HomeThe memo that changed the hiring conversation
On a Tuesday morning in April 2025, Tobias Lutke posted an internal Shopify memo publicly on X. Within hours it had spread across every management Slack channel, LinkedIn feed, and founder group chat that mattered. By the end of the week it had been quoted in boardrooms from San Francisco to Singapore. The memo was four paragraphs long. It did not announce a layoff. It announced something more structurally significant: a new hiring gate, and the gate was AI.
The operative sentence, as reported by CNBC and corroborated by Lutke's own public posting of the document, was this: before any team at Shopify could request additional human headcount, they were required to demonstrate why AI could not accomplish the work instead. "Using AI effectively is now a fundamental expectation of everyone at Shopify," Lutke wrote. He added that AI use would be built into performance reviews and peer evaluations — not as a bonus competency, but as a baseline requirement of the job itself.
Read that again slowly, because the implications travel much further than one Canadian e-commerce company's internal policy.
Shopify is not a fringe experiment. It powers a material share of global e-commerce, employs thousands of people across engineering, product, marketing, operations and support, and is led by a founder with a documented, unusually coherent philosophy about technology and work. When Lutke speaks about the nature of software and organisations, serious operators listen — not because he is always right, but because he has a track record of being directionally right early. He was writing about "software eating the world" implications before most CEOs had noticed the first bite.
What the memo announced was not a layoff. It announced a new default. The old default in every growing company is: when you need more output, hire more people. Lutke's memo flipped that default: when you need more output, first prove you cannot get it from AI. Only if you can make that case — and the bar is explicit — does a human hire enter the conversation.
This is a subtle but seismic shift. It is the difference between AI as a tool your employees optionally use and AI as the first resort your organisation mandatorily attempts. It is the difference between productivity software and a structural redesign of the hiring calculus itself. And it is, whether or not Singapore's business community has registered this yet, the direction the entire cohort of sophisticated operators is moving. The memo is the policy expression of a philosophy that will arrive at the desks of Singapore's HR managers, finance directors, and CEOs — the only question is whether they arrive prepared or surprised.
This piece is for the operators and investors who want to be prepared.
A confident executive in a modern Singapore office reviews a tablet displaying an AI workflow diagram, city skyline visible through floor-to-ceiling glass, navy and warm amber tones, shallow depth of field, cinematic
The world-class move: what Shopify actually did and why it is generationally significant
To understand why the Shopify memo is worth this level of attention, you need to understand what it is not. It is not a cost-cutting measure dressed up in AI language. It is not a PR stunt. It is not a response to a specific operational problem. It is the formalisation — in policy, in performance review, in hiring process — of a philosophy about where value in an organisation actually comes from.
Lutke has been explicit about this philosophy in various forums over the years: software compounds in ways that human headcount cannot. A hundred engineers writing great software can create leverage that ten thousand people doing manual work never achieves. The memo applies that compounding logic to every function in the company, not just engineering. It says: before you assume a person is the answer, prove that AI is not the better one.
That is a genuinely world-class move, and here is why. Most companies adopt AI the way they adopted every previous wave of productivity software: they buy the tools, train the willing, and let adoption drift in from the edges. The org chart does not change. The headcount model does not change. The defaults do not change. What you get, predictably, is a marginally more efficient version of the old operating model — and a line in the investor deck about "AI-enabled efficiency" that means approximately nothing.
Lutke's approach is structurally different. By making AI competency a gate on new headcount, he forces a conversation that most organisations systematically avoid: what is this role actually for, and can a machine now do a meaningful part of it? That conversation is uncomfortable. It challenges assumptions that have not been questioned in years. It requires managers to decompose jobs into tasks rather than treating job titles as atomic units. It exposes the difference between roles that exist because they create genuine value and roles that exist because they were created in 2017 and nobody thought to revisit them.
This discipline — forced task decomposition before every hire — is precisely what most organisations need and almost none impose on themselves. The reason is structural. Hiring more people is the path of least resistance. It is the culturally safe answer to "we are overstretched." It does not require the manager to think hard about which tasks actually need a human. It does not require a conversation about which parts of the role could be automated. It just requires approval from a VP and a job description recycled from three years ago. Shopify's memo makes that path of least resistance unavailable. The friction has been deliberately moved to the default answer so that the better answer gets a fair hearing.
