Some Professionals Are Designing the AI Wave. The Rest Are About to Be Redesigned By It.

Category: The Intersection  |  Structure: Expert vs AI Prediction

Walk into almost any organisation today and you will find two things that seem to contradict each other.

Everyone is talking about AI. And most of the actual work — the reports, the customer calls, the financial models, the content briefs — looks remarkably similar to how it looked three years ago.

That contradiction has led many thoughtful professionals to a quiet conclusion: AI is more hype than substance. The gap between what is promised and what is delivered is simply too wide.

That conclusion is wrong. And dangerously so.

The gap is not evidence that AI has underperformed. It is evidence that the real wave has not arrived yet.

---

The Gap Has a Name

In March 2026, Anthropic published a research report that measured something most AI commentary ignores: the difference between what AI can theoretically handle in a professional role, and what organisations are actually using it for. They called this the gap between Theoretical Capability and Observed Exposure.

The findings are striking.

Computer programmers — the most AI-exposed occupation in the study — sit at 74.5% observed exposure. Yet the entire Computer and Math category sits at only 33% observed against 94% theoretical capability. Business and Finance shows similarly high theoretical reach but low actual deployment. Legal sits at roughly 15–20% observed against 80% theoretical potential.

Most professional functions are operating at less than a third of AI’s theoretical capability. Not because the technology is missing. Because the deployment is.

That is the gap. And it is about to start closing — not gradually, not over a decade, but in the next one to two years across specific functions. The closing will look sudden to those who were not watching. It will not be sudden at all.

Read the full Anthropic report here: Labor Market Impacts of AI, March 2026 →

---

What AI Predicts

I gave an AI the Anthropic report’s framework and asked it directly: which professional functions will cross the 50% observed exposure threshold first, and by when?

Here is what it said:

Software developers — by mid-2026. Already at 74.5% observed exposure. Coding agents are scaling fastest and enterprise tooling is maturing rapidly.

Customer service representatives — by end-2026. Already at 70.1%. Core tasks are automatable and enterprise deployment is accelerating.

Data entry and records processing roles — by early 2027. At 67.1%. The primary task of reading and entering data is almost entirely automatable.

Financial analysts and reporting roles — by mid-2027. At 57.2%. AI agents for forecasting and reporting are rolling out across financial institutions.

The AI’s prediction is logical. It ranks functions by current observed exposure and draws a straight line forward. The roles closest to 50% will cross first.

Clean. Data-driven. And largely wrong about the reasons — which means it will also be wrong about the timing in ways that matter.

---

What Experience Predicts

Here is the distinction the AI cannot make.

The functions that will cross the 50% threshold first are not simply the ones with the highest current observed exposure. They are the ones where the force pulling the gap shut is strongest — where organisational incentive, business case clarity, and implementation speed are all aligned at once.

Observed exposure data tells you where AI has already arrived. It does not tell you where the conditions for rapid adoption are ripening right now.

Here are the four functions I predict will cross first — and why the reason matters as much as the timing.

Software developers and technical roles — within one year

Agreement with the AI on timing. Disagreement on the reason.

The AI attributes this to tooling. Coding agents, integrated development environments, enterprise AI deployments. These are real. But they are not the primary driver.

The real driver is feedback loops. Technical roles have the shortest gap between deploying AI and measuring its impact. A developer who uses AI to write code knows within hours whether the output works. That tight loop between action and result builds confidence fast — and confidence is what converts augmentation into automation.

When professionals can see AI’s contribution immediately and precisely, they deploy more of it. Technical roles have this advantage above all others. That is why adoption here will accelerate beyond what tooling alone predicts.

Customer-facing service roles — within one year

The AI sees a linear progression from 70.1% upward. What it misses is the cascade dynamic.

Organisations have been cautious about automating customer interactions because of brand risk. The fear is not technical — it is reputational. What happens when AI gets it wrong in front of a customer?

That caution is eroding. Not because the risk has disappeared, but because the first movers are demonstrating that well-designed AI deployments in customer service produce more consistent outcomes, not less. When one organisation in a sector crosses successfully and visibly, others follow within months. The decision to automate stops being a leap of faith and becomes a competitive necessity.

We are at that tipping point now across multiple sectors simultaneously. The cascade has begun.

Content, communications and marketing roles — within one to two years

This is where my prediction diverges most sharply from the AI — and where the distinction between augmentation and automation becomes critical.

The Anthropic report measures automated, work-related AI usage. Much of what is happening in marketing right now is augmentative — humans using AI to draft, refine, and iterate on content that they still own and approve. This registers as lower observed exposure than the reality of AI’s involvement.

The shift from augmentation to automation in content production is already underway. When AI moves from helping a marketer write to writing on behalf of a marketer — with human review rather than human creation — the observed exposure data will register a sudden jump.

That jump will look like a surprise. It will not be. It is the inevitable endpoint of a transition that has been building quietly for two years. Marketing and communications functions are further along this path than the data currently shows.

Financial reporting and compliance roles — within one to two years

The AI predicts this function crosses the threshold because of efficiency gains — faster reporting, lower cost, better analytics.

The deeper driver is risk reduction.

Compliance requirements are actively accelerating AI adoption in finance because AI produces more consistent, auditable outputs than human-generated reports. This is not primarily about doing things faster. It is about doing them more reliably and with a cleaner audit trail.

When the incentive for adoption is regulatory and reputational rather than purely economic, the adoption is stickier. Organisations do not reverse course on compliance-driven automation the way they might on efficiency-driven experiments. Financial reporting and compliance functions are being pulled toward AI by forces that are more durable than cost savings alone.

---

The Question Underneath The Prediction

The Anthropic report contains one observation that deserves more attention than it has received.

Thirty percent of workers currently have zero observed AI exposure. Their tasks appear too infrequently in usage data to register at all. For some, this is because their work is genuinely beyond AI’s current reach — physical, situational, deeply relational.

But for others, the observed exposure is zero not because AI cannot do the work, but because nobody has yet designed the solution.

That distinction — between cannot and have not yet — is where the next two years will be decided.

The gap between theoretical capability and observed exposure is not a measure of AI’s limitations. It is a measure of the thinking that has not yet been applied to deployment.

In every function listed above, the organisations that cross the 50% threshold first will not do so because they found a better tool. They will do so because someone in that organisation designed a solution — thought rigorously about the problem before reaching for the technology, understood the workflow before automating it, and built something that held together under real organisational conditions.

That person is the AI solution architect. The role did not exist by that name two years ago. It is now the most consequential professional capability of the decade.

---

The Debate Worth Having

Some will read this post and conclude that these functions are under threat. That is the wrong frame.

Every function on this list is under transformation. The threat is not to the function — it is to the professionals within it who are waiting to see what happens rather than deciding what happens.

The wave is not arriving. It is already here, gathering at the shore. The gap is closing. The question is not whether your function appears on this list.

The question is whether you are designing the crossing — or whether the crossing is being designed around you.

That answer is still yours to write. But the window for writing it is narrowing faster than most people realise.