Most organisations decide what a supplier is called before they decide what it owes. The word strategic arrives in a steering deck, gets repeated in a governance charter, and by the time anyone tests it against the contract it has been true for two years by assertion.
I led a post-mortem review of a failed offshore software build at a European banking group's IT services subsidiary. The build had been taken back in house before user acceptance testing, the statement of work had terminated, and the review answered a set of client questions across delivery approach, project management and the software itself, then filed recommendations against each. The very first of those recommendations, numbered R-1.1 and filed under delivery approach, is the one I keep coming back to, because it does something the rest of the register does not. It treats the label as a purchase.
The sentence that makes status conditional
The review opens its recommendations by classifying what the arrangement actually was rather than what it was called. That classification, and the client-side control it justified, is already in print in How to Screen an Offshore Supplier's Developers Before the Build Begins, so I will not relitigate it here. What matters for this piece is the sentence that follows it.
If the supplier wishes to be a strategic partner, the review says, it must take end-to-end accountability on roles, responsibilities and strong in-house skills.
Read the grammar. There is a conditional clause, a modal obligation and a list. The status is not something the review recommends granting, withholding or reviewing annually. It is an outcome of a transaction, and the transaction is stated in a single line with the consideration on the supplier's side of it. A supplier that has not taken end-to-end accountability is not a strategic partner that is underperforming. It is a supplier that has not bought the title.
That is a more useful construction than it looks, because it inverts the direction most sourcing conversations run in. The usual sequence is to award the relationship a tier, then attempt to derive obligations from the tier, and then discover in year two that nothing in the signed documents actually changed when the tier did. Here the obligations are the definition. There is no tier to argue about separately.
Why the third item costs more than the first two
The list has three parts and they are not the same kind of thing, which is easy to miss when they sit in one line.
Roles is an org chart question. Responsibilities is a matrix question. Both can be satisfied inside a fortnight by any supplier with a delivery methodology and a competent bid team, and both produce artefacts a client can file and feel reassured by. I have watched several such artefacts get produced. They are usually accurate on the day they are signed.
Strong in-house skills is a different order of commitment. It says the capability has to sit on the supplier's own books. Not borrowed from a partner network for the duration of a bid, not assembled from whoever cleared a previous engagement that month, not subcontracted at a margin. In-house means hired, retained and carried through the gaps between projects, which is precisely the cost that a project-by-project commercial model is designed to avoid paying.
Roles and responsibilities can be produced in a week from a template. Strong in-house skills is a hiring and retention commitment, and it is the only one of the three a supplier cannot satisfy by rearranging a document.
So the recommendation, in three words at the end of a list, asks the supplier to absorb the standing cost of capability. That is the real content of end-to-end accountability, and it explains why the upgrade is rarely taken up when it is offered honestly. A supplier can accept accountability for outcomes it does not have the standing skill to produce, and many do, but only until the first difficult delivery. The review put the skills clause in the same breath as the accountability clause because one is unenforceable without the other.
For a buyer, this turns a vague ambition into a question with an evidential answer. Ask which of the capabilities you depend on are employed by the supplier, on permanent terms, and retained between your engagements. The answer is checkable, and it is the same answer whether the capability in question is a Java architect or an evaluation engineer.
The client's half of the bill
The part of R-1.1 that gets overlooked is that the client pays too, and the payment is specified rather than gestured at.
The recommendation calls for a comprehensive, advanced intercultural programme, and it names three outcomes: an understanding of the role cultural dimensions play in international teams, improved communication and cooperation skills, and a set of common rules for efficient international teamwork. Three outcomes, not a workshop. The third one in particular is an artefact somebody has to write and both sides have to work to.
Two things about that are worth carrying into your own arrangements.
The first is that the review put culture in the recommendation register at all, next to accountability and contracting, rather than in an annex about ways of working. Treated that way, it becomes a delivery control with a named deliverable, and it can be tracked like one.
The second is where the ownership landed. The programme is assigned to the subsidiary's sourcing function, not to the delivery managers who were living the friction and not to a human resources team. Sourcing writes the relationship, so sourcing carries the obligation to make the relationship operable across two working cultures. That placement is a quiet argument that the capability to run an international supplier relationship is part of the sourcing function's job description, not an overhead the delivery teams absorb.
- R-1.1
- The first recommendation in the register, filed under delivery approach
- 3
- Components of the accountability the supplier was asked to accept
- 3
- Named outcomes required of the client's intercultural programme
- Recommendation
- Status of both halves, written after the statement of work had ended
The square brackets are the mechanism
Every recommendation in the register is stamped the same way: a number, a theme, and in square brackets the party accountable for acting on it. R-1.1 carries two owners because it contains two obligations, one on each side.
I have come to think that tag is the most transferable thing in the entire document. A review that produces findings without owners produces reading material. The register format forces the author to answer, for each recommendation, who is going to be asked about this in ninety days, and it makes the answer visible on the page where the recommendation lives rather than in a separate action log that diverges from it within a month.
It also makes asymmetry impossible to hide. A register in which every owner tag names the supplier is telling you something about the review, and a register in which every tag names the client is telling you something else. R-1.1 splits, which is the honest shape for a relationship problem.
The same test, with a newer kind of supplier
The reason I am writing this up now rather than filing it as an artefact of an older sourcing era is that the label problem has returned in a more expensive form. Organisations are currently appointing AI partners, and the appointment usually precedes any statement of what the partner owns when a model degrades, when a retrieval layer starts returning stale documents, or when an agent takes an action nobody can reconstruct.
Run R-1.1 against that. Does the supplier hold end-to-end accountability for the outcome, or for the increment it bid for. Are the roles and responsibilities written down, and are they written at the level of the system rather than the sprint. And the expensive question: are the skills in house. Data engineering, evaluation design, model monitoring and incident forensics are exactly the capabilities a delivery firm is tempted to rent for the duration of a proposal, and exactly the ones you need standing when something goes quietly wrong eight months in.
What it produced
Nothing, on this engagement. The statement of work had terminated months before the review reported. There was no upgrade left to grant, no joint delivery for the accountability clause to govern and no working relationship for the intercultural programme to improve. Both halves of R-1.1 are recommendations addressed to a relationship that had already stopped.
What survives is the construction. A supplier tier is worth exactly the obligations you can point at, and the moment to establish those is before the word goes into the governance deck, because afterwards it defends itself.
Drawn from the first recommendation page of an independent post-mortem review of a failed offshore software build at a European banking group's IT services subsidiary, which I led. R-1.1 was written and delivered after the statement of work had already been terminated: the review produced findings and recommendations, not delivered results. The wording of the accountability clause, the three named programme outcomes and the owner tags are as recorded in it. Reading the in-house skills clause as the expensive part, and the owner tag as the register's enforcement device, is ours.
“Roles and responsibilities can be produced in a week from a template. Strong in-house skills is a hiring and retention commitment, and it is the only one of the three a supplier cannot satisfy by rearranging a document.”
Get in touch
Put RealAI’s applied-AI team on your hardest data problem.
We help enterprises move from pilots to production: sovereign models, governed data, and agents you can audit. Start with a value-first assessment.