Consider the second element of the memo: AI proficiency built into performance reviews. This is not incidental. A company that evaluates people on their AI use is not running an experiment; it is redefining what it means to be effective at your job. When you measure something, you change the behaviour around it. When Shopify makes AI proficiency part of how it assesses performance, it creates an incentive structure that propagates the AI-first operating norm throughout the organisation without requiring top-down policing. The memo becomes self-enforcing, not through surveillance but through alignment of incentives with the philosophy. That is organisational design, not HR communications.
The third element — that teams must demonstrate AI is insufficient before requesting human headcount — creates what economists would call a soft constraint that reveals preferences. In practice, a team that has to make the case for a human hire in writing, against the explicit alternative of an AI solution, will think about the problem differently than a team that simply sends a headcount request to the talent acquisition queue. Some of those teams will discover that AI actually can do what they needed. Others will discover that they need the human and can now articulate clearly why — which is itself valuable, because roles that exist for reasons their owners can articulate are more durable than roles that exist by inertia.
The broader context matters here. Shopify had gone through significant workforce reductions in 2022 and 2023 — cuts that were, at the time, explained partly in terms of pandemic-era over-hiring and a return to leaner operations. By the time Lutke posted the April 2025 memo, the company was not in a moment of financial stress. It was in a moment of strategic choice: how do you build an organisation that captures the full compounding potential of AI, rather than merely deploying AI on the margins of a traditionally structured company? The memo is the answer to that question, formalised as policy. It is a declaration that the compounding logic of software, so familiar in the engineering function, now applies to the whole organisation.
"Using AI effectively is now a fundamental expectation of everyone at Shopify." The word 'fundamental' is doing heavy lifting there — not 'encouraged,' not 'welcomed,' not 'a competitive advantage.' Fundamental. As in: the same category as being able to read, write, and work professionally with other people.
That framing is deliberate and significant. When AI proficiency becomes fundamental rather than advanced, it stops being a differentiator and becomes a floor. The organisations that treat it as a floor will pull away from those that treat it as a ceiling.
The misread: why most observers got this exactly wrong
The coverage of the Shopify memo fell, with depressing predictability, into two camps. One read it as "Shopify does not want to hire humans anymore." The other read it as a visionary CEO unlocking exponential productivity. Both missed the actual point, but the first misread is the more dangerous one — because it is the one that produces the wrong operational response.
Shopify's memo is not about replacing people with AI. It is about redesigning how organisations make decisions before people are brought in. That distinction sounds pedantic until you trace the implications.
The replacement narrative produces a specific set of bad decisions. Boards pressure leadership to cut headcount as a demonstration of AI adoption. Managers feel that keeping their teams intact signals a failure to "go AI-first." Employees interpret any AI rollout as preparation for their elimination and quietly disengage. The result is a smaller, more anxious organisation that has not actually redesigned its work — it has just done it with fewer people under more stress. That is not operating leverage. That is cost reduction with a technology narrative stapled on.
The redesign narrative produces a completely different set of decisions. Teams are asked to map their work into tasks and think honestly about which tasks are genuinely human and which are automatable. New roles are defined around the judgment, creativity, and accountability that AI cannot replicate. Existing people are reskilled upward into those roles rather than replaced. The organisation gets leaner in the right places and more capable in the places that matter. That is what the memo was actually asking for.
The distinction between replacement and redesign is not semantic. It lives in the task decomposition.
A job is not an atom. A job is a molecule — a bundle of tasks, each with its own characteristics, each with its own answer to the question "can AI do this?" Consider a mid-level marketing manager at a Singapore SME. Across a given week she is writing briefs, reviewing agency outputs, analysing campaign data, managing relationships with vendors, presenting to leadership, attending client calls, and deciding — using experience, pattern recognition and relationship knowledge — what the next strategic move is. Generative AI and AI analytics tools can do meaningful parts of that list: drafting briefs, first-pass data analysis, generating presentation structures. They cannot do the rest: the vendor relationship that took three years to build, the judgment about which metric actually matters to this leadership team, the read of the room in the client call that changes how you pitch next quarter.
