at the end of the month — eight points fewer than the previous year [1]. More technology, less trust, less material slack: this is the triad PwC interprets as a symptom of insufficient redesign. The same firm points, moreover, to its Global CEO Survey, where most CEOs still admit they see no financial return from AI investment [4] — and attributes the deficit to the operating model.

The January–February 2026 issue of the Harvard Business Review opens with the exact opposite thesis. Darrell Rigby and Zach First, in «Get Off the Transformation Treadmill,» argue that «Too much change can traumatize your organization. The remedy is to minimize the need.» [2]. Where PwC sees a transformation deficit, HBR sees transformation saturation. And the materials in the same issue support that reading: according to HBR, citing a Writer survey, 31% of workers admit to actively resisting their companies' AI initiatives, often out of fear of being replaced, and one in ten says they have falsified performance metrics; 62% of companies point to poor cross-functional fit and 63% to the need to readjust workflows as the main barriers; and, in another survey cited by the magazine, only 25% of CEOs say they are fully prepared to deploy AI at organizational scale [2]. The same observable phenomenon — initiatives that don't pay off — receives opposite causes depending on who is observing it.

Notice what the two voices do together, despite diverging in their prescription: both remove the bottleneck from technology. PwC locates it in the operating model [1]; HBR states that many initiatives fail not because the algorithms are weak but because people refuse to use them [2]. This establishes, without appeal, that the problem is one of governance and not of technical capacity. And this is where the reading flips. If 31% actively resist and less than half the workforce trusts those who decide [1][2], resistance stops being friction to be removed through training and communication, and becomes information about who is in charge. It is not a system malfunction. It is a vote.

Transformation has stopped being a project and become a market

Heidrick enters this dispute not as opinion but as an order count — and that is what gives it arbitral authority. Demand for human capital specialists rose 129% year over year; demand for digital, data, and IT leaders grew 58% since 2023; healthcare and life sciences, consumer markets, and technology and services lead demand for independent talent [3]. Read alongside PwC, the human capital figure is almost ironic. The anchor report asks organizations to protect learning time rather than piling it on top of daily tasks, and to give managers license to redesign work [1] — that is, it asks for permanent, internal capacity to design work. Heidrick shows that this capacity is, in large part, being rented out [3]. The function to which PwC entrusts the remedy is one of the functions migrating fastest into rental mode.

There is a moral asymmetry in this material that deserves to be named. PwC treats the 29% of front-runners ready to leave as flight risk, a symptom of organizations that reward caution, a retention problem to be solved with autonomy and progression routes [1]. Heidrick describes exactly the same detachment, one layer up, as a deliberate and desirable strategy [3]. Not staying is pathology below and virtue above. And neither side names the other: PwC measures workers' distrust of top leaders without counting how much of that top is temporary; Heidrick celebrates temporariness without cross-referencing it against the downstream cost of trust it may generate.

It is an old mistake, and it has a name in contract theory. Ian Macneil distinguished the transactional contract — discrete, complete, settled on the spot — from the relational contract, whose value depends on the expectation of continuity rather than on what is written. Albert Hirschman, in Exit, Voice, and Loyalty, showed that exit and voice are substitutes: whoever can leave stops investing in improving what they leave behind. What these three sources describe together is an organization in which exit has become the normal behavior of the top and of the best-equipped 14% [1][3], while voice is exercised by those who stay — and exercised, according to HBR, in the least legible form of all: silent resistance and falsified metrics [2].

From here follows an observation that none of the three sources formulates, and that only juxtaposition makes visible: trust is a slow-forming stock, whose accumulation rate depends on the mandate of decision-makers being longer than the change cycle they impose. If that is so, the seven points lost in PwC's figures [1] may be, in part, a lagging indicator of leadership rotation rather than an effect of artificial intelligence. The hypothesis is falsifiable, and should be treated as such — none of the three sources directly links top rotation to falling trust, and anyone who disagrees has a simple test available: show organizations with high C-suite rotation and rising internal trust. Until that case appears, the reading stands. And it extends what this publication has already argued in «The asset that became a flow»: renting top cognition consumes a stock that the very logic of renting does not replenish. On the demand side, that stock was called discernment. On the side of the entire workforce, it is called trust — and the bill arrived in June 2026, discounted by seven points.

