The sectors the AI use-case reports forget

Deloitte's AI Institute publishes a compendium called The AI Dossier. The current edition runs to 86 use cases, each one specifying the business function it supports and whether it uses agentic AI. It is genuinely useful, and it is free.

It also covers six industries: Consumer, Energy & Industrials, Financial Services, Government & Public Services, Life Sciences & Health Care, and Technology, Media & Telecommunications.

Read that list again if you run a membership body, a union, a professional institute, a university or a charity. You are not in it. You are not in Gartner's sector cuts either, and you are not in most of what McKinsey publishes on applied AI.

The patterns transfer. The framing doesn't.

This matters less than it looks, because the underlying patterns are not industry-specific. What is specific is the framing — and framing is what makes a board say yes.

Take three from the Government & Public Services section, which is the closest analogue to member-led organisations because it deals with entitlement, casework and public trust.

Eligibility and entitlement

Deloitte's case covers benefits eligibility: conversational agents that handle intake, explain determinations and reference the rules they applied. The transferable pattern is any situation where someone asks "am I entitled to this, and why?"

For a union, that is the single most common member question there is. Am I covered. Does this apply to my grade. What happens if I raise it. Answering that consistently, at three in the morning, with a citation to the actual policy, is not a futuristic use case. It is the job.

Casework triage

Public sector permitting is about applying consistent criteria to varied applications at volume. Every professional body doing accreditation, every union running individual casework and every university processing applications has the same shape of problem — and the same risk if the answers vary by who picks up the file.

Documentation and knowledge retrieval

Health care's automated clinical documentation case is really about capturing what happened so the next person has it. Substitute a caseworker's notes, a regional officer's meeting record, or a course leader's feedback and the value is identical.

What the reports don't tell you

Each of those Deloitte cases carries a "managing risk and promoting trust" section, and for member-led organisations that section matters more than the use case itself.

The reason is structural. A retailer that gets a recommendation wrong loses a sale. A union that gets an entitlement answer wrong loses a member's case — and a member's trust in the union. Consistency, explainability and the ability to show which rule was applied aren't compliance overhead here. They are the product.

That changes the sequencing. In sectors with high transactional volume and low individual stakes, you pilot in production and learn fast. In member-led organisations, the first deployment has to be the one you would be comfortable explaining at an AGM.

Where to start

• Take the use case, not the industry. Pattern-match on the shape of the problem — entitlement, triage, retrieval — rather than looking for your sector on the contents page.
• Start where the rules are already written down. If your answer exists in a policy document, AI can find and explain it. If it lives in someone's head, fix that first.
• Ask what happens when it's wrong, before asking what happens when it's right. In member organisations the downside case sets the design.
• Pick something members already complain about. You will not have to manufacture a business case, and you will know within weeks whether it worked.

The absence of your sector from the major reports isn't evidence that AI has nothing to offer you. It's evidence that nobody has done the translation. That's a gap worth exploiting rather than waiting out.

Source: The AI Dossier, Deloitte AI Institute. Referenced here for analysis; the report is Deloitte's and is available free from their website.