The Ownership Test: Score a Digital-Workforce Deal Before You Sign

Manuel Conde

17 July 2026 · Capability Ownership

Flat geometric illustration of a proposed deal at a fork: a beige contract document at the base sends a single black path upward that splits in two. The left branch leads to a solid black panel with a lime core held inside it, grounded on a short teal base — capability owned inside. The right branch leads to a hollow black panel outline with a lime disc floating beside it rather than within, marked by a small red risk ring and hovering over a dashed line it never touches — dependency rented beside. Teal and beige circles, squares and bars surround the two outcomes.

Exploring this for your team? Talk to CrmWare →

The dependency you can't yet read

The argument this test operationalises is that owning your digital workforce matters. This is how you read that verdict off a deal before you commit to it.

Most buyers cannot yet read it. Even IBM's own commissioned research — from a vendor that sells the multi-vendor architecture the finding recommends — surveyed 1,000 senior executives across 16 countries and found that 91% do not fully understand their dependencies across AI vendors, models and infrastructure, 71% say switching their primary vendor would be difficult, and 72% would accept a 20% cost increase to maintain AI vendors if it improved strategic flexibility.[1] When even the seller's own numbers show the gap, the gap is not in doubt.

And what breaks next is decided here, not in the model. Gartner predicts over 40% of agentic AI projects will be cancelled by the end of 2027 — for escalating costs, unclear value and inadequate controls, every one a governance failure the deal structure decides, not a model that misbehaved.[2]

Ownership is a property of the deal, not the vendor

Control does not live in a vendor's reputation. It lives in the contract terms — which means the own-vs-rent verdict is legible from the structure of an arrangement before anyone signs it. That is an argument, not a lifted fact, and the strongest evidence for it is that regulators bother to legislate exactly these terms.

The European Banking Authority is explicit that outsourcing a function cannot delegate the management body's responsibility; the institution remains fully accountable, and must be able to exit an arrangement without undue disruption.[3] Accountability and exit are treated as contractable properties, not vendor courtesies. The EU went further and legislated switching itself: under the Data Act, egress charges between data-processing providers are removed outright from January 2027, because lock-in proved structural enough to require law.[4]

If control is written into terms, then the deal decides — not the counterparty. A large-firm arrangement can pass this test and a boutique one can fail it. The variable is never the logo. It is the structure.

The six tests: score the deal

Run six tests on the proposed arrangement. Score each one Rent (0), Contestable (1), or Own (2), sum them, and read the band — Own, Contestable or Rent — off a 0–12 scale. Do it before signing, not in the post-mortem.

A six-row scoring matrix titled the own-vs-rent scorecard. The rows are six deal dimensions — stop authority, exit cost, accountability, drift control, retention, and transparency — and the three columns are scores: rent 0 (red), contestable 1 (teal), and own 2 (lime). A worked example deal is plotted with one cell marked per row: stop authority, drift control and transparency each score rent 0; exit cost and retention each score contestable 1; accountability scores own 2. A score column on the right reads 0, 1, 2, 0, 1, 0. Below the matrix a verdict band splits the 0–12 total into rent 0–4, contestable 5–8 and own 9–12, with a pointer showing this deal's total of 4 of 12 landing inside the red rent zone. — Figure 1. The own-vs-rent scorecard — six deal dimensions scored 0–2 and summed to a pre-signing verdict; this worked deal scores 4 of 12 and lands in Rent.

Stop authority. Does the contract give you a unilateral halt or override you can exercise without the vendor? The EU AI Act requires a stop function and the power to disregard or reverse a high-risk system's output[5]; NIST calls for mechanisms to supersede or deactivate one.[6] Rent if only the vendor can stop it.
Exit cost. Are switching, portability and data egress priced and contracted, or is departure left undefined?[4] The EBA expects exit without undue disruption.[3] Rent if leaving means a rebuild.
Accountability. Does accountability stay non-delegably yours, with the evidence to discharge it?[3] A deal that pretends to transfer accountability is renting.
Drift control. Do you get drift-detection, evaluation authority and the right to supersede on failure — or is the vendor the only party who can judge the system?[6] Rent if you cannot measure it or turn it off.
Retention. Does the structure transfer runbooks, decision logic and skills so your own team can operate it?[7] MIT's NANDA study ties the enterprise returns gap to exactly this learning gap — tools that do not land inside the organisation.[8] Rent if the capability never lands inside.
Transparency. Can you read the full stack of vendors, models and infrastructure the arrangement commits you to before signing?[1] Rent if the dependency is opaque at signature.

Weight the dimensions that can stop your business hardest — selective sovereignty, not total independence[1] — and let the sum give you the verdict.

Before you sign, require the transfer

Where the test reads rent, one move changes the verdict: require an ownership-transfer clause. This is CrmWare's own dependency-audit method, not a procurement rule handed down from outside — a contracted commitment that runbooks, decision logic, evaluation methods and operating skills transfer to your team, so the capability lands inside the organisation rather than beside it. The UK National Audit Office reached the same place from the public-sector side: the remedy for rented expertise is to retain and upskill your own people to reduce reliance on the supplier.[7]

That is what ownership means in practice — and what it means to own your digital workforce is the definition this test scores a deal against. Rent the technology where it makes sense. Require the capability.

References

[1] IBM Institute for Business Value (with Oxford Economics). IBM Study: Limited control and rising dependencies leave enterprises exposed in the age of AI ("The calculus of AI sovereignty"). 2026. https://newsroom.ibm.com/2026-06-17-ibm-study-limited-control-and-rising-dependencies-leave-enterprises-exposed-in-the-age-of-ai

[2] Gartner. Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027. 2025. https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027

[3] European Banking Authority. Guidelines on outsourcing arrangements (EBA/GL/2019/02). 2019. https://www.eba.europa.eu/sites/default/files/documents/10180/2551996/38c80601-f5d7-4855-8ba3-702423665479/EBA%20revised%20Guidelines%20on%20outsourcing%20arrangements.pdf

[4] European Commission. Data Act explained. 2024. https://digital-strategy.ec.europa.eu/en/factpages/data-act-explained

[5] European Commission. AI Act Service Desk — Article 14: Human oversight. 2024. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-14

[6] NIST. AI Risk Management Framework 1.0 — Core (AI Resource Center). 2023. https://airc.nist.gov/airmf-resources/airmf/5-sec-core/

[7] UK National Audit Office. Lessons learned: Government's use of external consultants. 2025. https://www.nao.org.uk/wp-content/uploads/2025/11/governments-use-of-external-consultants.pdf

[8] MIT Project NANDA. The GenAI Divide: State of AI in Business 2025 (preprint v0.1). 2025. https://web.archive.org/web/20260101051157/https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf

Talk to CrmWare