The board approved the AI budget in October. By the following July, the pilot still hadn't started. The vendor wasn't the holdup. They had a proposal ready in three weeks. The holdup was that nobody had confirmed whether the planning team's ERP instance would grant read and write access without a separate IT project, and the person everyone assumed would own the rollout had already moved into a different role by December.
That company isn't an outlier. First Analysis's July 2026 coverage of Sage's 2026 State of Supply Chain Report puts a number on the pattern: 95% of supply chain leaders call AI vital to their company's future success, and only 10% have it live in an actual workflow. An 85-point gap between conviction and execution, in an industry that keeps calling this the year AI finally moves off the roadmap.
That gap holds steady across company after company we've reviewed before a rollout starts. Leadership conviction is close to universal. Deployment is rare. And the 85 points between them aren't sitting where most executives assume.
AI deployment readiness is the set of decisions made before a vendor is selected: whether the ERP already grants the access an AI system needs, who owns each exception category it will surface, whether the sponsor can actually sign the contract, and how narrow the first deployment is scoped. Miss one, and a funded initiative stalls before a vendor gets picked.
From the work
Written from readiness reviews across 5+ manufacturing and distribution engagements, 2025–2026, run before a rollout started, not after one stalled.
Why budget was never the blocker in the supply chain AI deployment gap
Assume the constraint is budget and the data stops making sense. If 95% of leaders already call AI vital, the money conversation happened months ago in nearly every one of these organizations: board decks approved, line items created, a vendor shortlist started in some cases. What stalls the other 90% is a set of decisions nobody put on a meeting agenda: whether the ERP already grants access, who owns the exception category the AI will surface, whether the sponsor can actually sign a contract, and how narrow the first deployment is scoped. None of those four show up on a vendor RFP. All four get discovered the hard way, somewhere between the signed proposal and a go-live date that keeps slipping.
This is a different failure than the one that gets most of the attention. Once AI is live, a separate set of problems takes over: signals that never convert into a governed decision (see the supply chain AI execution gap), or pilots that stall before reaching production (see supply chain AI at scale). Both of those assume the AI is already running somewhere. The 90% in the Sage data haven't gotten that far. Their gap opens earlier, in procurement and ownership decisions that were never made before the vendor call, not in a governance model that was never built after.
The readiness signals that separate the 10% from the 90%
Across the engagements where AI reached production on schedule, the same four conditions were true before the first vendor call happened. Across the ones still circulating a business case a year later, at least two were missing, and it was the same two nearly every time.
| Readiness signal | What the 90% were still figuring out | What the 10% had before the vendor call |
|---|---|---|
| ERP data access | Access request opens after a vendor is picked, becomes a 4 to 6 month IT ticket | Read/write API access already provisioned, no new IT project required |
| Decision ownership | Assigned after go-live, during the first incident the tool should have caught | Named owner per exception category, set before the pilot scope was written |
| Executive sponsor | Champions the idea in a town hall, has no signature authority over the contract | Can approve a $50K+ vendor agreement without a second committee |
| Pilot scope | Framed as a platform evaluation meant to eventually replace planning and procurement together | Scoped to one exception category, for example stockout risk on A-class SKUs only |
ERP data access, decided before the RFP goes out
The single most common reason a funded AI initiative never launches is that nobody checked, before the vendor was picked, whether the ERP would grant the access the AI needs. In Dynamics 365 Business Central, Epicor, or Infor environments, that access request routes through IT as a new project once someone actually asks for it, and a new project means a queue. Organizations that deploy on schedule ask the access question during the readiness review, months before a vendor is shortlisted, not during onboarding after the contract is already signed.
A named decision owner, identified before the pilot starts
The second most common failure looks like success right up until go-live. The AI ships, the signals fire correctly, and then nobody is quite sure whose job it is to act on a stockout risk alert or an overdue PO flag. Ownership gets assigned reactively, during the first incident the tool should have caught, which means the organization spends its first real test of the system arguing about accountability instead of using it. The 10% name an owner per exception category while the pilot scope is still being written, before the AI has surfaced a single alert.
What the readiness gap costs while it sits unresolved
None of this shows up as a missed payment or a broken contract. It shows up as a business case that gets re-presented at the same planning cycle a year later, with the same 95% conviction and the same stalled status. The internal hours spent on a vendor RFP that never closes don't get logged as a loss anywhere. They just don't produce a deployment. Meanwhile, the 10% that cleared these four items compound an advantage every quarter the other 90% spend re-running the same evaluation: faster response on the exception categories AI already covers, a live system generating the deployment data the next expansion decision needs, and a procurement and access process that's already proven it works for the next tool.
How to close the gap before your next vendor call
The order matters. Each of these is a decision your organization can make before a vendor is on the phone, and each one removes a reason the pilot stalls after the contract is signed.
- Confirm ERP read/write access exists before the RFP goes out, not after a vendor is picked. If the access request is still theoretical, the timeline is too.
- Name the decision owner for the first exception category before the pilot scope document is final. If nobody can name that person today, the tool won't create them once it's live.
- Confirm your sponsor's actual signature authority, not just their enthusiasm. A sponsor who can't approve the contract without a second committee isn't a sponsor yet.
- Scope the first deployment to one exception category, not a platform swap. Stockout risk on your A-class SKUs is a pilot. Replacing planning, forecasting, and procurement at once is a program, and it takes the readiness question four times over.
Frequently asked questions
What is the supply chain AI deployment gap?
The supply chain AI deployment gap is the distance between how many supply chain leaders say AI is vital to their company's future and how many actually have it running in a live workflow. Sage's 2026 State of Supply Chain Report puts that distance at 85 points, 95% call AI vital, 10% have it live, and the gap is driven by organizational readiness decisions made or skipped before a vendor is ever selected, not by budget or technology maturity.
Why do 90% of supply chain organizations never get AI live, even with budget approved?
Budget approval answers one question and leaves four others unanswered: whether the ERP grants the access an AI system needs, who owns each exception category it will surface, whether the sponsor can actually sign the vendor contract, and how narrow the first deployment is scoped. Organizations that skip these decisions discover them one at a time after the contract is signed, which is why a funded initiative can stall for a year without anyone officially killing it.
What readiness signals separate the 10% who deployed AI from the 90% who didn't?
Four signals repeat across organizations that reached production on schedule: ERP read/write access provisioned before the vendor search started, a named decision owner per exception category set before the pilot scope was written, an executive sponsor with real signature authority over the contract, and a pilot scoped to one exception category instead of a full platform swap.
How long does it take to close the AI deployment readiness gap before a pilot starts?
Confirming ERP access and naming a decision owner are answerable within a single working session, not a project. The harder gap is scope discipline: keeping the first deployment narrow enough to reach production instead of expanding into a platform evaluation. Organizations that hold that line move from readiness review to a live pilot in weeks, not the multi-quarter cycle that shows up when the same four questions get answered reactively after a vendor is already under contract.
The organizations in the 10% didn't win by moving faster once the AI arrived. They won by answering four questions before it did. That's the finding worth repeating to whoever owns your next AI business case: the vendor call isn't where the deployment gap opens. It's where you find out whether you closed it already.