The capacity being rented is the capacity being prescribed

There is a third collision, quieter than the previous two, and it is the one that most clearly exposes the problem. PwC, in the Global Workforce Hopes and Fears Survey 2026 [1], ends where all surveys of this kind end: with a set of recommendations for employers. Two of them are precise and deserve to be taken seriously — protecting learning time rather than adding it to the daily load, and giving middle managers effective license to redesign their teams' work. Both presuppose the same thing: permanent, accumulated internal capacity to design work. Yet the 2026 High End Independent Talent Report, by Sunny Ackerman of Heidrick & Struggles [3], shows that this is exactly the capacity leaving the building faster than any other: demand for human capital specialists rose 129% year over year, and demand for digital, data, and IT leaders grew 58% since 2023. It is worth qualifying the figure — it concerns demand passing through the firm itself, an internal and interested series, solid as a trend in orders and fragile as a market estimate. But the direction is unequivocal: the function to which PwC entrusts the remedy is one of those migrating fastest into rental mode.

The compounded effect is unpleasant to state. Whoever designs the work is hired on a project basis; whoever executes the design stays; whoever evaluates the result is already gone. The January–February issue of the Harvard Business Review [2] offers, without intending to, the doctrinal complement to this architecture: Antonio Nieto-Rodriguez argues, in «The Project-Driven Organization,» that CEOs should place projects, rather than operations, at the center of value creation. The thesis is defensible in terms of capital allocation and indefensible in terms of trust. A project is, by definition, a form of work with an end date; operations are what remains once the project ends. An organization that shifts its center of gravity toward projects simultaneously shifts authority toward those with a short horizon — and leaves legitimacy on the side of those with a long horizon and no authority. The same issue measures the price of this in an apparently minor note in the «Idea Watch» section: the cost of AI workslop, machine-generated work that someone, downstream, has to clean up.

The bottleneck is not technical

At the two points where the anchor and HBR converge, that convergence is more informative than any of the divergences. PwC refers to its own Global CEO Survey to recall that most CEOs still see no financial return from investment in artificial intelligence [1], and attributes the deficit to the operating model. HBR, citing a Writer survey and another survey of CEOs [2], arrives at the same place through another door: 31% of workers admit to actively resisting the company's AI initiatives, often out of fear of being replaced, and one in ten says they have falsified performance metrics; 62% of companies point to poor cross-functional fit and 63% to the need to adjust workflows as the main barriers; only 25% of CEOs say they are fully prepared to deploy AI at organizational scale. None of these numbers are about algorithms. All of them are about governance. The bottleneck sits upstream of the model and downstream of the executive committee, in that in-between space where it is decided who is in charge of what.

From here comes the twist that neither of the two formulates. If 31% actively resist and less than half of the global workforce trusts top leaders — seven points below the previous year [1] —, resistance stops being friction to be removed and becomes information about who is in charge. One in ten falsifying metrics is not a digital literacy failure; it is a vote, cast at the only ballot box available to those with no formal voice and no credible exit. It should be noted that these are perceptions: nothing in PwC's survey proves that front-runners produce more, only that they feel more valuable and more secure. But perception is precisely the substance of trust, and trust is the substance of execution. An organization that treats 31% informed refusal as a communication problem is asking the wrong department to resolve a dispute over authority.

Who stays

Crossing PwC with Heidrick produces, at different layers, the same description: the long employment relationship has stopped being the vehicle of competence. Competence circulates; the organization rents. Ackerman says it bluntly — «In a world where disruption is the baseline, independent talent are moving to the center of how critical work gets done» [3]. The question neither of them asks is who is left over. If 14% of the workforce are front-runners, with 51% daily GenAI use and 75% trust in the top, and 29% of those consider it very or extremely likely that they will change employers in the next twelve months [1]; and if a good part of the C-suite is now interim — +151% since 2021, +14% in the last year, with COOs at +250% and interim CFOs accounting for 51% of all interim leadership requests [3] — then those who remain are precisely the engine room: 56% of the workforce, 11% daily GenAI use, the majority having never used AI, 33% trust in top leaders, half with no confidence in their job security [1].

