RA LAB · Referrals · AI · 45 minutes + a 7-day test
Customers are satisfied but rarely recommend us
AI can review de-identified feedback to find repeated reasons to recommend. Do not generate fake reviews or request someone else’s contact details without consent.
RA LAB practical guide · based on the library’s GRI / GTM sources.
GOAL
Find a proven moment of value and test a convenient referral path.
VALUABLE FINAL DELIVERABLE
A referral-trigger map, invitation flow and test of new lead quality.
Separate satisfaction from willingness to recommend. Identify what the customer found useful, who might benefit and when they are comfortable sharing the experience. Do not present AI interpretations as customer testimony.
For an initial test, use voluntary feedback and de-identified excerpts you are entitled to analyse. Compare every finding with the original statement.
AI helps with timing and explanation
AI can group recurring reasons to recommend and draft short invitations. A person checks accuracy, tone, language and the absence of pressure.
Let a customer share a link themselves. Do not request address-book uploads, a friend’s phone number or reviews the person did not write. Any reward and conditions should be clear in advance.
Example referral map
Illustration. Value received → voluntary invitation → link visit → new customer’s own enquiry → task fit → purchase. The number of shared links does not replace enquiry quality.
Revise the flow if participants misunderstand terms, complaints increase or enquiries do not fit the offer. Distinguish a customer-confirmed referral source from an assumption based on the last click.
Common questions
Can AI write customer reviews?
Do not fabricate reviews. AI can analyse permitted feedback; a public quote should preserve a real statement’s meaning and have a valid basis for publication.
How do we know referrals work?
Compare confirmed enquiries, quality, purchases and costs over a defined period. With a small sample, record facts and limitations rather than declaring a causal effect.
AI helps identify patterns and generate hypotheses. Verify outputs against source evidence, use de-identified data and keep a person accountable for decisions.
Review de-identified feedback and repeat-purchase facts. AI groups themes; a person verifies actual value.
What did the customer achieve, when, and what supports it?
Step deliverable: What did the customer achieve, when, and what supports it?
2. Draft invitation
Invite a voluntary recommendation after a successful outcome. Avoid hidden contacts and generated testimonials.
Trigger, invitation wording, consent and opt-out.
Step deliverable: Trigger, invitation wording, consent and opt-out.
3. Test referral path
Check whether a customer can share a link with little friction. Attribute enquiries without storing third-party personal data without a lawful basis.
Path from invitation to enquiry and lead-quality metric.
Step deliverable: Path from invitation to enquiry and lead-quality metric.
4. Evaluate impact
Compare qualified enquiries and purchases, alongside complaints and opt-outs.
Result, limitations and improvement to the referral flow.
Step deliverable: Result, limitations and improvement to the referral flow.
How to judge whether it helped
You should leave with a document that supports a decision and a next action. Completing steps does not guarantee a commercial result. Measure the chosen change separately over an agreed period. If evidence is missing, state that and plan how to collect it.
A baseline, actual result and data limitations are recorded.
A person made the decision after reviewing quality and side effects.
How this differs from S01–S13
This is a user route for a specific problem. The GRI AI research stages S01–S13 and the 13-week programme remain in the library as separate methodological work. This route is not a validated GRI diagnostic scale.
Your route0 / 4 · 0%
Completed 0 · remaining 4
Next action: Find value momentThe percentage measures completed steps, not proven business growth.
PRACTICE / 1
Find value moment
Review de-identified feedback and repeat-purchase facts. AI groups themes; a person verifies actual value.
Step deliverable: What did the customer achieve, when, and what supports it?
No names, contacts, client secrets or sensitive details. Saving happens only when you press the button.
Discuss opens WhatsApp. Your answers are not sent automatically.
Data & privacy
Matching works without saving your query. Practice answers remain in this tab until you enable saving. Do not enter names, contacts, client secrets or sensitive information.
Server saving: answers and steps are linked to a random key in a secure browser cookie for up to 90 days. This is pseudonymous, not anonymous data. It does not sync across devices.
Server saving is not enabled yet: the data controller and hosting configuration are being confirmed. You can work in this tab and download your results.