When you automate the drafting and the first-pass analysis, you have not eliminated the marketing manager. You have changed the centre of gravity of her role toward the things that were always actually valuable. She spends less time generating the commodity outputs and more time on the judgment, relationships, and strategic decisions that no model can own. Same person. Materially higher-value day. That is redesign, not replacement.
The WEF Future of Jobs 2025 report, which should be the empirical grounding of every serious conversation on this topic, projects approximately 170 million new roles and roughly 92 million displaced globally by 2030, for a net positive of around 78 million. Around 86% of employers expect AI to transform their businesses in that period. Read that carefully: the projection is not elimination but churn and recomposition. Tasks dissolve and recombine into new roles faster than job descriptions can keep up. The organisations that lose are those whose people are locked into task bundles that AI has hollowed out, with no path to the higher-value reconfiguration. The organisations that win are those that actively manage the recomposition.
But honesty demands a calibration alongside the optimism. A net global gain of 78 million roles is a macro statistic that provides zero comfort to the specific person whose specific task bundle was just automated in a specific firm in a specific industry. The displaced 92 million and the created 170 million are not the same people in the same place with the same skills. The role that disappears is immediate and concrete. The role that appears requires reskilling that takes time, support, and a labour market willing to absorb the transition. The aggregate optimism is warranted. The local and individual disruption is real. Both things are true at once, and a good operator holds both rather than defaulting to whichever one is more comfortable.
The Shopify memo, read correctly, is on the right side of this. It does not say "replace your people with AI." It says "before you add a person, prove AI cannot do it." That is a constraint on the growth of the team, not a mandate to shrink the existing one. The redesign the memo implies is: get more from the people you have by moving them into the task bundles that AI cannot touch, while routing the tasks AI can handle to AI. That is the compounding story. The shrinkage narrative is a misread, and a costly one.
Redesign, not replacement: the three-bucket model
So how does the Shopify philosophy translate into an operational framework that a Singapore business can actually use? Start with the discipline Lutke's memo forces — task decomposition — and formalise it into a model. Not a consultant's abstraction, but a concrete sorting exercise that any team can run in a half-day workshop and act on by the following Monday.
Take any role in your organisation. Decompose it into every distinct task performed across a typical week. Do this honestly, at a granular level, not "manages marketing" but the seventeen specific things that constitute managing marketing. Then sort each task into one of three buckets.
Bucket one: what AI does better. High-volume, pattern-based, repetitive tasks where the output is good enough with a human edit and the cost of AI error is low. First-draft copy and content. Data formatting and first-pass analysis. Generating report structures and presentation skeletons. Meeting summaries and action-item extraction. Research compilation from known sources. Repurposing one asset into ten. Answering the standard customer enquiry from the script. These tasks belong to AI — not because we do not value the humans who did them, but because routing routine production through AI and human through to judgment is the redesign that makes both more valuable. The cost of keeping these tasks human is not just the labour cost; it is the opportunity cost of a skilled person spending sixty percent of their week on work that does not need their skill.
Bucket two: what only humans should own. Accountability for outcomes. Relationships built on trust and history. Judgment about what is actually good — not technically correct, but genuinely right for this client, this culture, this moment. Ethical decision-making in ambiguous situations. The accountability chain that makes an organisation trusted by its customers, regulators, and partners. Original creative leaps that depend on genuine experience and point of view. AI cannot own any of these, not because it is not sophisticated, but because ownership and accountability are definitionally human in every legal, cultural, and reputational sense that matters. In Singapore's FEAT-governed financial sector, this is explicit and auditable. In every other sector, it is the implicit deal between a business and its customers: a human is accountable for this. Gut that layer, and you gut the trust that revenue is built on.
