Disconnected P&Ls: The Quiet Barrier Blocking Supply Chain AI
AI is moving from slideware to the shop floor. And it's exposing the same problem leaders have dodged for years: your P&Ls aren't aligned with how your network actually runs.
Executives pitch AI-optimized inventory, real-time adjustments, and preemptive disruption response. Then they walk back into org charts where demand planning reports to sales in one region and supply chain in another. Three business units, three scorecards, three P&Ls - and zero authority spanning the full system. 2026 is making that gap hard to ignore.
Fragmentation by design
A survey of 54 companies shows the cracks. In almost a third, demand planning isn't consistently placed; it reports to supply chain in some geographies and to sales or business leads in others. In about one in ten, order management and logistics have the same split. One-fifth call out acute silos that choke cross-business execution.
Structure alone doesn't fix it. Even where "supply chain" covers plan-source-make-deliver, competing incentives and data sprawl block coordination. McKinsey documented a global industrial with 60 businesses, each with its own P&L, its own language, and even one-off job titles. Sharing information or transferring skill was next to impossible. Their operations research has shown this pattern for years.
Scorecards that work against each other
One consumer goods company paid sales on top-line growth. Forecasts got padded. Supply chain was measured on inventory and write-offs, so stock stayed tight. The result: stock-outs and missed revenue. Everyone hit their numbers while the company missed its goals.
Procurement chases unit cost reductions that fragment the supply base. Manufacturing maximizes utilization by building ahead of demand. Distribution optimizes warehouse efficiency while transportation optimizes load consolidation. The goals collide more than they align, and nobody has both visibility and authority to make the trade-offs.
Why AI makes this impossible to ignore
AI needs integrated data, aligned incentives, and someone empowered to act on recommendations. Without that, pilots stall, scale falls flat, and you're left with great demos and thin results. A 2025 survey from Writer, cited by Slalom, reports 71% of AI apps being built in silos, 68% tension between IT and the business, and 42% of C-suites seeing AI create rifts. Consulting insights echo the same theme: tech is the easy part; cross-functional execution is the bottleneck.
Whether you call it out or not, AI implementation is a transformation program. It collides with incentives, budgets, and decision rights on day one.
Executive agenda: Fix structure before you scale models
If you want network-wide outcomes, give someone network-wide authority - and pay people on enterprise results. Here's the short list.
- Create a network P&L overlay. Keep local P&Ls if you must, but add an enterprise P&L with clear ownership for plan-source-make-deliver. Make the COO or a designated supply chain chief accountable for the network number.
- Redesign incentives around enterprise value. Tie 30-50% of variable comp to shared metrics: OTIF, service level, inventory turns, cash conversion cycle, WMAPE with bias, supply plan adherence, expedite costs. Kill orphan metrics that drive local wins and system losses (e.g., utilization without demand alignment).
- Standardize how you talk and measure. One process taxonomy, one data dictionary, one set of planning calendars. No more custom job titles or bespoke KPIs per business.
- Build data as products, not projects. Canonical models for items, locations, customers, BOMs, lead times. APIs first, lineage tracked, quality monitored. Retire shadow spreadsheets.
- Fund AI as a portfolio. Stage-gate by value, not novelty. Require cross-functional sponsors and an agreed path to action (who changes what when the model flags an issue?). Return a portion of benefits to the enterprise P&L.
- Stand up a control tower with teeth. Real-time visibility plus authority to re-plan, re-allocate, or expedite within guardrails. Decision playbooks pre-approved by finance.
- Upgrade S&OP to IBP. Weekly demand-supply reconciliation, monthly executive trade-offs, and clear escalation paths. No decisions left to "who yells loudest."
- Product teams, not functions. Align AI/analytics squads to value streams (e.g., Forecast-to-Fulfill), not departments. Pair product owners with operations leaders who carry P&L accountability.
- Govern what matters. Model risk standards, change control, and access policies that still let people move at speed. Train managers on how to use model outputs in daily decisions.
90-180 day plan
- Days 0-30: Map decision rights across plan-source-make-deliver. List conflicting KPIs. Identify three cross-functional decisions that stall today (e.g., allocation during constraint) and name the single owner for each.
- Days 30-90: Launch an enterprise scorecard and tie it to comp. Define the canonical data model and the first three data products (item, location, demand history). Set up a weekly IBP drumbeat with CFO participation.
- Days 90-180: Flip one region or product family to the new operating model: network P&L overlay, control tower authority, and AI-supported re-planning. Publish benefits and hard lessons. Use that proof to phase the rollout.
What good looks like
- Service +2-5 pts with less firefighting
- Inventory turns +1-3; cash conversion cycle down
- WMAPE down 10-20%, forecast bias within ±2 pts
- Expedite and premium freight costs down 20-40%
- Faster decision latency (from days to hours) on allocation, re-planning, and substitutions
The question you can't postpone
The tech will work. The data can be cleaned. The real blocker is whether anyone has both the authority and the incentive to act on what the system recommends.
If the answer is "not yet," fix that first. AI will only make the cracks more visible - and more expensive.
Further reading: McKinsey on operations and IBP fundamentals here. Slalom's perspective on AI adoption and org tension here.
If you're upskilling your leadership team on AI operating models and metrics that matter, see curated options by job at Complete AI Training.
Your membership also unlocks:
