Every Competitor Can Buy the Same AI

Almost every large organisation has now bought AI. Copilot sits inside Microsoft 365. ChatGPT Enterprise or Claude has a signed contract and a security review behind it. Salesforce, Adobe, SAP and the rest have shipped AI features into software that was already paid for. The budget was approved, the licences were distributed, and the internal announcement went out.

Then the reporting cycle came round, and the question changed. It stopped being whether AI works and became what it has actually changed.

McKinsey's State of AI survey, published in November 2025, gives a reasonable read on where most organisations sit. Nearly two-thirds of respondents said their organisations had not yet begun scaling AI across the enterprise, and just 39 percent reported any EBIT impact at enterprise level. Among those who did report an effect on profit, most put it below five percent of total EBIT. Adoption is close to universal. Measurable outcomes are not.

The gap between those two facts is where the next few years of competitive advantage will be decided.

The assumption most organisations are still making

When results fall short of the business case, the instinct is to buy more capability. Another platform. A wider licence pool. A specialist team, hired externally, to build the things the business apparently cannot build for itself.

Each of these is defensible in isolation. None of them changes the daily behaviour of the several hundred or several thousand people whose work makes up the organisation's cost base.

A licence is permission. It grants access to a text box. What it does not supply is the judgement to know which parts of your own job are worth handing over, how to give the model enough context to be useful, and how to tell a sound output from a confident one. That judgement is specific to the work. It cannot be procured, and it is not evenly distributed. In most organisations it currently sits with a small number of enthusiasts who worked it out on their own time.

Specialist hires have a related limit. They are expensive, hard to find, and they concentrate capability rather than spreading it. A central AI team can build twelve excellent things. It cannot sit next to a credit analyst and work out what her Thursday afternoon should look like.

The bottleneck is repeatable work

The organisations struggling with AI are rarely short of technology. They are short of tasks that have been converted into something a person does the same way every week.

This is more concrete than it sounds. Consider what it looks like function by function.

In finance, monthly reporting is a strong candidate. Variance commentary is written from the same numbers each month, in the same structure, for the same audience. The work is drafting, checking against the ledger, and rewriting for the board pack. An AI-assisted version of that task is not an experiment. It is a defined process with a defined input, a defined check, and a defined output, run twelve times a year.

In HR, job descriptions, interview scorecards and the first pass of performance review summaries all follow patterns. The pattern is what makes them suitable. The risk sits in a slightly different place, since a description that quietly imports someone else's assumptions about a role is worse than no description at all, so verification matters more than speed.

In operations, the candidates are process documentation, supplier correspondence, incident write-ups and recurring status reporting. These are the tasks people describe as necessary and unrewarding, which is usually a sign that the shape of them is stable enough to be improved.

In learning and development, the shift is more fundamental. If capability has to be built inside the work, the function's job stops being course delivery and becomes something closer to enablement design, with reinforcement built into the manager's routine rather than a calendar invitation.

None of these are transformation programmes. They are individual pieces of recurring work, made faster and more consistent by one person who has learned how to do it well, and then made durable by writing down how.

What the organisations pulling ahead do differently

BCG's fourth annual AI at Work survey, published in June 2026 and covering 11,749 employees across 14 markets, found something that ought to concern any executive who has approved an AI budget. Among frontline employees who use AI regularly, 42 percent save upwards of a full day each week, but 66 percent are given no guidance on what to do with the time they save, and more than half do not redirect it towards strategic work.

Time is being released. It is not being captured. That is an operating question rather than a technology one, and the same survey points at where the leverage sits. BCG found that clear strategy lifts AI's impact by 25 percentage points, against five points for better tools.

What that looks like in practice is fairly ordinary.

People know which of their tasks are appropriate for AI and which are not, because they have tried both and been told which is which. They verify outputs against a source rather than reading for plausibility. Managers ask for the workflow, not the anecdote, so a good method spreads from one analyst to a team instead of staying with the analyst. Learning happens against real deliverables, in the systems people already use, rather than in a two-hour session that competes with the day job. Over time AI stops being an application someone opens and becomes part of how a process is run.

McKinsey found the same shape at the other end of the distribution. The highest-performing organisations stand out for thinking beyond incremental efficiency gains, redesigning workflows and treating AI as a catalyst for change in how the organisation operates.

Where the advantage settles

The models will keep improving, and every competitor will have access to them at roughly the same time, at roughly the same price, with roughly the same capabilities. Whatever edge exists in the technology itself has a short half-life, measured in months.

The durable variable is the number of people in an organisation who can apply it well to work they already understand, and the speed at which a method discovered by one of them reaches all of them.

That is a capability question, and it responds to the things capability has always responded to. Practice on real work. Feedback from someone who knows what good looks like. Evidence that someone can actually do the thing rather than a record that they attended.

Where Etiq fits

Etiq helps organisations turn AI investment into practical, repeatable and measurable outcomes by helping existing employees apply AI inside their real workflows.

Learning happens on the job. Rather than a course completed away from the work, people build with their own role's context and data, against tasks that already sit in their week, with guardrails that check outputs and flag claims the data does not support. Capability is verified against what someone has produced, so leaders see evidence of applied skill across teams and departments instead of completion rates. Deployment runs in public cloud, private cloud or on-premise, using your own LLM keys.

The closing thought

The advantage was never the software. Everyone has the software.

What separates the organisations that see a return from the ones that see a line item is quieter and harder to copy: the number of people inside them who know how to do their own work better with the tools already sitting on their desk. That capability is not bought. It is built, in the work, by the people already doing it. Every organisation will eventually run the same models. Very few will have taught their people to use them well, and that is the only part of this that compounds.

Discover how at etiq.ai.

‍