← Research
Research

Is AI-first a strategy or a slogan?

Two chief executives declared it in April 2025. One held the line and one walked it back in a month. What the phrase commits an organisation to, and what it leaves undecided.

Last reviewed: 15 September 2026

AI-first is a slogan until it names which decisions the machine may make, which stay human and who is accountable. Shopify and Duolingo both declared it in April 2025; Duolingo's chief executive walked it back within a month. What the phrase commits you to and what it leaves undecided. An evidence review by Rahim Hirji; every figure resolves to a graded entry in the evidence base that says what it does not show.

Question this page answersAll 811 questions this research covers

AI-first is a preference order, and a preference order is a rule for one decision: try the machine before the person. It says nothing about the decisions that follow. Which work the machine may do unsupervised, which it may only draft, who checks it, who answers when it is wrong, and what the people still on the payroll must stay capable of doing. In April 2025 two chief executives declared AI-first within three weeks of each other. One published his memo and held the line. The other walked it back in a month, on LinkedIn, after his customers turned on him. The difference between the two is not courage. It is that one of them had written down what the phrase committed the company to and the other had not.

The answer, in one line

Not on its own. AI-first is a preference order: try the machine before the person. A strategy says which decisions the machine may inform, recommend or execute, which stay human, who is accountable for each, and what the organisation's people must remain capable of.

Share as a card

What the two memos said#

Tobi Lütke posted Shopify's internal memo on X on 7 April 2025 after it leaked. The sentence everybody quoted: 'Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI.' The sentence fewer people quoted was the question he wanted every team to ask: 'What would this area look like if autonomous AI agents were already part of the team?' That is a design question, and an answerable one. It asks the team to describe the work, not to describe the tool.

Luis von Ahn told Duolingo staff in late April that the company would 'gradually stop using contractors to do work AI can handle', that AI use would count in performance reviews, and that headcount would be added only 'if a team cannot automate more of their work'. The backlash was public and fast, on the app's own social channels. On 23 May he wrote on LinkedIn: 'I do not see AI as replacing what our employees do (we are in fact continuing to hire)' and 'I see it as a tool to accelerate what we do, at the same or better level of quality.' A month from replacement to acceleration, with no change in the technology in between.

What AI-first decides, and what it does not#

Read as a rule, AI-first decides exactly one thing: the order in which options are considered. It does not decide the boundary, which is the list of decisions the machine may inform, recommend or execute, and the ones it may never own. It does not decide accountability, which is who answers for a machine-assisted decision that goes wrong. It does not decide capability, which is what the organisation keeps practising so that somebody can still tell when the machine is wrong. And it does not decide the reinvestment, which is what happens to the time or the money the machine frees. Those four are the strategy. The preference order is the slogan that sits on top of it.

The reason the slogan travels is that it sounds like a decision. A chief executive who says AI-first has said something bold and has committed to nothing specific, which is a comfortable place to stand until somebody asks a specific question. The Duolingo case shows what happens when the specific question arrives from outside: the company had no written boundary to point to, so it retreated to the softest available position. Shopify's memo was not more careful about the boundary, but it was published, dated, and framed as a question about work rather than a statement about people, which gave it somewhere to stand.

The record on what follows#

The evidence on what happens after the declaration is thin and points one way. Klarna, the most-cited AI-first customer service case, had its chief executive say in May 2025 that cost had been 'a too predominant evaluation factor' and that the result was lower quality; the company recruited people again. IBM automated a few hundred human resources roles and, by its chief executive's account, put the money into engineers and salespeople so that total employment rose. Both are operator accounts and neither published figures, which is the grade they carry in the evidence base. What they share is that the outcome turned on a choice made after the slogan: what to do with what the machine freed. The declaration decided nothing about that, in either direction.

McKinsey's 2025 survey found that the two things most associated with profit from generative AI were the chief executive owning the rules for how it is deployed and the redesign of workflows, held by 28 and 21 per cent of organisations respectively. Neither is a preference order. Both are the second sentence that AI-first never gets round to.

