How to prepare a renewal risk plan with AI

Start with the renewal decision

Record the agreement end date, notice period, contracted scope, customer objectives, renewal owner, and next decision milestone. Then define what the team needs to learn: whether the customer still has the problem, whether the product is used in the relevant workflow, whether credible outcome evidence exists, and whether stakeholders can support continuation.

Do not begin with a red-amber-green score. A color compresses information before the team has established what it means.

Assemble an evidence board

Use five lanes:

| Lane | Evidence | Important distinction | | --- | --- | --- | | Commercial | scope, terms, renewal date | Contract fact versus forecast | | Adoption | active users, workflow events, recency | Access versus meaningful use | | Outcomes | agreed objectives and observed evidence | Observation versus attribution | | Stakeholders | roles, meetings, decisions, changes | Known view versus inferred sentiment | | Blockers | defects, missing capability, process constraints | Symptom versus root cause |

The GOV.UK Service Manual recommends defining metrics from service purpose and finding appropriate data sources. It also combines performance data with user research. The transferable point is that a count should connect to a user outcome and decision; the guidance does not establish a commercial renewal model.

Worked synthetic account

Suppose a customer’s agreement ends in 90 days. Twenty administrators can access the workspace, six created assessments last month, and two teams completed the intended workflow. A sponsor changed roles. Support records show repeated questions about template setup. The original objective was to make evidence-backed AI-fluency assessments easier to run.

A weak AI analysis says, “Low usage and sponsor departure mean renewal probability is 35%.” There is no validated model behind that number, and account access alone does not reveal value.

A better risk statement is:

Observed: Two of five intended teams completed the workflow in the last 30 days; setup questions recurred; the named sponsor changed roles.

Interpretation: Adoption may be constrained by setup friction and stakeholder continuity.

Unknown: Whether the remaining teams still plan to run assessments, whether completed assessments met the customer’s evidence needs, and who now owns the program.

Decision implication: Do not forecast continuation until the new owner and remaining teams are understood.

Design interventions as tests

Each action should test a hypothesis and produce evidence:

Assign an owner, date, expected evidence, and decision consequence. “Increase engagement” is not an intervention.

• Interview the new program owner to test whether the original objective remains funded and relevant.
• Observe one team configuring an assessment to locate the setup failure instead of assuming training is the answer.
• Review two completed assessment workflows with the customer to determine whether the resulting evidence supports their stated use.
• Offer a narrowly scoped setup session, then compare completion of the blocked step before and after.

Produce a renewal risk plan

The plan should include a dated fact sheet, signal register, stakeholder map, open questions, intervention log, and decision calendar. Where the team uses a risk rating, define it from observable conditions and show its evidence. Keep commercial judgment with the accountable owner.

Verification checklist

Confirm that contract facts match the signed record; usage definitions describe meaningful events; outcome statements do not imply causality without evidence; stakeholder sentiment comes from direct interaction; support issues are linked to cases; unknowns remain visible; and every intervention has a test and decision date.

Sources and claim map

The account, activity, and intervention results are synthetic. This educational plan does not predict renewal, claim validated benchmarks, or promise revenue or customer outcomes.

• GOV.UK Service Manual, How to set performance metrics for your service: supports connecting metrics to service purpose, hypotheses, and data sources.
• GOV.UK Service Manual, Learning about users and their needs: supports treating internal opinions as assumptions and using direct user evidence.

Sources: [1] [2]