Staying has stopped being rewarded loyalty and has started to function as an indicator of exclusion. This is the inversion that the crossing makes impossible to ignore, and the most uncomfortable one for anyone sitting on a board. Two-thirds of that engine room still say they go beyond what is required [1], at a time when only 34% of workers can pay their bills and still have money left at the end of the month — eight points below the previous year. Delivery holds because human energy compensates for architectural failures, not because the architecture works. This publication has already argued, in «Belonging does not scale without degrading,» that belonging withstands expansion poorly; what these three reports together show is the financial variant of the same phenomenon — belonging also does not withstand the rotation of those who administer it.

The moral asymmetry is the point where the dispute becomes indefensible. The same behavior — not staying — is read as strategic virtue at the top and as flight risk at the bottom. Heidrick records requests for CEOs and chairs rising 38% year over year [3] and treats temporariness as a mature response to unpredictability; PwC treats the 29% of front-runners eyeing the door as a symptom of organizations that reward caution [1]. The two groups share the same incentive structure and the same reading of the market. Only the classification differs. This is where PwC's most quotable sentence reveals its other side: artificial intelligence is «professionalising» some jobs, reshaping them to require even more human skill [1] — but it professionalizes, above all, those who already could leave. For the others, it does not upskill: it exposes.

And this is where PwC's other sentence — «twice as many workers think their bargaining power has fallen over the last three years than think it has risen» [1] — should be read as a market fact and not a state of mind. Bargaining power did not fall just because technology replaces tasks. It fell because the employer side has become equally mobile. There is no stable counterparty to negotiate with: the director who promised the progression route leaves by month fourteen; the interim COO who redesigned the workflow delivers the report and vanishes; the human capital specialist who designed the competency model invoiced and moved on. Negotiating requires a counterparty with memory and exposure to breach of promise. The erosion of that counterparty is the part of the story that appears in none of the three documents, because none of the three measures the horizon of whoever signs.

The practical consequence is narrow and verifiable, and it is the only recommendation these three documents, read together, authorize. The missing metric is not the AI adoption rate — that has already risen ten points, to 64%, with daily GenAI users going from 14% to 22% [1], and has brought no visible return. The missing metric is the average tenure of whoever signs off on work redesign, measured from the date of signature. A board that wants to know whether trust will recover does not need another climate survey: it needs to cross-reference the duration of its transformation sponsors' mandates with the lifecycle of the transformations they sponsor. Where the second number exceeds the first, trust will keep falling — and will keep being attributed to technology.

[BOX] RESIDUAL MANDATE: THE MISSING VARIABLE IN THE THREE REPORTS

Residual mandate is the time remaining, as of the decision date, in the predictable mandate of whoever makes it. It is neither seniority nor contractual duration: it is the effective horizon of exposure to consequences. An interim CFO brought in to close out a restructuring cycle has a residual mandate of months; a managing director with a two-year contract and a three-year transformation plan has a negative residual mandate relative to what they are signing. The variable is trivial to calculate and none of the three documents records it.

Its relevance to the board stems from a simple asymmetry: the cost of a transformation is front-loaded, the benefit is diluted at the end, and trust behaves inversely — it is paid upfront and accumulates slowly. With +151% interim C-suite leadership since 2021 [3] and trust in top leaders falling seven points in a single year [1], the hypothesis that the two series are related deserves testing before it deserves dismissal. None of the sources establishes it; the juxtaposition merely makes it plausible and falsifiable.

The concept allows a more precise re-reading of HBR's cover thesis — «Too much change can traumatize your organization. The remedy is to minimize the need,» by Darrell Rigby and Zach First [2] — than the authors themselves claim. Trauma does not come from the amount of change in the abstract: it comes from how often the author of the change changes. Ten years of continuous adjustment under the same mandate produces learning; three reorganizations in three years under three different sponsors produce scarring. The same volume of change, opposite results.

For the front-runner, a short residual mandate is an opportunity: 29% already consider it very likely they will change employers within twelve months [1], and every transformation abandoned midway is a résumé line with the outcome yet to be verified. For the engine room, which represents 56% of the workforce and votes with 33% trust in the top [1], it is the opposite: it is the certainty that the next promise will have a different author. The difference between the two groups lies not only in access to technology — it lies in each one's horizon relative to the company's horizon.

