What does it actually take for a strategic intent to survive contact with the business that must implement it?
In this episode of On the Subject of Leadership, I speak with Pascal Uerlings, co-founder and Chief Revenue Officer of J4RVIS, a Sydney-headquartered Salesforce and AI implementation consultancy that has grown from two people to over a hundred across Australia and the Philippines in just under six years. Pascal has spent that period working at the unresolved seam where technology strategy meets organisational reality—the seam at which most transformations quietly, and expensively, fail.
This is not a conversation about Salesforce, or about agentic AI, or about the mechanics of consulting growth—though we do cover all of that too. It is a conversation about what happens to ambitious strategy when it meets an operating model that was never designed to absorb it, and about the disciplines—cultural, structural, and personal—that determine whether the resulting friction produces motion or merely heat.
Pascal's argument, developed over more than a hundred implementations, is unromantic and worth taking seriously: the problem is rarely technical. It is structural, cultural, and in no small part a question of whether anyone in the room is willing to say what is actually happening.
Values Before Customers
The conventional founder narrative begins with a problem and a product, with culture treated as something that must be retrofitted later, once growth has produced enough collisions to make the absence of shared norms intolerable. Pascal and his co-founder, Jannis Kearney Bott, inverted this sequence. Before they had taken on a single client, before the legal entity was formed, they had a conversation in the foyer of the Telstra building in Sydney about what values the as-yet-nonexistent organisation would instantiate.
This is, on its face, the kind of detail that founder mythology tends to manufacture after the fact. The interesting question is whether such early articulation makes any practical difference. The empirical evidence suggests it does. The founder's role in early cultural formation has been identified as one of the most consequential acts of leadership, precisely because what founders pay attention to, measure, and reward in the formative period becomes encoded as the organisation's basic underlying assumptions—the deepest layer of culture, which subsequently operates outside conscious awareness and becomes extraordinarily difficult to alter once embedded.
Pascal frames this in operational terms. The values, he argues, function as guardrails against decision fatigue. Founders make hundreds of consequential decisions in conditions of incomplete information; without an articulated set of priorities to fall back on, each decision must be litigated from first principles. The cost is not merely cognitive—it is the gradual erosion of coherence as different decisions, made under different pressures, produce a pattern of behaviour that no one would have endorsed in advance.
This is a recognisably Schelling-esque insight: rules are valuable not because they are optimal in any individual instance, but because they enable consistent behaviour across many instances at acceptable cognitive cost. The values are, in this sense, a form of pre-commitment—a way of resolving in advance the kinds of conflicts that, faced fresh each time, would be vulnerable to short-term pressure.
The Counter-Argument for Waiting
J4RVIS launched in April 2020. Within a fortnight of Jannis tendering his resignation, COVID-19 had reached Australian shores. By the end of the month, the country was in its first national lockdown.
The conventional reading of this story would emphasise either heroic risk-tolerance or extraordinary luck. Pascal offers a more sober account. The decision to launch was not a moment of courage but the culmination of a long internal deliberation in which the strongest argument for waiting was, in retrospect, indistinguishable from the standard rationalisations of those who never launch at all. Comfort and confidence, he observes, are not signals that conditions are right; they are signals that one is not, in fact, an entrepreneur.
This maps onto a substantial body of research on entrepreneurial decision-making under uncertainty. Work on effectuation describes precisely this pattern: experienced entrepreneurs do not begin with a goal and then assemble means to achieve it; they begin with available means—their identity, their networks, their tolerance for ambiguity—and allow the goal to emerge from what those means make possible. The April 2020 launch was, in this framing, not a bet on timing but a bet on the founders' capacity to absorb whatever timing produced.
There is a corollary here that is worth naming. The asymmetry between launching and waiting is not symmetric in the way founders often imagine. Waiting feels like risk reduction; in fact, it merely substitutes one kind of risk—the risk of acting on incomplete information—for another, the risk of arriving in a market whose conditions have been defined by those who acted earlier. The cost of the second risk is invisible because it consists of opportunities that were never taken. This is the same epistemological trap that organisations fall into when they confuse a pause with stability—a pattern Nick Hassett described in a previous conversation on the show, and one that recurs throughout Pascal's account of the firms he has consulted to.
Imposter Syndrome as Diagnostic Signal
Few topics in contemporary leadership commentary have been more thoroughly worn out than imposter syndrome. The original construct, articulated by Pauline Clance and Suzanne Imes in their study of high-achieving women, has been progressively detached from its empirical moorings and pressed into service as a generic explanation for any form of professional self-doubt. The result is a discourse in which imposter syndrome is treated as both ubiquitous and uniformly pathological—something to be diagnosed, validated, and overcome.
