The order is signed. The announcement has gone into Slack. Sales has handed the account to implementation, and the project plan contains a line called “go-live”. Everyone has a date. Nobody has yet written down the first observable change in the customer’s work.
This is how a commercial yes becomes confused with customer value. The signature is real evidence: a buyer made a commitment under particular terms. It does not prove that the product will help, that the intended users will adopt it, or that the delivery team shares the promise the customer heard.
Those questions do not need a grand outcomes framework. They need an honest sentence early enough to be challenged.
If you cannot say what first value looks like without naming your own delivery milestone, the promise needs another conversation.
Rebuild the promise from more than the contract heading
Imagine a company buying an AI assistant for its service team. This is a composite teaching situation, not a Stratogenic customer story. The proposal promises faster response preparation and more consistent use of approved policy. The contract covers a twelve-week pilot for forty advisers. In the sales calls, one leader also mentioned reducing escalations, although no baseline or mechanism was agreed.
Three versions of success are already present:
- procurement: the pilot starts on time and stays within scope;
- the service leader: advisers prepare good replies with less rework;
- the executive sponsor: escalations fall.
They are connected, but they are not the same outcome. A timely launch can coexist with no useful change. Fewer escalations can reflect demand, policy or staffing rather than the assistant. Less drafting time can create more review time if advisers do not trust the output.
Put each promise into one of four states: contracted, stated by the customer, inferred by the provider, or still unknown. This does not settle a contractual dispute. It makes the differences visible before the project quietly chooses one.
Define first value close to the work
A first-value statement should name the users, change, evidence window and safeguard. For the pilot it might read:
By week six, participating advisers prepare routine policy-based replies with less median handling time than the agreed baseline, while sampled policy accuracy and escalation safeguards remain within the customer-approved boundary.
Every part needs confirmation. Median handling time may be the wrong measure. The sample may need independent review. A safeguard may be more important than speed. That is why the sentence is useful before it is final: the customer can disagree with something concrete.
The Government Digital Service’s standard on defining success and performance data is written for UK public services. Its transferable value here is the separation between delivering activity and showing whether a service solves the intended problem. Applying that principle to a private implementation is editorial judgement, not government endorsement of this measure.
Give the evidence a route through the handover
Once first value is clear enough, the handover changes shape. Sales carries the source of the promise and any unresolved expectation. Implementation carries configuration, readiness and delivery evidence. Customer Success carries continuity of the measure. The customer names the people who can confirm whether the work has changed.
For the assistant, that could mean:
- Agree the routine enquiry types inside the pilot.
- Measure current preparation and review effort for a small baseline sample.
- Define policy-accuracy review and cases that must never be automated.
- Record adviser corrections and reasons, not only acceptance.
- Review the first-value statement with advisers after two weeks.
NIST’s AI Risk Management Framework supports attention to context, testing, monitoring and ownership across the life cycle. It is voluntary guidance and is being revised. It does not prescribe this customer measure or certify the pilot.
Leave room for value to be disproved
Teams sometimes write outcomes that can only succeed. “Improve the adviser experience” can be claimed after any positive comment. “Drive efficiency” can absorb a spreadsheet of selected savings. A useful measure has a failure condition.
If preparation time falls but total checking time rises, the efficiency claim has not landed. If advisers avoid complex cases to protect the metric, the boundary needs attention. If the customer cannot supply the baseline, the team should label the claim unmeasured rather than reconstructing it from memory.
From Promise to First Value came from seeing the commercial promise, delivery boundary and evidence owner drift apart. The package helps reconstruct that decision and handover record. It cannot guarantee adoption or manufacture an outcome the customer has not confirmed.
Ask one sentence three times
Take one newly signed account. Ask sales, delivery and the customer sponsor separately to finish this sentence: “The first useful change will be visible when…” Compare the answers without blending them.
Choose the difference most likely to damage trust. Give it an owner, an evidence source and a date before the next irreversible delivery step.
Sources and limits
- GOV.UK Service Standard: define success and publish performance data — used for the distinction between delivery activity and evidence of solving the intended user problem. It governs UK public services; transfer to private delivery is editorial judgement.
- NIST: Artificial Intelligence Risk Management Framework — used for context, evaluation, monitoring and ownership around the illustrative AI workflow. It is voluntary US guidance under revision, not certification, legal advice or proof of safety.
Sources reviewed 7 September 2026. The account is explicitly composite and must not be represented as customer evidence.