The application is immediate and requires no new system. Before approving a transformation program, the board should demand a single line: what is the sponsor's residual mandate, and how does it compare with the program's cycle. Where the gap is unfavorable, there are three honest options — shorten the program, extend the mandate, or appoint a permanent guarantor who answers for the outcome after the author has left. The fourth option, the one actually practiced, consists of approving it anyway and measuring trust the following year.

What changes at the boardroom table

The first practical consequence of this crossing is one of human capital governance, not of technology: the missing indicator on board dashboards is the residual mandate of transformation sponsors, measured in months and compared with the duration of the programs they sign off on. PwC surveyed 49,364 workers across 48 countries and 29 sectors between May and June 2026 and measures the outcome with precision — trust in top management falling seven points in a year, less than half the workforce trusting top leaders [1] — without measuring the variable that Heidrick & Struggles tallies on the other side of the counter: interim C-suite leadership requests +151% since 2021 and +14% in the last year, with COOs +250%, CMOs +100%, CEOs +24%, and CFOs +14% [3]. No board today has a line cross-referencing the two series within its own house, and it is trivial to build one. One need only record, for every program approved above a given investment threshold, the expected completion date and the likely departure date of whoever proposed it.

The second implication concerns investment sequencing. PwC records 64% of workers using AI in the past year, up ten points year on year, and daily GenAI users rising from 14% to 22% [1], while pointing to its CEO survey's finding that most still see no financial return from the investment. HBR's January-February issue, citing a Writer survey, points to the cause that no technology budget solves: 31% of workers admit to actively resisting the company's AI initiatives, often out of fear of being replaced, and about one in ten says they have falsified performance metrics [2]. A board reading these two findings together concludes that the next tranche of investment carries diminishing marginal returns until trust stabilizes — and that the trust variable is cheaper to fix than the infrastructure one, because it depends on appointments, not on licenses.

The third implication touches the architecture of the human resources function and should unsettle whoever oversees it. PwC prescribes protecting learning time instead of adding it to daily tasks and giving line managers effective license to redesign work [1]; yet Heidrick records that demand for human capital specialists rose 129% year over year and that demand for digital, data, and IT leaders grew 58% since 2023 [3]. The capacity to which the anchor report entrusts the remedy is precisely migrating into rental mode — and Heidrick's figures, it should be said, measure orders passing through the firm itself, which makes them solid as a trend and fragile as a market estimate. Even so, the direction is unequivocal: the competence to design from within is being outsourced.

The fourth implication concerns differential retention and an honest reading of risk. Among the 14% of front-runners — 51% daily GenAI use, 75% trust in the top —, 29% consider it very or extremely likely they will change employers within twelve months [1]; among the 56% of the engine room, with 11% daily use and 33% trust, mobility is residual. A board that reads this as a digital talent retention problem spends badly: what the number describes is not dissatisfaction, it is optionality. Those with alternatives use them; those without stay and sustain delivery. PwC measures perceptions, not productivity — nothing in the survey proves that front-runners produce more, only that they feel more valuable and more secure [1] — and treating perception as performance is the quickest way to build a retention policy on sand.

The fifth implication is financial and direct. Only 34% of workers can pay their bills and still have money left at the end of the month, eight points below the previous year [1]. Any transformation program designed on the assumption of cognitive availability and risk tolerance on the part of those who execute it is presuming a material condition that no longer holds for two-thirds of the base. Heidrick describes the most intense demand for independent talent in healthcare and life sciences, consumer markets, and technology and services [3] — sectors where the operational layer simultaneously has the least financial cushion and absorbs the most change designed by those who pass through.

The mobility paradox

The central tension is one of moral asymmetry, and it is visible as soon as the two sources are overlaid. Heidrick asserts, through Sunny Ackerman's voice, that «In a world where disruption is the baseline, independent talent are moving to the center of how critical work gets done» [3] — detachment as market maturity, with CEO and chair requests rising 38% year over year and interim CFOs accounting for 51% of all interim leadership requests. PwC describes exactly the same behavior, one layer down, as risk: the 29% of front-runners about to leave are a symptom of organizations that reward caution [1]. The same act — not staying — is strategy at the top and desertion at the base.