Bucket three: what they do better together. This is the leverage layer, and it is where the compounding actually happens. The strategist who now generates and tests twenty positioning hypotheses in the time it used to take to brainstorm two. The analyst who runs ten data scenarios before breakfast and arrives at the board meeting with a point of view rather than a printout. The operations manager who lets AI monitor a hundred process variables and flags only the ones that need her attention. The sales professional who uses AI to prepare so thoroughly for each meeting that every call feels like the third conversation, not the cold first. In each case the human is not removed from the loop. The human is amplified by the loop — doing more valuable work, more often, at higher quality, without the cognitive cost of the routine work that was always in the way.
Three glowing buckets arranged on a sleek dark surface — left blue, centre warm gold, right a blend of both — representing AI tasks, human tasks, and collaborative tasks, macro cinematic render, no text
This three-bucket model is the operational chassis that the Shopify memo implies without articulating. When Lutke says "demonstrate why AI cannot do this before requesting headcount," he is asking teams to do bucket-one triage. When he says AI proficiency is a fundamental expectation, he is asking every employee to master bucket-three collaboration. The implicit message about bucket two — what humans must own — is the part the memo leaves to the judgement of leaders, and it is the part that most implementations get wrong by omitting it entirely.
The most expensive version of getting the three-bucket model wrong is not putting too much in bucket one. It is putting too much in bucket two as a defensive reflex — insisting that AI cannot really do that, that the human touch is essential everywhere, that quality requires the full traditional role unchanged. That is the other failure mode, and it is the more common one in organisations where AI adoption is being driven by anxiety rather than strategy. The organisations that win are not the ones that automate everything and hope for the best. They are the ones that make clear-eyed, task-by-task decisions about which bucket each activity belongs in — and then redesign roles, incentives and workflows to match the new distribution.
The Shopify memo is, at its core, a forcing function for that discipline. It makes the three-bucket analysis mandatory at every new hire decision. The companies that get ahead will not wait for the hire decision to force the analysis. They will run it across every function, on a regular cadence, as a standard part of strategic operations. As we explore in the death of the job description in Singapore, the traditional static role definition is itself the structural problem — a job description written in 2022 is a snapshot of a task bundle that may look very different by 2026, and almost certainly will look different by 2028.
What this means for Singapore
Singapore should read the Shopify memo not as a story about a foreign technology company managing its headcount in a competitive North American labour market. It should read it as a preview of the management philosophy that is about to become the global baseline — and ask, with the urgency the timeline deserves, what it means here, specifically.
The first thing it means is that Singapore's SME-heavy economy has both a vulnerability and an advantage that are mirror images of the same structural fact. The vulnerability: most Singapore SMEs run lean, have not invested in formal AI capability-building, and are about to face hiring conditions — from multinationals, from well-capitalised local competitors, from clients with new expectations — where AI competency is increasingly a prerequisite. The advantage: Singapore has a national institutional infrastructure for managed workforce transition that almost no other economy can match. The question is whether that infrastructure gets activated ahead of the disruption or after it.
Consider the industries where the Shopify logic lands hardest in Singapore. Professional services — the legal, accounting, and consulting firms that constitute a large share of the economy's high-value employment — are already watching AI recompose the task bundles of junior roles at speed. The junior analyst who spent three years building spreadsheet models may find that an AI can replicate the mechanics of the model in minutes; the value is now in interpreting the output and advising the client, which requires experience and relationship the junior analyst is building — but needs to build faster now, because the mechanical repetitions that used to be the training ground are being automated away before they have had time to compound into expertise. The profession does not disappear; the apprenticeship model that produced competence inside it may need fundamental redesign.
Financial services face a regulatory layer that makes the three-bucket model not optional but mandatory. The Monetary Authority of Singapore's FEAT principles — Fairness, Ethics, Accountability, Transparency — mean that AI systems touching customers in financial contexts must be explainable, auditable, and governed by a clear human accountability chain. For DBS, OCBC, UOB and the wider financial ecosystem, this is the daily operational reality: AI can accelerate and augment, but the human who owns the decision is not a rubber stamp — she is a genuine accountability node who must be capable of explaining and defending what the system did. FEAT makes bucket two non-negotiable in every institution that falls under MAS supervision, which in practice covers a significant share of Singapore's white-collar employment. That is not a constraint on AI adoption; it is a design specification for how human roles must be redesigned to accommodate it.
