By the time the renewal terms arrive, the negotiating position has already been set. The Index reads it in advance.
Episode 7 made the argument: vendor dependency is a capital-architecture commitment that forms quietly between signing and renewal. Every integration, every dataset, every workflow rerouted through the vendor adds switching cost that no one is tracking. By the renewal cycle, the accumulated total is the leverage you do — or do not — have.
This issue turns that argument into an instrument. Six dimensions, each scored Low, Moderate, or High exposure. Run your top AI vendor relationships through it and the dependency profile tells you where the risk sits and when to act on it.
The discipline the Index enforces is simple: price the exit option at entry, not at renewal.
HOW TO SCORE
Pick one AI vendor relationship — the one carrying the most weight in your portfolio. Score it across all six dimensions below. Each dimension resolves to one of three states.
Low exposure — the dependency is shallow, reversible, or already allocated. Moderate — the dependency is real and growing but still manageable with deliberate action. High — the dependency is deep, compounding, or unallocated, and the exit option is narrowing.
Dimensions one through three measure how the dependency forms. Dimension four quantifies it. Dimension five asks whether the contract anticipated it. Dimension six is the result — the negotiating position the other five produce. Read the profile top to bottom: the decision, the accumulation, the cost, the contract, the constraint.
The profile is the diagnostic — not any single score. A relationship that scores High on integration and data but Low on contractual allocation is a different problem from one that scores Moderate across the board. Where the High exposures cluster is where the renewal-cycle risk concentrates.
DIMENSION 1. Integration Depth
Core question: How deeply is the vendor's system wired into your core workflows — and how many processes would break if you removed it?
Low: The vendor sits at the edge of one or two non-critical workflows. Removal is disruptive but contained.
Moderate: The vendor is embedded in several workflows, including at least one the business depends on daily.
High: The vendor is wired into multiple core workflows. Removing it would disrupt operations across the organization.
What this returns: Integration depth is the first source of switching cost. Shallow integration is reversible; deep integration is a structural commitment whether or not it was approved as one. This dimension rises on its own as the vendor is wired into more processes — without a new decision being made.
DIMENSION 2. Data Accumulation & Portability
Core question: How much organizational data now lives inside the vendor's environment — and can you extract it in a usable form?
Low: Little data is held by the vendor, or the data exports cleanly in a portable, tested format.
Moderate: A meaningful volume of data sits with the vendor; export is possible but untested or partial.
High: Substantial data lives in the vendor environment with no tested export path or no portable format.
What this returns: Data that cannot be moved is a dependency that deepens automatically with time. This dimension compounds passively: every month of accumulated data raises the exposure, independent of any new decision. A tested export path is the single most effective way to hold this dimension down.
DIMENSION 3. Model & Platform Dependency
Core question: Is your AI capability tied to a specific vendor's model or platform that cannot be substituted without re-engineering?
Low: The capability is model-agnostic or built on a substitutable standard; switching models is a configuration change.
Moderate: Some capability is tuned to the vendor's platform; substitution is possible but requires meaningful rework.
High: Core capability is tuned to the vendor's specific model or platform and cannot be substituted without re-engineering.
What this returns: When capability is tied to a vendor's model, that vendor's roadmap, pricing, and deprecation decisions become your constraints. Model dependency converts the vendor's strategic choices into your operating risk — a dependency that is invisible until the vendor changes something.
DIMENSION 4. Switching Cost
Core question: What is the full cost of exit — re-integration, data migration, retraining, re-engineering, and operational disruption — expressed as an estimable number?
Low: Exit is straightforward and the cost is known and modest.
Moderate: Exit is costly and disruptive but has been estimated and is survivable.
High: Exit cost is large and — the more common case — has never been estimated at all.
What this returns: Switching cost is the dimension the episode argues should be sized at entry and is usually discovered at renewal. The gap between those two moments is where leverage is lost. An unestimated switching cost is not a low cost — it is an unmeasured one, which at renewal tends to resolve as high.
DIMENSION 5. Contractual Risk Allocation
Core question: Does the contract assign responsibility for incorrect outputs, regulatory non-compliance, and substantial-modification risk — and do those allocations survive renewal?
Low: The contract explicitly allocates the major failure modes, and the allocations are durable through renewal.
Moderate: The contract addresses some failure modes but is silent on others, or the allocations reset at renewal.
