Research with AI often begins well: you get a clear summary, useful-sounding links, and a confident conclusion. The harder question comes next: Which parts can I actually use for a decision?

In this Course, you take responsibility for one research task at Harbour Community Support, a synthetic 12-employee nonprofit in England. Management is considering whether to do more internal work on a possible four-day-workweek pilot. The Course supplies the case facts and identifies every prepared source packet or standard handoff. You decide what to ask AI, what to check, what to correct, and what management can responsibly decide next.

You begin by turning a broad topic into one immediate decision. Then you build an evidence gate, use AI to discover candidate sources, open decision-changing originals, and record what each accessible layer can support. You compare the source with AI's wording, correct claims that go too far, preserve mixed evidence and unknowns, and stop when the next answer must come from Harbour rather than the public web.

The challenge grows gradually. One familiar label may hide different arrangements. A real source may be accessible through one route but restricted through another. A positive result may measure wellbeing rather than service performance. A short test may reveal operational problems without proving a lasting effect. You will make these distinctions yourself instead of watching a hidden researcher clean them up.

By the end, you will have a practical way to:

• frame a research request around a real decision;
• explain where the supplied material came from;
• use AI for discovery without treating it as the source;
• authenticate a citation and record its access state;
• check and correct a decision-changing claim;
• separate source fact, synthesis, inference, and unknown;
• decide when the evidence supports action, no action, or “not enough to decide”;
• hand over a conditional brief with clear human owners and recheck triggers.

You build requests in understandable layers and see why each field matters.

Finishing the Course means the public research has an honest status and a usable handoff. It does not mean Harbour has chosen a model, approved a pilot, changed anyone's work, or completed the internal and professional checks that still belong to people.

Frequently asked questions

Do I need research or statistics experience?

No. The Course explains the evidence distinctions when you need them and builds each AI request in understandable layers. You do not need an academic background or a perfect long Prompt.

Do I need my own AI account or unrestricted web access?

No. The Lessons provide the recorded AI work, key source material, and access results needed for the guided decisions. Opening original links is optional exploration; applying the method to your own topic requires current browser access and whichever AI or search tools you choose, subject to their access rules.

Must I be able to access the full text of every source?

No. You learn to record whether you opened a summary, abstract, manuscript, or full text and to limit each claim to that accessible layer. A restricted source may remain a lead, but you—not AI—decide whether the available evidence is sufficient to use.

Can I apply the method to my own topic?

Yes, for low-risk questions that can be investigated through public evidence. Replace the Harbour case with your own decision, scope, and applicability checks. Medical, legal, financial, immigration, and other high-stakes decisions still require appropriate professionals and should not rely on this Course alone.