The governance playbook for Singapore organisations deploying AI alongside human decision-makers is worth reading alongside this piece, because the FEAT-compliance dimension reframes the entire Shopify-memo logic for the Singapore context: the question is not just "can AI do this task?" but "what human accountability structure must exist around the AI that does it?" The answer to the second question defines what bucket-two roles look like in a FEAT-governed organisation, and those roles are not small or peripheral — they are the governance layer that makes the whole system legitimate.
Retail and e-commerce — Shopify's own heartland — are undergoing the same recomposition locally. Singapore's retail sector has been navigating margin compression for years; AI is now the most significant lever available for doing more with leaner operations. The question for local retailers is whether AI adoption leads to a smaller team doing the same operational tasks with less slack, or a redesigned team doing fundamentally higher-value customer-facing and strategic work. The first is a cost measure. The second is a capability investment. The Shopify memo is a call to the second, and Singapore's retailers are capable of answering it — but they need the framework and the will to do the task decomposition that makes the redesign real.
Marketing, content, and creative services — where the Shopify operating model has direct adjacency — face the three-bucket logic at every layer of the production chain. Singapore's agencies and in-house teams are already deploying AI for drafting, variation testing, and first-pass localisation. The question is whether they are simultaneously redesigning what their human teams do with the time that creates, or simply producing more volume with the same headcount configuration and calling it "AI-powered." The first produces compounding capability. The second produces a brief surge of output followed by exhaustion and diminishing quality, because the human roles have not been redesigned to match the AI-augmented workflow.
Microsoft's 2026 Work Trend Index named this specifically: a "redesign gap" where AI-driven productivity gains are outpacing organisational redesign across the economy. Companies are buying tools faster than they are rebuilding the work around them. In Singapore's SME-heavy economy, that gap is both the risk and the opportunity. The businesses that close the redesign gap first — that map their task bundles, sort them into the three buckets, and rebuild roles around the human layer that AI cannot replicate — will pull ahead of competitors who treated AI as a feature rather than a redesign mandate.
The honest acknowledgement is that the Shopify memo, arriving in a Singapore context, has implications that extend beyond any single company's hiring policy. It signals that the global talent market is moving toward a world where AI competency is a hiring baseline, not a differentiator. A Singapore job-seeker or incumbent employee who cannot demonstrate genuine AI proficiency — who cannot operate in bucket-three collaboration mode at a meaningful level — will face a progressively narrower market. That is not a prediction designed to cause anxiety; it is a description of the direction of travel, offered early enough that the people it affects can move before the gap becomes a cliff.
The Singapore enablers: a system built for this moment
Here is where the Singapore story diverges sharply from the Silicon Valley narrative — and where the divergence is an advantage, not a softening.
The Shopify memo dropped into a labour market context where the implicit social contract is: companies optimise, workers adapt or lose out, and the government provides a safety net for the ones who fall through. Singapore's architecture is different. It is tripartite by design and by conviction — government, employers, and unions building the labour market together, sharing information, sharing risk, and coordinating the managed transition that pure market dynamics would handle brutally. That is not sentimentalism. It is hard infrastructure, and for the specific transition that AI is producing, it is exactly the right infrastructure.
Workforce Singapore and NTUC's e2i together run the most sophisticated job-redesign and Career Conversion Programme infrastructure in the region. These are not retraining schemes in the abstract sense — the generic "go learn something new" advice that is simultaneously obvious and useless. They are structured programmes that work at the level of the specific role, in the specific industry, at the specific firm, mapping the task bundle that is changing and building a funded pathway to the redesigned role that absorbs the change. A Singapore SME whose marketing coordinator role has been restructured by AI is not on its own in figuring out what the redesigned role looks like and how to move the incumbent into it. There is institutional expertise, there is co-funding, and there is a framework for doing the redesign in a way that is fair, documented, and genuinely useful to the person going through it.
