diff --git a/src/__tests__/ai-reasoning-admin-guard.test.ts b/src/__tests__/ai-reasoning-admin-guard.test.ts
index f40ab3103..221419db5 100644
--- a/src/__tests__/ai-reasoning-admin-guard.test.ts
+++ b/src/__tests__/ai-reasoning-admin-guard.test.ts
@@ -149,3 +149,28 @@ describe("every call that carries reasoning took it from the resolver", () => {
     ).toEqual([]);
   });
 });
+
+/**
+ * The Coach turn receives its level on `TurnInput.reasoningLevel` and carries
+ * it through the loop under its own names, which the setter shapes above do
+ * not see. The one place that level is decided is the chat route, so the
+ * route is pinned: it resolves the level, and it does not fall back to a
+ * constant that would skip the operator's switch.
+ */
+describe("the Coach turn's level comes from the resolver", () => {
+  const route = stripComments(
+    readFileSync(join(SRC, "app/api/insights/chat/route.ts"), "utf8"),
+  );
+
+  it("resolves the level in the chat route", () => {
+    expect(route).toMatch(
+      /reasoningLevel[^=]*=\s*await\s+resolveCoachTurnReasoningLevel\s*\(/,
+    );
+    expect(route).toMatch(RESOLVER_IMPORT);
+  });
+
+  it("never hands the turn a constant level", () => {
+    expect(route).not.toMatch(/\bDEFAULT_REASONING_LEVEL\b/);
+    expect(route).not.toMatch(/reasoningLevel\s*:\s*["']/);
+  });
+});
diff --git a/src/app/api/insights/chat/__tests__/golden-harness.ts b/src/app/api/insights/chat/__tests__/golden-harness.ts
index 7590d7ceb..867f12e45 100644
--- a/src/app/api/insights/chat/__tests__/golden-harness.ts
+++ b/src/app/api/insights/chat/__tests__/golden-harness.ts
@@ -175,6 +175,8 @@ export function resetGolden(): void {
     },
   );
   m.getEvent.mockReturnValue({
+    // A request event, so the per-request cache memoises as in production.
+    getKind: () => "request",
     setError: (err: Error) =>
       record("event.setError", { message: err.message }),
   });
diff --git a/src/app/api/insights/chat/__tests__/golden-transcripts.test.ts b/src/app/api/insights/chat/__tests__/golden-transcripts.test.ts
index 96f297edb..1703970b6 100644
--- a/src/app/api/insights/chat/__tests__/golden-transcripts.test.ts
+++ b/src/app/api/insights/chat/__tests__/golden-transcripts.test.ts
@@ -63,6 +63,8 @@ vi.mock("@/lib/db", () => ({
     user: { findUnique: h.m.userFindUnique, update: h.m.userUpdate },
     coachConversation: { findFirst: h.m.conversationFindFirst },
     coachMessage: { findMany: h.m.coachMessageFindMany },
+    // v1.41 — the operator's reasoning controls; an untouched instance.
+    appSettings: { findUnique: vi.fn(async () => null) },
   },
 }));
 vi.mock("@/lib/rate-limit", () => ({
diff --git a/src/app/api/insights/chat/route.ts b/src/app/api/insights/chat/route.ts
index 7fe9a17de..971f391f8 100644
--- a/src/app/api/insights/chat/route.ts
+++ b/src/app/api/insights/chat/route.ts
@@ -37,10 +37,8 @@ import { coachChatRequestSchema } from "@/lib/ai/coach/types";
 import { listConversations } from "@/lib/ai/coach/persistence";
 import { detectRefusal } from "@/lib/ai/coach/refusal";
 import { runCoachTurn } from "@/lib/ai/coach/turn/pipeline";
-import {
-  DEFAULT_REASONING_LEVEL,
-  type ReasoningLevel,
-} from "@/lib/ai/reasoning/levels";
+import type { ReasoningLevel } from "@/lib/ai/reasoning/levels";
+import { resolveCoachTurnReasoningLevel } from "@/lib/ai/reasoning/controls";
 import { streamProviderError, streamRefusal } from "@/lib/ai/coach/turn/sse";
 
 /**
@@ -182,10 +180,13 @@ async function handleChatRequest(request: NextRequest): Promise<Response> {
   }
 
   // v1.41 — the person's effective reasoning level: their choice after the
-  // operator's switch and highest level. Integration points this one call
-  // site at the shared resolver (`src/lib/ai/reasoning/resolve.ts`); until
-  // then every turn thinks at the default level.
-  const reasoningLevel: ReasoningLevel = DEFAULT_REASONING_LEVEL;
+  // operator's switch and highest level, from the shared resolver. Who pays
+  // is not known until the turn resolves its chain, so the turn applies the
+  // operator-paid ceiling itself (`turn/reasoning.ts`); the provider's own
+  // ability is left to its client.
+  const reasoningLevel: ReasoningLevel = await resolveCoachTurnReasoningLevel(
+    auth.user.coachPrefsJson,
+  );
 
