AI Readiness Assessment for Revenue Teams: A Manual Scorecard
A team is not ready for revenue AI because it bought software or cleaned one contact list. Readiness means one bounded workflow has trustworthy inputs, accountable owners, safe handoffs, measurable outcomes, and a way to stop when evidence deteriorates.
This is a manual self-assessment. It stores no input and predicts no ROI. Complete it with the people who own sales, service, data, systems, risk, and the selected workflow.
Score the six domains
Use only documented evidence:
0— absent, unknown, contradictory, or dependent on an individual.1— partly defined and usable for a supervised pilot, with gaps recorded.2— documented, tested on the chosen workflow, owned, and reviewed on a cadence.
Add the six scores. Equal weighting keeps the method inspectable; do not change weights midway to obtain a preferred result.
| Domain | 0 | 1 | 2 |
|---|---|---|---|
| Workflow boundary | No clear start/end or excluded decisions | One workflow and exclusions are drafted | Start/end, exclusions, exception routes and version owner are tested |
| Data fitness | Required fields and sources are unknown | Minimum fields exist but gaps/duplicates need manual review | Sources, definitions, permissions, quality checks and correction path are documented |
| Ownership and handoff | No accountable operator or acceptance event | Primary owner exists; backup/SLA is incomplete | Owners, backups, acceptance, timeout and escalation are tested |
| Controls and risk | No prohibited actions or stop rule | Guardrails exist but are not replay-tested | Access, prohibited actions, logs, review and pause/rollback are tested |
| Measurement | Activity counts only or no baseline | One outcome and quality measure have definitions | Baseline, denominator, window, quality/risk and review decision are agreed |
| Change capacity | No time, training or feedback owner | Pilot resources exist but adoption plan is incomplete | Users are trained; feedback, correction and expansion gates have owners |
NIST's AI RMF organizes risk work through Govern, Map, Measure and Manage. OpenAI and Anthropic likewise emphasize bounded tasks, guardrails, human intervention and starting with simpler workflows. The scorecard turns those principles into a revenue-operations discussion; it is not a certification. See NIST, OpenAI and Anthropic.
Stop conditions override the total
Do not average away a critical gap. Stop the pilot when any of these is true:
- the workflow can make a legal, financial, clinical, safety, eligibility, pricing or contractual decision without an authorized person;
- required data lacks a lawful/approved use, access control, correction path or system of record;
- nobody accepts handoffs and incidents;
- the team cannot reconstruct what information and rule produced an action;
- the success metric ignores quality, rework, complaints or harmful failures; or
- no one can pause and roll back the workflow.
Record each stop condition separately from the score.
Manual worksheet
Copy this block into a blank document. Name one workflow, then score each domain only after writing the evidence or missing item beside it; if a stop condition appears, record it before adding the total.
Chosen workflow:
Decision supported:
Evidence review date:
Workflow boundary (0–2):
Evidence / missing item:
Data fitness (0–2):
Evidence / missing item:
Ownership and handoff (0–2):
Evidence / missing item:
Controls and risk (0–2):
Evidence / missing item:
Measurement (0–2):
Evidence / missing item:
Change capacity (0–2):
Evidence / missing item:
Total (0–12):
Stop condition present (yes/no):
Weakest domain:
Named action, owner and due date:
Define interpretation before scoring
The scorecard supplies no universal action bands. The accountable team records its own thresholds before seeing the total:
Stop-rule owner:
Minimum score required in every domain:
Total range for replay only:
Total range for supervised pilot:
Total range for considering expansion:
Required evidence at each gate:
Who approves the decision:
Any stop condition overrides the total. A score applies only to the named workflow and evidence date; it never means expected ROI or broad organizational readiness.
Completed fictional assessment
Fictional services company Northstar assesses automatic qualification of inbound website leads—not “all sales AI.” Before scoring, it chooses: any domain 0 means documentation/replay only; total 5–8 with no zero permits supervised replay; total 9–11 with every domain at least 1 permits consideration of a limited pilot; 12 permits independent review for expansion. These are Northstar's illustrative choices, not defaults.
Chosen workflow: Automatic qualification of inbound website leads
Decision supported: Whether to move from supervised replay to consideration of a limited pilot
Evidence review date: 2026-08-16
Workflow boundary (0–2): 2
Evidence / missing item: Versioned start/end, exclusions and human-review route
Data fitness (0–2): 1
Evidence / missing item: Required CRM fields listed; duplicate-account cleanup incomplete
Ownership and handoff (0–2): 2
Evidence / missing item: Primary/backup queues and acceptance event replay-tested
Controls and risk (0–2): 1
Evidence / missing item: Prohibited decisions written; rollback has not been rehearsed
Measurement (0–2): 1
Evidence / missing item: Accepted-handoff definition exists; baseline window not closed
Change capacity (0–2): 2
Evidence / missing item: Two pilot reps trained; weekly correction owner named
Total (0–12): 9/12
Stop condition present (yes/no): No
Weakest domain: Data fitness; Controls and risk; Measurement (tied at 1)
Named action, owner and due date:
- Clean up duplicate accounts — Data owner — 2026-08-18
- Rehearse rollback — Systems and risk owners — 2026-08-19
- Close the baseline window — Revenue operations measurement owner — 2026-08-22
Recalculation: 2 + 1 + 2 + 1 + 1 + 2 = 9. Northstar does not expand. It completes the three named actions before the accountable owner considers a limited pilot. The score does not estimate revenue.
Sensitivity and invalid states
If any 1 was awarded without a document or observable test, reduce it to 0: Northstar would move from 9 to 8 and remain supervised. If a missing permission or unowned handoff emerges, the stop rule applies regardless of total. Missing domains, half-points, percentages and weighted averages are invalid for this version.
Frequently asked questions
These questions prevent a local score from being mistaken for a universal readiness standard.
Is 9/12 a universal definition of readiness? No. Northstar chose that illustrative gate before scoring. Your accountable owners must set and document their own gate.
Can a high total offset a privacy or ownership stop? No. Stop conditions override arithmetic.
Should one score cover every AI use case? No. Reassess each bounded workflow with its own evidence date and owners.
Next steps, method, and limits
Act on the lowest domain first. For data foundations, use the CRM cleanup playbook. For a narrower intake check, use the lead qualification scorecard. For the implementation contract, use the qualify-and-route workflow.
Method and limits. This page performs no calculation in the browser and stores no answers. The user adds six whole-number inputs from 0 to 2; invalid or missing values produce no result. Equal weights reflect a conservative dependency view, not empirical prediction. The assessment needs review by the relevant data, risk and workflow owners before any live decision.