SkillsFuture extends this logic to the individual level, providing the credits and the course ecosystem that let a worker invest in their own reskilling without depending entirely on their employer to fund it. For the bucket-two professional whose role is being redesigned — not eliminated — SkillsFuture is the mechanism that makes "upgrade your AI collaboration skills" a realistic instruction rather than an unfunded aspiration. The credit system is not generous by global consulting standards, but it signals something more important than the dollar amount: a national commitment to the proposition that transition is managed, not abandoned.
The tripartite model does something subtler and more durable still. Because government, employers, and unions operate in the same conversation — about sector skills frameworks, about fair retrenchment practices, about what "responsible AI deployment" looks like in employment — there is a shared language and a shared accountability structure that most other economies lack. A Singapore business implementing the Shopify-style AI-before-headcount logic is not doing it in an institutional vacuum. It is doing it in a context where the National AI Strategy 2.0 has set the direction, where MAS FEAT principles have set the governance floor for the regulated sector, where tripartism means that a union has a seat at the redesign conversation, and where the alternative to a managed redesign is not "just cut and move on" but a reputational, regulatory, and talent-market cost that most businesses would rather avoid.
This is not a naive picture. The Singapore system has limitations, and they matter. Career Conversion Programmes are most powerful for workers who are already in formal employment with an employer willing to co-invest in the transition. Gig and contract workers — a growing share of the Singapore workforce — sit at the edges of this infrastructure, with less access to the structured support. The SkillsFuture credits, while meaningful, are not a substitute for an employer who is actively redesigning roles rather than quietly eliminating them. And the tripartite model, which is genuinely excellent at managing predictable transitions in stable industries, faces real pressure when the pace of AI-driven change outstrips the speed at which the institutional conversation can move. None of those limitations change the fundamental verdict: Singapore's infrastructure for this transition is better than almost anywhere else in the world. They simply mean that the infrastructure needs to keep evolving at the pace the technology demands.
For the governance and grant dimensions of this transition — especially for SMEs trying to understand what IMDA, ESG, and MAS frameworks mean for their AI deployment strategy — FMC Collective specialises in exactly this navigation, making the institutional architecture practical rather than theoretical. And for the implementation layer — actually standing up AI systems, designing the human-in-the-loop workflows, and ensuring the redesign produces working capability rather than a deck and a vendor contract — Freemansland operates at that intersection of strategy and execution. The point is not to advertise; it is to make concrete that the enabler ecosystem in Singapore is thick enough to actually support the redesign that the Shopify memo implies.
The tripartism, the Career Conversion infrastructure, the SkillsFuture system, the sector frameworks, the FEAT governance layer — together these constitute a redesign-before-you-reduce environment that is, frankly, the most favourable context in the region for a business that wants to do this well. The Shopify memo arrives in Singapore at a moment when the national architecture is more prepared to support a good response to it than almost any comparable economy. The question is whether Singapore's business community reads the memo and activates that support — or waits until the hiring market has moved and reacts from behind.
The operator's playbook: five numbered moves for Singapore businesses
Enough framing. If you run a Singapore business — SME, professional services firm, agency, in-house team — the Shopify memo compresses into five practical moves you can begin this quarter. They are ordered deliberately: execute them out of sequence and you get the costs without the benefits.
1. Run the task audit before any headcount decision — without exception.
The Shopify memo makes this mandatory for new hires. Make it mandatory for yourself before that external pressure arrives. Pull a representative month of output across every function and decompose it into tasks, not roles. Be granular: not "manages the content function" but the thirty specific things that managing the content function actually consists of across a week. For each task, ask three questions honestly: can AI do a meaningful version of this with human oversight? Does this require genuine accountability and judgment that must sit with a person? And does this task become more valuable when a human and AI do it together? The ratio you discover — how much of each role sits in bucket one, two, and three — is your redesign blueprint. You cannot redesign what you have not decomposed, and you cannot decompose what you have not looked at. Most organisations have not looked. The task audit is the prerequisite to everything else.