High: The contract is silent on the major AI failure modes; risk is unallocated by default.
What this returns: Silence in the contract is not neutrality — it is unallocated risk sitting with whoever is closest to the system when something goes wrong. EU AI Act obligations and substantial-modification exposure under Article 25 are a 2026 pressure point, though the implementation details continue to harden. The durable question is whether the contract creates a record of who carries the risk.
DIMENSION 6. Renewal-Cycle Leverage
Core question: Given the five dimensions above, what is your actual negotiating position when this contract renews?
Low: Shallow dependency, known switching cost, allocated risk. You can credibly walk, and the vendor knows it.
Moderate: Real dependency partially offset by a known exit path or allocated contract terms. Some leverage remains.
High: Deep dependency, high or unestimated switching cost, unallocated risk. The renewal is effectively a price-taker position.
What this returns: This dimension is not an input — it is the result the other five produce. It is read, not scored independently: high dependency plus high switching cost plus unallocated risk equals weak leverage. The value of the Index is that it surfaces this position with enough lead time to change it before the renewal email arrives.
APPLIED EXAMPLE
Consider a vendor-built customer-service AI platform. It is integrated across CRM, ticket routing, the knowledge base, and quality monitoring. Two years of conversation data have accumulated inside the vendor environment. The contract renews in Q4. Scoring it across the six dimensions:
Integration Depth: wired into four core workflows. Removing it disrupts each one. → HIGH.
Data Accumulation & Portability: two years of conversation data in the vendor environment, with no tested export path. → HIGH.
Model & Platform Dependency: response behavior and escalation logic are tuned to the vendor's model and would need re-engineering elsewhere. → HIGH.
Switching Cost: re-integration across four workflows, plus data migration, plus retraining — never estimated at signing. → HIGH.
Contractual Risk Allocation: the contract is silent on incorrect-output responsibility and substantial-modification risk, and AI-specific terms reset to vendor standard language at renewal. → UNALLOCATED.
Renewal-Cycle Leverage: the five dimensions above set the position. → WEAK.
Diagnosis Five dimensions High, one Unallocated, leverage Weak. This is not a marginal dependency — it is a structural commitment that was approved two years ago as a capability purchase. The profile concentrates risk in two places: the data that cannot be moved and the contract that never allocated the failure modes. The Q4 renewal is the action point. AI-specific clauses for incorrect-output responsibility and substantial-modification risk, plus a switching-cost estimate, have to be drafted before the renegotiation opens — because once the renewal email arrives, the leverage to negotiate them is already gone. The relationship may still be the right one to keep. The point is to keep it as a decision, not as a drift. |
THREE QUESTIONS TO ASK MONDAY
Run your single highest-spend AI vendor relationship through all six dimensions. Where do the High exposures cluster — in how the dependency formed, in its cost, or in the contract that failed to allocate it?
The cluster location is the diagnostic. Highs in dimensions one through three mean the dependency formed faster than anyone tracked. A High in dimension five means the contract is carrying unallocated risk into renewal.
For that relationship, has anyone ever estimated the switching cost — the full cost of exit, not the cost of staying? If not, that number is the missing input to every renewal conversation you will have.
Switching cost is estimable before signing and discoverable at renewal. The gap between those two moments is where leverage is lost. Sizing it now, mid-contract, is the second-best time.
When does that contract next renew — and are the AI-specific clauses being drafted now, or will they be negotiated under time pressure when the renewal terms land?
Renewal is the natural point to reallocate pass-through risk. If the clauses are not drafted before the renegotiation opens, the risk stays where it sits today — with whoever is closest to the system when something goes wrong.
WHAT’S NEXT
The Index measures one form of a larger pattern. An organization makes a decision, a constraint forms quietly around it, and by the time the risk becomes visible, the exit options have already narrowed. Vendor dependency is the version that shows up in contracts and renewal cycles.
But the same pattern runs through governance structures, board decisions, platform choices, and strategic commitments — anywhere a decision is made before the decision architecture around it is fully understood. That is the layer the channel opens next.
Watch Episode 7
Episode 7 walks through the dependency lifecycle and all six dimensions of the Index. If this instrument surfaced a high-exposure relationship in your portfolio, the most useful next step is to bring the profile — and the renewal date — to whoever owns that vendor contract.