Work Intelligence Fellowship

Three shortages. One mechanism.

Around 900 people are killed at work every day, and most of it comes down to how the work was planned, judged or done not to a machine failing. Universities need real capstone clients. Graduates want work that matters and cannot get near it. Each of those shortages is what another one needs.

The Convergence

Three problems usually discussed in three different rooms.

Put them next to each other and they turn out to be one shape.

01

Industrial safety

Operators in oil, gas, mining, construction and utilities

Not knowledge of the causes — those are documented and familiar. What is missing is first-hand evidence of how they keep recurring inside operations that are, by every formal measure, well run.

02

Capstone clients

Universities running final-year programmes

Corporate sponsors who arrive with a brief, a named contact and a defined deliverable — rather than a logo, a guest lecture and a vague invitation to engage with industry.

03

A way in

Recent graduates

Not education. Something concrete to point at, and a named person who can vouch for what they actually did. 42% of recent US graduates are working jobs that never required their degree.

Students need real work with real stakes. Universities need clients with structure. The industry needs candid conversations it cannot have with a vendor in the room.

The Subject

Seven causes. Each research team takes one.

More than seven in ten oil and gas accidents involve a human or organisational cause rather than an equipment failure. That is not a claim about carelessness it means the cause sits in the part of the job no system records. These are the findings investigations keep arriving at, written the way they appear in a report.

01

Never Practised

The crew was doing this procedure for the first time, at the equipment.

02

Lost experience

The person who knew how to do it that way retired in March.

03

When things go wrong

They have only ever practised it going right. Nobody rehearses the bad day.

04

Told, not tested

We told them — but we never verified they could actually do it.

05

The plan did not match the job

The permit did not cover the entire job, and nobody verified the isolation.

06

No guidance in the moment

He skipped a step, on the wrong asset, and nobody saw it.

07

The same mistake again

It happened before. Nothing changed. So it happened again.

None of this is secret. Any safety professional recognises all seven within seconds. The shortage is not knowledge — it is structured, first-hand evidence of how they keep recurring, gathered by people with nothing to sell.

How it works

Five things, over one academic term.

01

A brief, not a blank page

A defined research question, a selection brief and a background pack. Teams that spend three weeks deciding what to study deliver nothing in twelve.

02

A named partner on every team

Someone at EON who supervises scope, handles escalation and is reachable for the whole term. Unowned teams fail, and it is preventable.

03

The research itself

Structured interviews with operations, safety and training professionals about a single one of the seven causes.

04

A deliverable that is theirs

A sector evidence pack — findings, method, sources — returned to the institution and to the professionals who contributed, and assessable as coursework.

05

A reference that says something

Each student gets a letter describing the work they actually did and who they actually spoke to.

The Honest Part

On artificial intelligence, and where the criticism lands.

It would be convenient to present this as a rebuttal to the criticism of AI. It is not, and pretending otherwise would be the sort of move that earns the criticism in the first place.

Criticisms we think are fair

  • Entry-level displacement. The junior tasks that used to teach people their profession are the tasks most exposed to automation. How much of the graduate squeeze is attributable to AI is genuinely contested — but the concern is legitimate, and dismissing it is not an argument.
  • Optimising for engagement rather than benefit. A great deal of consumer technology is tuned to hold attention. That is a design choice with documented consequences.
  • Deployment without accountability. Systems placed into consequential decisions where nobody can inspect what they did.

What is different here — and how you can check

  • The success measure is external. Incidents and fatalities are counted by regulators and industry bodies, not by us. If the numbers do not move, we do not get to redefine the metric.
  • It augments the expert rather than replacing them. The system captures what an experienced person knows. Remove the expert and there is nothing to capture.
  • The knowledge belongs to the customer. Nothing leaves their environment without written agreement — a contractual commitment, not a policy statement.
  • This programme creates entry-level experience. Twenty-four students in the first cohort. That is a small number, and we report it as a small number.

What we are not claiming

  • That AI prevents accidents. People prevent accidents. The technology makes what competent people already know available at the moment and place it is needed.
  • That this solves graduate unemployment. Twenty-four students is twenty-four students.
  • That any figure about our own effect is measured. Where a number is ours rather than published, we say so.

The Rules

Published, so that they are enforceable.

The most likely way this fails is not weak research. It is one student somewhere representing it as something other than what it is. So the rules are public, and the sponsorship is on the record.

  • Students are not employees, are not paid, and are not selling.
  • They disclose their sponsor and the purpose of the research in the first sentence of every approach.
  • No commercial conversation — no pricing, no proposals, no negotiation. Those questions go to their EON partner.
  • Interviewees may decline, speak off the record, or withdraw their contribution afterwards.
  • Misrepresentation ends participation.

The first cohort

Twenty-four students, nine universities, starting September.

Ten degree programmes. Nothing in the design depends on these particular institutions, on North America, or on oil and gas — which is why we are opening it to more universities for the next intake.

Take Part

Join the Fellowship.

EON Introduces Verified Work Episodes to Turn Real-World Operational Experience into Proprietary Work Intelligence

One fellowship. Several ways to contribute.

Universities can explore hosting a future cohort. Students can register their interest in meaningful research experience. Operations, safety and training professionals can contribute their first-hand knowledge through a structured interview.

Whatever your role, the Fellowship team will explain how you can participate, how the research works, and what you will receive in return.