2. Automate bucket one — visibly to your team, not as a surprise.
Once you have identified the bucket-one tasks — the routine, high-volume, pattern-based production that AI genuinely handles well — move them to AI tools with human oversight. Do this transparently: tell your team explicitly that you are automating the grind so that they can own the work that actually matters. The single fastest way to kill AI adoption inside an organisation is to let people believe it is covert preparation for their elimination. The fastest way to accelerate it is to make clear that the automation creates capacity for higher-value work — and then actually deliver the higher-value work for people to move into. If you automate bucket one and fail to redesign bucket two, you have not redesigned the organisation; you have just extracted labour time and handed it back as free time or (worse) additional bucket-one work. Do not do that.
3. Redesign bucket-two roles upward — and let title, scope and pay reflect the upgrade.
The redesigned role is not the old role with the boring parts removed. It is a fundamentally different role with a higher centre of gravity: more judgment, more accountability, more creative authority, more strategic ownership. Write the new role description honestly. Give it a title that reflects the upgrade. Pay for the new scope, not the old one. This is the move that determines whether AI transformation produces a retained, motivated team doing more valuable work, or a residually resentful team doing the same work faster under the impression that they survived a culling. The redesigned role is the investment that makes the automation sustainable. It costs more in the short term than simply removing the automated tasks and banking the difference. It produces far more in the long term, because the people who own the judgment layer are your institutional knowledge, your client relationships, and your quality ceiling — and those do not regenerate quickly if you let them leave. The discipline is: redesign before you reduce, and redesign generously enough that the people going through it understand that this is an upgrade, not a layoff with a new name.
4. Use Singapore's institutional rails — they are built for exactly this.
This is the move most Singapore SMEs leave on the table. WSG's Job Redesign initiative provides co-funding and structured support for the precise work of decomposing roles and rebuilding them around AI-augmented workflows. E2i's Career Conversion Programmes fund the transition of an incumbent from the old role shape to the new one — covering training costs and, in some cases, salary support during the conversion period. SkillsFuture credits are available to your people individually for AI-related upskilling across a wide ecosystem of providers. Do not run this redesign as a unilateral internal exercise when there is co-funding available to do it properly. The institutional rails exist because Singapore has made a policy decision that managed transition is preferable to market-driven disruption. You are, quite literally, leaving subsidised assistance on the table if you do not engage them. For the FEAT-governed layer — anything in or adjacent to financial services — MAS guidance and the FEAT principles framework are the design specification for your human accountability structure, not an optional compliance overlay. Build to the spec before you deploy.
5. Build AI competency into your hiring and performance criteria — now, not when it is industry-standard.
The Shopify memo made AI proficiency a formal performance expectation. That is a wise structural move that Singapore businesses can and should adopt ahead of the industry norm, because adopting it early costs very little and adopting it late, when the talent market has already divided into AI-competent and not, costs a great deal. In your next hiring cycle, include genuine AI competency as an evaluated criterion — not "familiar with ChatGPT" but demonstrable ability to use AI tools to produce work, edit AI output, and operate in the bucket-three collaboration mode that defines the high-value human layer. In your next performance review cycle, include AI adoption as an assessed dimension. You are not penalising people who have not yet become AI-fluent; you are signalling, clearly and early, that the direction of travel is non-negotiable and the support to make the journey is available. The businesses that make this signal early retain the employees who respond positively to it and lose the ones who do not — which is the exactly correct selection at this particular moment.
A sixth move is implicit in all five: read the question of who manages the AI agents at your organisation with the same urgency as the Shopify memo. The organisational redesign that AI requires is not just about the task bundles of individual roles. It is about governance — who is accountable for AI system outputs, how errors are caught and corrected, what the human override structure looks like, and how the organisation learns from AI failures rather than hiding them. The Shopify memo is about the hiring gate. The governance question is about the operating structure on the other side of that gate. Both matter, and treating either as an afterthought is where the redesign fails.
The investor close: operating leverage and the metric that actually matters
For anyone allocating capital or running a business with an eye on its eventual value, the Shopify memo is ultimately a lesson in where AI value accumulates on the income statement — and it is not where the simplest version of the narrative points.
The naïve read of "prove you can't use AI before hiring a human" is: Shopify's headcount grows slower, costs grow slower, margins expand, multiple re-rates. That story is true as far as it goes, but it is the beginning of the analysis, not the end. The more important story is operating leverage, and operating leverage is not the same as headcount reduction.