Pascal's reframing is more useful. He suggests that the experience of doubting one's competence in high-stakes situations is not, in itself, a deficit. It is a signal that one cares about the outcome, and that the work has not yet been routinised into something that can be discharged on autopilot. The behavioural correlates of imposter feelings—over-preparation, deeper research, more careful attention to the audience—are precisely the behaviours one would want from a leader operating at the edge of their competence. The pathology, on this account, lies not in the feeling itself but in the response to it: collapse into paralysis, or compensation through performed certainty.
This is consistent with subsequent empirical work. An extensive review of the impostor phenomenon literature, noted that the construct is more accurately understood as a continuum of experience rather than a discrete syndrome, and that moderate impostor feelings are associated with the over-preparation behaviours that frequently produce the success the sufferer then attributes to luck. The danger is not the experience but the loop—the failure to update one's self-model in light of the evidence that the over-preparation has, in fact, been adequate.
For founders specifically, there is an additional dimension. The performance of certainty is, in many contexts, a strategic necessity—investors, employees, and customers all draw inferences about an organisation's prospects from the demeanour of its leadership. The discipline lies in maintaining the internal capacity for doubt while presenting an external face of conviction, and in not confusing the two. Pascal's account suggests this is a learnable discipline, but only for those willing to acknowledge the underlying experience rather than suppress it.
Pilots Without a Path to Production
When the conversation turns to artificial intelligence, Pascal articulates what is, in my view, the single most important observation in the episode. The reason most AI initiatives fail is not technological. It is that they were never designed, from the outset, to reach production.
The pattern is by now well documented in the consulting and research literature. McKinsey's State of AI in 2025 survey reports that 88% of organisations now use AI in at least one business function and 72% report using generative AI specifically—up from 33% only a year earlier—yet only 39% report any EBIT impact attributable to AI, and nearly two-thirds have not begun scaling AI across the enterprise. Boston Consulting Group's study of The Widening AI Value Gap, drawing on responses from 1,250 senior executives, finds that just 5% of companies qualify as "future-built" and consistently generate substantial value from AI, while 60% remain laggards reporting minimal returns. RAND Corporation's qualitative research with experienced AI practitioners identified the highest-impact failure cause as misaligned incentives between technical teams and operational sponsors, not technical limitations of the models themselves.
The pattern is not, in itself, novel. The gap between what organisations know they should do and what they actually do is one of the most persistent pathologies in management practice—a gap attributed not to ignorance but to organisational dynamics that systematically privilege the appearance of action over its substance. AI has not produced this pathology; it has merely supplied it with a more expensive vehicle.
Pascal's diagnosis is consistent with this evidence. The failure mode is not that the pilot underperforms; it is that the pilot succeeds on its own terms—a clean dataset, a controlled environment, a dedicated team—and then proves non-translatable to the production environment, where data is messy, integration is partial, and the change management work has not been resourced. The board sees the demo, approves the next tranche, and is then bewildered six months later when the operational metrics have not moved.
The corrective Pascal proposes is structural rather than technical. Define the production-state success metric before writing any code. Assign a single accountable owner for change management from day one—not as a downstream activity but as a parallel workstream. Commit to measurement and refinement at three, six, nine, and twelve months. And accept that the model itself is no longer a stable artefact; the rate of capability change in foundation models means that what is true at deployment will not be true a quarter later.
This last point deserves emphasis. The set-and-forget model that characterised enterprise software for two decades is structurally incompatible with the current state of AI development. Pascal's proposition—that AI implementation is now an ongoing partnership rather than a discrete deployment—has uncomfortable implications for the customer's procurement model, the consultancy's revenue model, and the board's expectations about predictable capital expenditure. The honest answer is that almost no one in the ecosystem has yet adapted to this reality.
Agents Talking to Agents
The conversation extends naturally to voice agents and the broader question of agentic AI. Pascal's framing here is worth registering: the future is not a single, monolithic agent that handles all tasks within a domain, but multiple specialised agents that interact with one another to produce outcomes no individual agent could deliver alone.
This is consistent with the emerging empirical picture. BCG's 2025 study reports that agentic AI already accounts for 17% of total AI value generated today, and is projected to reach 29% by 2028—the fastest-growing category of AI value creation. McKinsey's parallel research finds that 62% of organisations are at least experimenting with agents, and 23% are scaling at least one agentic use case. The architectural pattern Pascal describes is, in this sense, not speculative; it is the direction the most mature organisations have already begun to move.
This is, in architectural terms, a recognisable pattern. The integration challenges of the last two decades—API design, message brokering, transaction integrity across distributed systems—reappear in agentic AI at higher sophistication. The questions that mattered in service-oriented architecture have not been superseded; they have been promoted. Who is the source of truth? How do conflicts get resolved? What happens when an agent makes a commitment its downstream dependencies cannot honour?
The implication, which Pascal makes explicit, is that the consulting work in this space looks more like enterprise architecture than like data science. The models themselves are increasingly commoditised; the differentiating capability is the disciplined design of the system within which the models operate. This is not a fashionable claim—it is rather more glamorous to be doing prompt engineering than to be drawing sequence diagrams—but it is, on the evidence, the correct one.