   // v1.41 — the person's turn slot. The counter is the rate-limit row's own
   // (an atomic upsert), given back when the turn is over; a refused slot is
diff --git a/src/lib/ai/reasoning/__tests__/controls.test.ts b/src/lib/ai/reasoning/__tests__/controls.test.ts
new file mode 100644
index 000000000..11f6a196f
--- /dev/null
+++ b/src/lib/ai/reasoning/__tests__/controls.test.ts
@@ -0,0 +1,90 @@
+/**
+ * v1.41 — the Coach turn's level as the chat route resolves it: the stored
+ * preference against the operator's controls, read through the real loader
+ * against a stubbed settings row.
+ */
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+const { row } = vi.hoisted(() => ({
+  row: { current: null as unknown, fail: false },
+}));
+
+vi.mock("@/lib/db", () => ({
+  prisma: {
+    appSettings: {
+      findUnique: vi.fn(async () => {
+        if (row.fail) throw new Error("db down");
+        return row.current;
+      }),
+    },
+  },
+}));
+vi.mock("@/lib/logging/context", () => ({
+  annotate: vi.fn(),
+  getEvent: () => undefined,
+}));
+
+import {
+  loadReasoningControls,
+  resolveCoachTurnReasoningLevel,
+} from "../controls";
+
+beforeEach(() => {
+  row.current = null;
+  row.fail = false;
+});
+
+describe("loadReasoningControls", () => {
+  it("reads an untouched instance as on and uncapped", async () => {
+    expect(await loadReasoningControls()).toEqual({
+      enabled: true,
+      maxEffort: "high",
+    });
+  });
+
+  it("fails closed when the row cannot be read", async () => {
+    row.fail = true;
+    expect((await loadReasoningControls()).enabled).toBe(false);
+  });
+
+  it("reads an unknown stored cap as no cap", async () => {
+    row.current = { aiReasoningEnabled: true, aiReasoningMaxEffort: "xhigh" };
+    expect((await loadReasoningControls()).maxEffort).toBe("high");
+  });
+});
+
+describe("resolveCoachTurnReasoningLevel", () => {
+  it("runs a never-chosen preference at medium", async () => {
+    expect(await resolveCoachTurnReasoningLevel(null)).toBe("medium");
+    expect(await resolveCoachTurnReasoningLevel({ tone: "warm" })).toBe(
+      "medium",
+    );
+  });
+
+  it("reads a malformed stored level as the default", async () => {
+    expect(await resolveCoachTurnReasoningLevel({ reasoning: "max" })).toBe(
+      "medium",
+    );
+  });
+
+  it("clamps the stored level to the operator's cap", async () => {
+    row.current = { aiReasoningEnabled: true, aiReasoningMaxEffort: "low" };
+    expect(await resolveCoachTurnReasoningLevel({ reasoning: "high" })).toBe(
+      "low",
+    );
+  });
+
+  it("is off when the operator switched reasoning off", async () => {
+    row.current = { aiReasoningEnabled: false, aiReasoningMaxEffort: "high" };
+    expect(await resolveCoachTurnReasoningLevel({ reasoning: "high" })).toBe(
+      "off",
+    );
+  });
+
+  it("is off when the controls cannot be read", async () => {
+    row.fail = true;
+    expect(await resolveCoachTurnReasoningLevel({ reasoning: "high" })).toBe(
+      "off",
+    );
+  });
+});
diff --git a/src/lib/ai/reasoning/controls.ts b/src/lib/ai/reasoning/controls.ts
index 4beea7ef7..9751573ec 100644
--- a/src/lib/ai/reasoning/controls.ts
+++ b/src/lib/ai/reasoning/controls.ts
@@ -18,7 +18,12 @@ import {
   type CoachPrefs,
 } from "@/lib/validations/coach-prefs";
 
-import type { BackgroundReasoningJob } from "./levels";
+import {
+  DEFAULT_REASONING_LEVEL,
+  isReasoningLevel,
+  type BackgroundReasoningJob,
+  type ReasoningLevel,
+} from "./levels";
 import {
   coachReasoningCeiling,
   completionReasoning,
@@ -45,7 +50,17 @@ const UNREADABLE: ReasoningAdminControls = Object.freeze({
  * inside a background run, so switching reasoning off at 02:05 stops the
  * nightly pass that began at 02:00 for every account after the switch.
  */
-export function loadReasoningControls(): Promise<ReasoningAdminControls> {
+export async function loadReasoningControls(): Promise<ReasoningAdminControls> {
+  // The request cache itself can fail (an event without a kind); the
+  // controls then fail closed like an unreadable row, never the caller.
+  try {
+    return await loadMemoisedControls();
+  } catch {
+    return UNREADABLE;
+  }
+}
+
+function loadMemoisedControls(): Promise<ReasoningAdminControls> {
   return memoizePerRequest(
     "ai-reasoning-controls",
     async () => {
@@ -174,3 +189,26 @@ export async function resolveJobReasoning(
   });
   return completionReasoning(resolved, job);
 }
+
+/**
+ * The effective Coach level for a turn: the person's stored choice
+ * (`coachPrefsJson.reasoning`, absent read as the default) after the
+ * operator's switch and cap. Who pays is applied inside the turn, once its
+ * chain is known; the provider's ability is left to its client. Never
+ * throws: an unreadable preference reads as the default, unreadable
+ * controls as "off".
+ */
+export async function resolveCoachTurnReasoningLevel(
+  coachPrefsJson: unknown,
+): Promise<ReasoningLevel> {
+  const stored =
+    coachPrefsJson && typeof coachPrefsJson === "object"
+      ? (coachPrefsJson as { reasoning?: unknown }).reasoning
+      : undefined;
+  return resolveReasoning({
+    surface: "coach",
+    userPref: isReasoningLevel(stored) ? stored : DEFAULT_REASONING_LEVEL,
+    admin: await loadReasoningControls(),
+    costOwner: "user",
+  }).effort;
+}