Operations
Part of who should sign off AI marketing tools output?
Set AI marketing tools service standards and score suppliers against a rubric
Write service standards for your AI marketing tools supplier, then score accuracy, turnaround, data handling, accountability and exit on one rubric.
What to take away
- The decision in front of you is keep, renegotiate or replace, and all three need the same evidence: a written standard and a score against it.
- Each clause needs a number, a named owner and somewhere to file the proof.
- The ICO direct marketing checklist and government AI ethics guidance set the floor a service standard must clear.
- Score suppliers on one rubric at selection and at renewal, so continuity does not depend on a single account manager.
Define the standard before you compare suppliers
Marketing tools service standards only work if they exist before the shortlist closes. Once suppliers pitch, conversation drifts to platform features that are hard to hold anyone to.
Steps to set service standards
- Define standards before shortlist closes
- Describe inspectable outputs, not tool features
- Set two or three thresholds and hold them
- Record the date each threshold was agreed
- Get sign-off from marketing and data owners
The wider delivery picture is set out in our guide to AI marketing tools operations and delivery. Those standards sit inside a working cycle. Two people should sign the finished document: the marketing lead who owns delivery and someone outside marketing who owns data.
Describe the output, not the product
Write each clause around something you can inspect: a campaign brief returned within two working days, ad copy checked by a named editor, a monthly spend reconciliation. Tool features belong in a separate column. A supplier can change its model without breaking a promise about a checked output.
Set thresholds a supplier can meet
Choose two or three thresholds and hold them. For example, a team paying £1,500 a month might require 95% of ad copy to pass brand review first time. The same standard can set a reply to any data query within one working day.
Record the date each threshold was agreed. A target carried over from an old contract rarely survives its first difficult month, and tighter targets need a reason you can defend at renewal.
Audit the standard against regulation and ethics
A standard that ignores data protection or ethics will struggle in an audit, however neatly it is written. The test is whether you could hand the file to an auditor without a scramble. Build both into the monthly routine rather than bolting them on before a review.
Monthly compliance audit checklist
- Who reviews the model's output?
- How can a customer challenge a decision?
- Check consent records
- Check suppression lists
- Check record keeping
- File results beside service scores
Treat AI ethics guidance as a design input
Government guidance on understanding artificial intelligence ethics and safety asks organisations to weigh fairness and accountability before deployment. Turn that into two questions for a supplier: who reviews the model's output, and how can a customer challenge a decision made with it?
Check the campaign trail against the ICO checklist
The ICO direct marketing checklist gives you a repeatable audit of consent, suppression and record keeping. Run it monthly and file the result beside the service scores, so a compliance question is answered from a folder you already keep.
Score the supplier at selection and at renewal
A rubric makes two suppliers comparable on the same terms. It also makes a renewal decision something you can explain to finance without rewriting the story from memory. Weight the clauses before anyone scores: accuracy and data handling should carry more than presentation, otherwise a tidy weekly report outranks a clean consent record.
Build the skills that keep a standard alive
Standards fade when nobody can interrogate the output. Data and insights training, such as the courses offered by the Chartered Institute of Marketing, helps an analyst question a model's summary rather than pass it on. Fund it from the same budget line as the subscription.
Use one rubric for every supplier
Score each supplier out of five on the clauses below and record the evidence next to each score. A supplier that looks cheap but scores badly on audit trail is a risk, not a bargain. The AI marketing tools operating workflow in England shows where those scores feed into weekly delivery and sign-off.
| Clause | What a strong score looks like | Evidence to request |
|---|---|---|
| Output accuracy | A named editor signs off before publication | Sample of checked outputs |
| Turnaround | Brief returned within two working days | Timestamped delivery log |
| Data handling | Consent and suppression applied per campaign | Completed ICO checklist |
| Accountability | A named contact answers escalations within a day | Escalation record |
| Exit | Campaign data returned in a usable format | Written exit plan with dates |
Common questions
Should service standards sit in the contract?
Yes. A schedule attached to the contract turns a good intention into something enforceable. Keep the wording short and the measures specific.
How many clauses should we track?
Five or six is workable. A longer list tends to be reviewed once and then abandoned, which is worse than a shorter list you maintain.
Who owns the standard?
Name one person, usually the marketing operations lead, plus a deputy. Shared ownership between marketing and data teams often means nobody chases the evidence.
What if a supplier refuses to be scored?
Treat the refusal as a finding. Ask what they would measure instead and record the answer. A supplier unwilling to discuss evidence has told you something useful.