What nobody has measured#

No study compares organisations that declared AI-first with organisations that did not, and the two memos are two companies in one month. Shopify has not published what its headcount rule did to hiring or output. Duolingo's walk-back is a statement of intent, not a record of what changed. The McKinsey correlation explains a fifth of the variance in a self-reported outcome. This page argues from the shape of the decisions rather than from measured results, and says so.

The test#

A leadership team that has declared AI-first can find out in an hour whether it has a strategy. Write the list: every consequential decision the organisation makes, with each marked inform, recommend, execute or never. Put a name against each. State which work stays unaided so that people can still check the machine, and where the freed time goes. If that document exists, AI-first is the headline on a strategy. If it does not, it is a slogan, and the next public question will find that out. The rules for writing it are at rules before tools; the argument for who writes it is at AI leadership.

Key sources

The four decisions the slogan leaves open are the subject of AI leadership; the list itself is how AI decision rights should be allocated; the reversals are gathered at what happened to the companies that cut staff for AI; and the question of who should hold the pen is at who should own AI strategy.

About this research#

Rahim Hirji is the author of SuperSkills (Kogan Page, 2026), keynote speaker on AI and human capability, and founder of The SuperSkills Intelligence Company. He has run, grown, bought and advised businesses with AI in them. Findings are attributed to the studies and statements that produced them and kept separate from the interpretation. This is a living reference, reviewed and updated as significant new evidence appears.

How this research works  ·  Reviewed quarterly  ·  Found an error? Tell me and it is corrected on the page.

Evidence review · SS-2026-241 · Graded against the published rubric

Cite this page

Hirji, R. (2026). Is AI-first a strategy or a slogan?. The SuperSkills evidence base, SS-2026-241. https://thesuperskills.com/research/is-ai-first-a-strategy-or-a-slogan. Last reviewed 15 September 2026.

An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.

How citations and IDs work
Questions answered on this page

Is AI-first a strategy?

Not on its own. AI-first is a preference order: try the machine before the person. A strategy says which decisions the machine may inform, recommend or execute, which stay human, who is accountable for each, and what the organisation's people must remain capable of. Shopify's April 2025 memo contained the first and none of the second; Duolingo's contained the same and was walked back within a month after a public backlash.

What did the Shopify memo actually say?

Tobi Lütke's memo, which he published on X on 7 April 2025, said that 'before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI', and asked teams what an area would look like if autonomous AI agents were already part of it.

What did Duolingo walk back?

In April 2025 Luis von Ahn told staff Duolingo would 'gradually stop using contractors to do work AI can handle' and add headcount only 'if a team cannot automate more of their work'. On 23 May 2025 he wrote on LinkedIn: 'I do not see AI as replacing what our employees do (we are in fact continuing to hire)' and 'I see it as a tool to accelerate what we do, at the same or better level of quality'.

In this hub

Definitions

The terms this field uses, defined against their primary sources.

Ask the evidence
What does the evidence actually show?What should our board be asking about this?Where does Rahim disagree with the consensus?
Bring this into your organisation

If this describes something happening in your teams, say so.

Keynotes, board sessions and advisory work, drawing on research across more than 200 organisations in 30 countries. Tell me the room, the date and the shift you need. A reply within 24 hours.

Start a conversation

Topics and audiences  ·  All research

Declaring AI-first is the easy sentence. Writing the four decisions underneath it, with the executive team, so the next public question has an answer, is the engagement. Board advisory.

This argument is one a board usually meets for the first time in the room. There is the boards and leadership version, and the full range of topics and audiences.

Box of Amazing

Rahim’s free weekly letter on AI and human capability

If this was useful, the weekly letter is where the thinking happens first. Most of what ends up on this site starts there. Weekly essays on AI, capability and the future of work. Read by 25,000 people, every week since 2017. Free, and one click to stop.

Opens Substack to confirm. No pitch in it, unsubscribe in one click, and nobody follows up because you read something.