Practical guide
How to evaluate a partnership with AI
Evaluate a partnership against predeclared strategic, economic, operating, evidence, and exit criteria. AI can normalize proposals and expose responsibility gaps, but brand halo, friendly conversations, or a large audience should not substitute for an operating commitment. The output should make ownership, dependencies, downside, and test conditions explicit.
Evaluate a partnership against predeclared strategic, economic, operating, evidence, and exit criteria. AI can normalize proposals and expose responsibility gaps, but brand halo, friendly conversations, or a large audience should not substitute for an operating commitment. The output should make ownership, dependencies, downside, and test conditions explicit.
Define the partnership decision
State the objective, customer or beneficiary need, scope, alternatives, decision owner, time horizon, and non-negotiable boundaries. Compare partnership with buying, building, referring, or doing nothing where those are real options.
UK government guidance for school partnerships recommends agreeing aims, project leads, activities, performance measures, evaluation, funding, and a written agreement. The domain is specific, but the operating questions are broadly useful. Charity Commission guidance similarly emphasizes clear responsibilities, independence, risk, and formal agreements within its scope.
Freeze criteria before partner scoring
Use criteria such as strategic fit, evidence of need, contribution, economics, delivery capacity, governance, data and intellectual-property boundaries, reputational risk, learning value, and reversibility. Define what green, amber, and red mean before reviewing a specific partner.
Worked synthetic assessment
Suppose Two Prune is considering a training-provider pilot. The provider would recruit a 20-person learning cohort; Two Prune would configure a synthetic assessment and provide task-bounded reports for human interpretation.
| Criterion | Evidence | Status | Required commitment | | --- | --- | --- | --- | | Cohort readiness | Draft course calendar | Amber | Named cohort and date | | Evidence need | Founder interview notes | Amber | Programme owner confirms decision use | | Delivery capacity | Named facilitators | Green | RACI and rehearsal | | Data boundary | Synthetic-task proposal | Green | Approved handling plan | | Economics | No agreed budget | Red | Cost and payment owner | | Evaluation | General interest only | Red | Questions, measures, and review date |
The table is synthetic and does not describe an approved customer or actual partnership.
Build an operating charter
For each workstream identify accountable and responsible parties, inputs, deliverables, acceptance evidence, timing, escalation, and decision rights. Cover participant communications, consent, support, data handling, assessment interpretation, incident response, and final evaluation where relevant.
AI can inspect the charter for tasks assigned to both parties or neither party. It can also compare the economics narrative with the resource plan. Owners confirm every commitment.
Test economics and downside
Model cash costs, staff time, opportunity cost, dependencies, and who bears overruns. Run low-enrollment, delayed-launch, low-completion, and early-exit cases. Do not count access to an audience as revenue or a signed memorandum as demand.
Design a bounded pilot
Convert unknowns into testable questions. For the synthetic partnership: Can the provider recruit an appropriate cohort by the agreed date? Can both teams operate the synthetic-data boundary? Do programme owners find the evidence useful enough to inform a development discussion? Define evidence, owner, stop conditions, and review date for each.
Make exit possible
Specify term, termination rights, unfinished work, data return or deletion, participant communication, intellectual property, outstanding payments, and permission for any public reference. A partnership that cannot end cleanly carries hidden cost.
Verification checklist
Confirm criteria predate scoring; claimed assets have evidence; responsibilities have one accountable owner; economics reconcile; data and claim boundaries are approved; pilot questions are observable; conflicts and escalation are addressed; exit terms are workable; and no public proof is implied without permission.
Sources and claim map
The organizations, cohort, statuses, and commitments are synthetic. This resource does not claim partner demand, customer proof, commercial return, or a successful pilot.
- UK Department for Education, Guide to setting up partnerships: supports agreeing aims, leads, activities, measures, evaluation, funding, and formal arrangements in its stated domain.
- Charity Commission for England and Wales, Work with other charities: supports clear responsibilities, independence, risk assessment, agreements, and exit considerations for covered charities.
Sources: [1] [2]
Sources
- 1.Guide to setting up partnerships · UK Government
- 2.Work with other charities · UK Government