The paradox deepens when the anchor's most quotable sentence is read in reverse. PwC observes that AI is «professionalising» some jobs, reshaping them to require even more human competence [1]. The statement is true and insufficient: it professionalizes, above all, those who already had market value. For the 11% of daily users in the engine room, the same technology adds no remunerable demand — it adds surveillance. And PwC's other sentence, according to which «twice as many workers think their bargaining power has fallen over the last three years than think it has risen» [1], describes less a loss of power than the disappearance of the counterparty: there is no one to negotiate with when the deciding side is also temporary.

There is also the prescription paradox. Rigby and First argue, on HBR's cover, that excessive change traumatizes and that the remedy lies in minimizing the need for transformation [2]; PwC argues that the remedy lies in transforming more and better, redesigning work instead of injecting AI into old models [1]. Both are right about different halves of the problem, and Heidrick explains why they coexist: permanent transformation has stopped being a project and become a market, with suppliers, prices, and order series. Those who make a living designing change have no incentive to minimize the need for it — and HBR, in the same issue, measures the price of this when it addresses the cost of «AI workslop» and when Antonio Nieto-Rodriguez proposes placing projects, rather than operations, at the center of value creation [2]. It is a coherent and dangerous proposal: organizations made entirely of projects are organizations where no one signs off on the final result.

Scenarios

The first scenario is the institutionalization of renting. If Heidrick's series continues at the observed order of magnitude — +151% since 2021, +14% in the last year [3] —, a stable part of the C-suite will be hired by mission, with twelve-to-eighteen-month mandates, and the trust measured by PwC stabilizes at a structurally low level, around or below current levels [1]. There is no collapse: there is a new equilibrium, in which the company buys execution and stops producing belonging. The cost does not appear on the income statement; it appears in how fast any subsequent initiative needs to be imposed rather than adopted.

The second scenario is the deliberate reconstruction of the horizon. Some organizations — probably family-owned ones, listed companies with a reference shareholder, and those operating in long industrial cycles — start treating permanence as an explicit competitive advantage, lengthening mandates, tying variable pay to verifiable results three years after a program's completion, and appointing internal guarantors who answer for what the interim designed. HBR provides them with the structural argument in «Leading After the Founder,» by Samantha Hellauer and colleagues, which addresses the emotional and organizational costs of a founder's departure [2]. In markets where competitors rotate the top every eighteen months, the company that keeps the same decision-maker for five years acquires an advantage that cannot be bought: it is trusted.

The third scenario is contractual bifurcation. The 14% of front-runners converge, in terms and horizon, with the high-level independent talent Heidrick describes, and the distinction between employee and provider loses economic meaning at the top of the competence distribution. The 56% of the engine room consolidate as the only truly permanent population — and permanence stops being rewarded loyalty to become an indicator of exclusion. When stability signals a lack of alternatives, no culture program corrects the signal.

The fourth scenario is the least discussed and the most likely in the short term: none of this is decided, it simply happens. With 25% of CEOs declaring themselves fully prepared to deploy AI at organizational scale, according to the survey cited by HBR, and with 62% pointing to poor cross-functional fit and 63% to the need to adjust workflows as the main barriers [2], inertia will take the form of approved programs, sponsors who change midway, and trust indicators measured the following year with genuine surprise.

The arbitration, once done, resolves as follows: HBR is right about the symptom, PwC is right about the need, and Heidrick explains why the two coexist without cancelling each other out. Trust does not fall for lack of work redesign — it falls from an excess of transient redesigners. The hypothesis is falsifiable and should be treated as such: whoever disagrees must produce organizations with high top rotation and rising internal trust. This publication's archive has already described, regarding the renting of top cognition, a stock of discernment that the logic of renting consumes without replenishing; the seven points PwC lost are the bill for that operation presented to the layer below, where the stock in question is no longer discernment, but trust.

What remains is the narrow, verifiable consequence. The indicator to follow in the next cycle is not the AI adoption rate, which has already risen ten points and will keep rising without producing proportional returns [1]: it is the tenure of whoever signs off on the redesign. Until that number is known, any discussion of culture, resistance, or digital literacy is discussing effects. The engine room understood this before the boards did: when 33% trust the top and 56% keep delivering, what is being measured is no longer morale — it is patience. And the signature that matters in a transformation program was never the one that approves it, but the one still in the building when it ends.