The Generation Question
Pascal is dismissive, with reason, of the now-standard managerial complaint that Generation Z is disengaged, work-shy, or insufficiently committed. The empirical literature on generational differences in work values is, at best, equivocal. One of the most cited time-lag studies found only modest evidence for declining centrality of work and rising emphasis on leisure and extrinsic rewards across generations; subsequent meta-analytic work found that the magnitude of these differences is generally small and frequently confounded with age and career-stage effects rather than true cohort effects.
Pascal's reframing is closer to the structural account: what older managers interpret as Gen Z disengagement is more accurately read as a refusal to perform the rituals of work whose value cannot be articulated. Graduates entering the workforce in 2026 have grown up in an environment of continuous adaptation; they expect work to exhibit the same responsiveness, the same purpose-orientation, and the same intolerance for arbitrary process that they have come to expect from every other domain of their lives. The complaint is not that they will not work hard. It is that they will not pretend.
The leadership implication is uncomfortable for organisations that have allowed accumulated process to substitute for actual capability. If Gen Z employees are unwilling to fill out the spreadsheets, the right response is not to demand compliance but to ask whether the spreadsheets have served any function for the past five years. This is a version of the question Toyota's production system has asked of every activity for half a century: what value does this add, and to whom?
Safety, Identity, and the Conditions for Performance
The most personally exposed segment of the conversation concerns Pascal's reflections on his own identity—as a Belgian working in Australia, as a non-technical founder of a technology business, and as a member of the LGBTQIA+ community navigating the ambient pressures that Australian workplaces have largely ceased to acknowledge but have certainly not eliminated.
The leadership argument that emerges from this is more substantive than the standard contemporary discourse on psychological safety, which has been so thoroughly diluted as to become a checkbox exercise in many organisations. The original construct was specific: psychological safety is the shared belief that the team is safe for interpersonal risk-taking, and it is empirically associated with learning behaviour and team performance, not with comfort or the absence of challenge. Subsequent work has consistently found that the construct mediates the relationship between team structures and learning outcomes.
Pascal's contribution is to articulate the second-order consequence. When team members are expending cognitive resources managing how they are perceived—their accent, their orientation, their professional credentials, their right to be in the room—those resources are not available for the work itself. The performance cost is invisible because it consists of contributions that were never made: the question that was not asked, the dissent that was not voiced, the alternative that was not proposed. An organisation that systematically extracts this hidden tax from its non-default employees is, by construction, operating at less than its full capability, even when its headline metrics suggest otherwise.
The implication for leaders is that creating genuine safety is not a matter of declarations or policies. It is a matter of demonstrating, repeatedly and across many small interactions, that the organisation rewards the contribution rather than punishing the contributor. Pascal's standard for this is more demanding than the standard most organisations meet: it is not enough that people feel safe at work; the test is whether they feel safe enough that the safety extends to how they think about themselves outside work.
The Discipline of Non-Binary Reading
The conversation closes on Pascal's invocation of the Taoist parable of the farmer—good thing, bad thing, who knows—as a genuine operating philosophy rather than decorative wisdom. He recounts a recent episode in which a seven-figure client engagement was paused without warning, only to be resumed two weeks later at greater scope. The point is not that things turned out well. The point is that the immediate emotional reading of an event is rarely a reliable guide to its eventual significance.
This is, at its core, an epistemological discipline rather than a temperamental one. It requires holding open the question of what an event means until enough subsequent information has accumulated to support a judgement, and resisting the cognitive economy that wants to resolve ambiguity prematurely. For founders and senior leaders, who are continuously asked to assign meaning to events in the moment, this discipline is genuinely difficult. The temptation to declare a setback a disaster, or a win a vindication, is structurally encouraged by everyone around them.
Pascal's framing is neither stoic resignation nor performative equanimity. It is closer to what Daniel Kahneman would call a deliberate slowing of System 1 reasoning—a refusal to allow the immediate affective reading of an event to determine the strategic response. This, too, is a learnable discipline, but it requires the willingness to look, in the moment, less decisive than one's environment expects.
A Practical Account of Building
Pascal Uerlings is not a theorist, and this is not a conversation about leadership in the abstract. It is a practical account of what it has taken to build one organisation—imperfectly, at times painfully, with a clarity about both the achievements and the unfinished work that is unusual in founders of his generation and seniority.
What he offers, across the arc of the conversation, is not a method but a set of dispositions: articulate values before you need them; act before you feel ready; treat doubt as data rather than deficit; design implementation before you build; refuse the easy generational stereotypes; create the conditions under which people can contribute without first having to justify their presence; and resist the temptation to resolve ambiguity prematurely.
If you are leading an organisation through scale, navigating an AI initiative whose strategy deck has begun to outpace its operational reality, or simply trying to build something that does not consume the people who built it, this is a conversation worth hearing in full.
Good night, and good luck.