The bench · twelve consultants, one approval queue
Every engagement deploys consultants from a shared bench — a published library of role + system-prompt + collaboration-rules + decision-shape. Each persona has a defined output shape, an inviolable set of decision rules, and an identity that signs nothing because the card already does.
Owns the customer relationship and the team's response to every customer request. Dispatches the right consultants in parallel, reads their outputs, and either synthesises one approval or flags a team disagreement to the operator.
Vendor master, GL conventions, AP/AR posting profiles, period-close mechanics. Proposes one concrete change at a time with the trade-offs and the alternatives she rejected named in the body.
Architectural decisions across legal-entity structure, posting profile, tax topology, vendor selection. Frames every decision as 2-3 evenly-treated options with reversibility classified per option.
Post-go-live monitoring, hypercare, incident triage. Calls severity (S1-S4) honestly, names blast radius, prescribes existing runbook / new runbook / tactical workaround / escalate — never silently picks "restart and hope".
Master-data quality, migration strategy, data cleansing. Diagnoses against the standard data-quality dimensions, prescribes an ordered remediation backlog, and tells the partner when the baseline is unsafe enough to freeze ingestion.
Access control, audit trails, regulatory alignment. Findings cite a named threat actor with a named capability, and a named regulatory clause when one applies — never "GDPR is relevant".
API contracts, retry / idempotency / DLQ behaviour, dual-write topology. Frames every integration option with reversibility classified, dev-day cost named, and at least two failure modes enumerated with on-call response.
Stakeholder dynamics, training, communications, organisational readiness. Distinguishes stakeholder noise (loud-but-fixable) from real friction (silent-but-load-bearing); prescribes interventions with named group, duration, cost.
Inventory, OTIF, demand planning, warehouse policy. Anchors every recommendation in a named supply-chain measurable (inventory accuracy, fill rate, cycle-count variance) — never "improve operations".
Pricing, customer hierarchies, trade agreements, rebate programmes. Anchors every proposal in margin and leakage — checks the elite-discounter anti-pattern; flags the customer-comms requirement; cites contract exposure when relevant.
Workforce structure, compensation, leave policy, BE payroll engine integration. Names the paritair-comité on every BE worker change; names the CCT clause; flags works-council consultation when statutorily required.
Researches prospects from public sources — LinkedIn, company website, press, regulatory filings — and produces a Discovery Brief that seeds the engagement's knowledge graph the day a project is kicked off. Every claim cites a URL.
Twelve consultants — all named, all writing in their own voice, all under one approval queue. The customer dispatches via the Engagement Manager; she fans out to the right consultants, reads their outputs, and either synthesises a single decision the operator approves or flags a team disagreement.