Operating leverage is the ability to grow revenue — to serve more customers, produce more value, expand into more markets — without growing the cost base proportionally. AI creates operating leverage when it multiplies what each human can accomplish, rather than simply substituting for human labour at a one-for-one rate. The difference between those two things is the difference between a one-time margin improvement and a compounding capability advantage. The first is visible in a single year's income statement. The second builds over years and shows up as a business that is structurally more valuable than its competitors, not just currently more profitable.
A Singapore investor studies an upward-trending revenue-per-employee chart on a large screen in a dim boardroom, navy and warm gold ambient lighting, shallow depth of field, no text visible, cinematic editorial
The metric that distinguishes the two is revenue per employee. A business that cuts headcount and holds revenue constant shows a rising revenue per employee — but the mechanism is cost reduction, not capability growth. That is a diet, not a redesign. It has a natural floor: you can only cut so far before the capability deficit shows up in the revenue line. A business that keeps its team, redesigns roles, and moves people into bucket-two and bucket-three work while routing bucket-one tasks to AI shows a rising revenue per employee because the same humans are now directing more AI-augmented output. That mechanism has no floor. As AI capability compounds, the leverage available to each human in the redesigned role compounds with it. The business gets structurally more productive with each cycle of AI improvement — not because it keeps cutting, but because its people are positioned to capture each improvement as it arrives.
The question for any investor evaluating an "AI transformation" story is straightforward: is the revenue per employee rising because you cut the denominator, or because you grew the numerator? One is a diet that ends. The other is a redesign that compounds. They look identical in year one. They look very different in year five.
For Singapore specifically, this framing has an additional dimension. Singapore's competitive advantage in global markets is not low labour cost — it never has been. It is the quality, trustworthiness, and capability of its talent pool, operating within a stable institutional and regulatory environment. A Singapore business that pursues AI-driven headcount reduction as its primary strategy is trading away its core competitive advantage for a margin improvement that better-positioned competitors in lower-cost markets will replicate at lower base cost. The strategy that is actually defensible is the redesign: keeping Singapore's talent, moving it up the value stack, and building an organisation whose output quality and governance standards justify premium pricing in the markets where those things command premium prices.
That is the operating leverage story that the Shopify memo, read correctly, is pointing toward. Not "do more with less," but "do better with the same, and let the same become more capable with each passing year as the tools improve." The investor who can distinguish between those two stories in a management presentation will make better decisions than the one who treats any mention of "AI efficiency" as equivalent evidence for either.
The Singapore business that implements the three-bucket model, uses the institutional rails to fund the transition, redesigns roles upward rather than eliminating them downward, and builds AI competency into its talent pipeline will show the compounding version of revenue per employee over a three-to-five-year horizon. That is the Shopify memo's real promise — not a leaner organisation in 2025, but a structurally more capable one in 2028.
The hire that will not happen and the role that will
Tobias Lutke's memo is four paragraphs long. It will be studied, argued about, partially misunderstood, and selectively quoted for years. But its operating logic is simple enough to fit on a single slide: AI is no longer optional infrastructure. It is the first resort. Human hiring is the second.
That logic is spreading. Not because Shopify is uniquely influential, but because it is correct — and the companies that figure out that it is correct are the ones that will compound. The ones that treat AI as a software subscription layered onto an unchanged organisation are already being outpaced, even if their income statements have not made it legible yet.
Singapore sits in an unusual position at this particular moment. The technology is global. The disruption it creates is real and specific. But the institutional architecture to manage that disruption — the Career Conversion infrastructure, the SkillsFuture system, the tripartite model, the FEAT governance framework — is local, mature, and genuinely excellent. The window to use that architecture well, to be ahead rather than behind the redesign that the Shopify memo represents, is open now. It will not be open indefinitely.
The memo says: prove you cannot use AI before you hire a person. The better question it is really asking — the one worth taking into every leadership team conversation this year — is this: have you redesigned the work around what AI makes possible, or are you still building an organisation around what AI is about to change?
One of those questions has a comfortable answer. The other has the right one.
Redesign before you reduce. The memo just made it official.