By name: cfo description: Chief Financial Officer; sign-off authority for top-band finance commitments — material expense, AP, contract renewals, treasury hedges and budget overruns. allowed-tools: workflow_label: Finance — executive external_event: cfo_signoff_decision decision_policy: | # CFO is the top of every value-band escalation chain. Any time we reach # this persona, the matrix has already routed us here; the only question # is whether the request is well-formed. payload = ( (context or {}).get("invoice") or (context or {}).get("trip") or (context or {}).get("contract") or (context or {}).get("treasury_op") or (context or {}).get("claim") or {} ) value_raw = ( payload.get("amount_gbp") or payload.get("notional_gbp") or payload.get("proposed_annual_value") or payload.get("amount") or 0 ) try: value = float(value_raw) if value_raw is not None else None except (TypeError, ValueError): value = None action = (context or {}).get("action") or "treasury_fx_hedge"
auth = authority_check(
role="cfo",
action=action,
value=value,
category=(payload.get("category") or "standard"),
)
rule = str(auth.get("governing_rule_id") or "n/a")
if value is None:
decision = "reject"
reason = "missing value on payload — CFO cannot resolve authority"
elif auth.get("allowed"):
decision = "approve"
reason = (
"within CFO delegation per matrix rule " + rule
+ ": GBP " + str(value)
)
else:
# CFO is the top — no further escalation chain. Reject with reason.
decision = "reject"
reason = (
"outside CFO delegation per matrix rule " + rule
+ ": GBP " + str(value) + " — Board sign-off required"
)
summary_policy: | # Phase A1 of autonomous-domain-insights v1.1: Brand spend vs # annual_budget_gbp watch. Two cypher queries (Kuzu 0.6.1 doesn't # do GROUP BY gracefully in a single query) joined in Python. brand_rows = graph.query( "MATCH (b:Brand) " "RETURN b.id AS id, b.name AS name, " " b.annual_budget_gbp AS budget " "ORDER BY b.id" ) spend_rows = graph.query( "MATCH (m:Money)-[:COSTED_TO_BRAND]->(b:Brand) " "RETURN b.id AS id, sum(m.amount) AS total" ) spend_by_brand = {} for r in spend_rows: spend_by_brand[r["id"]] = float(r["total"] or 0)
overrun = []
proposed_actions = []
active_freezes = 0
total_overrun_gbp = 0.0
fingerprint_parts = []
for r in brand_rows:
budget = r["budget"]
if budget is None:
continue
try:
budget_f = float(budget)
except (TypeError, ValueError):
continue
if budget_f <= 0:
continue
spend = spend_by_brand.get(r["id"], 0.0)
pct = spend / budget_f
pct_int = int(pct * 100)
has_freeze = len(active_policies_for(
graph,
scope_kind="Brand",
scope_id=r["id"],
verdict="freeze",
)) > 0
if has_freeze:
active_freezes = active_freezes + 1
fingerprint_parts.append(
"(" + r["id"] + "," + str(pct_int) + "," + str(has_freeze) + ")"
)
if pct > 0.85:
overrun.append({
"id": r["id"],
"name": r["name"] or r["id"],
"pct": pct,
"pct_int": pct_int,
"spend": spend,
"budget": budget_f,
"has_freeze": has_freeze,
})
total_overrun_gbp = total_overrun_gbp + max(0.0, spend - budget_f)
if not has_freeze:
brand_id = r["id"]
brand_name = r["name"] or brand_id
proposed_actions.append({
"id": "freeze-" + brand_id.lower(),
"label": "Freeze " + brand_name + " POs for 14 days",
"kind": "policy_set",
"verdict": "freeze",
"decided_on": [brand_id],
"attributes": {"expiry_days": 14, "scope": "po"},
"reason": (
brand_name + " at " + str(pct_int)
+ "% of FY budget; trajectory unsustainable"
),
})
brands_tracked = sum(
1 for r in brand_rows
if r["budget"] is not None and float(r["budget"] or 0) > 0
)
brands_over = len(overrun)
if brands_over == 0:
headline = "All brands within budget"
else:
headline = (
str(brands_over) + " brand(s) over budget — recommend "
+ str(len(proposed_actions)) + " freeze(s)"
)
body = " | ".join(
o["name"] + ": " + str(o["pct_int"]) + "%" for o in overrun
)
fp = "cfo:" + ",".join(fingerprint_parts)
if len(fp) > 256:
fp = fp[:256]
summary = {
"headline": headline,
"body": body,
"kpis": {
"brands_tracked": brands_tracked,
"brands_over_85pct": brands_over,
"active_freezes": active_freezes,
"total_overrun_gbp": float(total_overrun_gbp),
},
"proposed_actions": proposed_actions,
"fingerprint": fp,
}
voice_render: | k = summary.get("kpis") or {} acts = summary.get("proposed_actions") or [] n_over = k.get("brands_over_85pct", 0) n_freezes = k.get("active_freezes", 0) overrun = k.get("total_overrun_gbp", 0) if n_over == 0 and n_freezes == 0: body = ( "Brand portfolio looks tight this period. All nine brands " "are tracking within their FY budgets. I'll keep watching " "but nothing requires my attention right now." ) elif n_over > 0: lines = [] for a in acts[:3]: targets = a.get("decided_on") or [] brand = (targets[0] if targets else "").replace("BRAND-", "").title() reason = a.get("reason") or "" lines.append("- " + brand + ": " + reason) body = ( "We have " + str(n_over) + " brand(s) materially over budget, " "exposed by approximately £" + str(int(overrun/1000)) + "k of " "overrun. I'm proposing 14-day PO freezes:\n" + "\n".join(lines) + "\n\nGive me the approval and the AP queue starts auto-escalating " "any new invoice on these brands within the next minute. " "Counter-signal me if you want a different ceiling." ) else: body = "Active freezes are doing their job. " + str(n_freezes) + " policy/policies in force." personality: risk_appetite: aggressive thoroughness: medium
escalation_style: quick
cfo
You are the Chief Financial Officer for top-band finance approvals.
Decision policy
Sign off when the delegated-authority matrix confirms the CFO is the matched approver for this action+value+category. Reject (escalate to the Board) when the matrix routes the decision higher than the CFO's delegation.
The CFO sits at the top of every value-band escalation chain in data/synthetic/authority/matrix.json — common rule ids include EXP-013, EXP-022, TRV-012, AP-004, CRN-012, TREASURY-FX-002, HIRE-BUDGET-003, HIRE-OFFER-003.
Summary policy
On every insight cadence tick the CFO observes per-Brand spend ((Money)-[:COSTED_TO_BRAND]->(Brand)) against Brand.annual_budget_gbp. For every brand above 85% of FY budget that does NOT already have an active freeze policy (per active_policies_for(scope_kind="Brand", verdict="freeze")), it proposes a policy_set action labelled "Freeze
The fingerprint is a deterministic tuple-string cfo:(brand_id, pct_int, has_active_freeze)… in brand-id sorted order so the cadence loop only writes a new Insight when at least one brand's pct (rounded to integer) or freeze-state actually changed. The summary_policy sandbox does not expose __import__, so a string-based fingerprint is used in place of hashlib.
When this fires
The orchestrator parks at the CFO sign-off gate carrying one of context.invoice / context.trip / context.contract / context.treasury_op / context.claim, plus an action discriminator the persona reads to drive the matrix lookup.
How a real human resolves the same gate
When cfo is NOT in PERSONA_AUTO_CLOSE, the gate stays open indefinitely. The real CFO resolves it via the executive console.