Turning assessment findings into operating rules: a financial services implementation case

Assessment finding

Decision areas where AI must not be used are not explicit.

Service B implementation

Create an AI-permitted workflow list and an AI-excluded decision list.

Assessment finding

Human review triggers are not defined.

Service B implementation

Define review conditions for customer disadvantage, contractual terms, fees, deadlines, and complaint risk.

Assessment finding

Human review criteria are insufficient.

Service B implementation

Create a Human Judgment Review procedure with review criteria and return conditions.

Assessment finding

Differences between AI drafts and final decisions are not recorded.

Service B implementation

Introduce a Decision Log template that records the AI draft, human edits, rationale, final judgment, and approver.

Assessment finding

No owner manages the Decision Boundary™ over time.

Service B implementation

Define Boundary Governance rules for ownership, change control, and periodic review.

Workflow

Customer inquiry responses

AI use

Summarization and response drafting

Design focus

Define review triggers and final judgment ownership by response type.

Workflow

Complaint and sensitive inquiry triage

AI use

Complaint risk classification and severity estimation

Design focus

Use sensitive terms, customer disadvantage, and strong dissatisfaction as review triggers.

Workflow

Sales screening memos

AI use

Meeting note summarization and screening issue extraction

Design focus

Limit AI to issue candidates and exclude approval or rejection decisions.

Workflow

Internal policy lookup

AI use

Policy search and summary generation

Design focus

Require source-text confirmation and version recording for formal decisions.

Deliverable

AI workflow scope list

Description

Classifies permitted, restricted, and excluded AI use by workflow and decision type.

Primary users

Business owners, IT planning, compliance

Deliverable

Decision Boundary™ design document

Description

Defines AI role, human review, and final judgment ownership by decision type.

Primary users

Operations managers, business owners, audit and control teams

Deliverable

Review trigger and threshold list

Description

Documents when human review, escalation, or approval is required.

Primary users

Contact centers, screening teams, compliance

Deliverable

Human Judgment Review procedure

Description

Standardizes what reviewers must check before accepting or changing AI output.

Primary users

Frontline staff, supervisors, approvers

Deliverable

Decision Log template

Description

Records AI drafts, human edits, rationale, final decisions, and approvers.

Primary users

Operational teams, audit, quality management

Deliverable

Boundary Governance rules

Description

Defines ownership, change requests, approvals, and periodic boundary reviews.

Primary users

AI governance owners, IT planning, business owners

Phase

1. Kickoff and scope definition

Duration

1 week

Work

Confirm target workflows, stakeholders, existing rules, and assessment findings.

Output

Implementation scope and stakeholder map

Phase

2. Decision type mapping

Duration

1-2 weeks

Work

Classify decisions across customer response, screening, policy interpretation, and complaint handling.

Output

Decision type inventory

Phase

3. Decision Boundary™ design

Duration

2-3 weeks

Work

Define AI support, human review, human final judgment, and AI-excluded decisions.

Output

Decision Boundary™ design document

Phase

4. Responsibility and review design

Duration

2 weeks

Work

Define review triggers, approvers, and escalation conditions.

Output

Review trigger and responsibility matrix

Phase

5. Evidence design

Duration

1-2 weeks

Work

Define Decision Log fields, storage units, and evidence use.

Output

Decision Log template

Phase

6. Governance design

Duration

1 week

Work

Define ownership, change control, periodic review, and governance forums.

Output

Boundary Governance rules

Phase

7. Pilot and adoption

Duration

1-2 weeks

Work

Pilot with target workflows and incorporate operational feedback.

Output

Final deliverables and training materials

Decision type

Responses involving contract terms, fees, or deadlines

AI role

Drafting only

Review condition

Always reviewed

Final owner

Business owner or authorized approver

Log requirement

Record AI draft, edit rationale, and final response

Decision type

Complaint or sensitive inquiry classification

AI role

Candidate classification only

Review condition

Sensitive terms, customer disadvantage, or strong dissatisfaction

Final owner

Supervisor or responsible manager

Log requirement

Record classification rationale and escalation decision

Decision type

Screening approval or rejection

AI role

Excluded

Review condition

Not applicable

Final owner

Screening team

Log requirement

Record as an AI-excluded decision category

Before

AI response drafts were edited based on individual judgment.

After

Review triggers and final judgment ownership are defined by response type.

Before

Contract, fee, and deadline responses were handled as frontline decisions.

After

Potential customer-disadvantage responses require business-owner review.

Before

AI-generated screening memos could influence screening decisions.

After

AI is limited to summaries and issue candidates; approval decisions remain human-only.

Before

Differences between AI drafts and final judgments were not retained.

After

Decision Logs record drafts, edits, rationale, and final judgments.

Before

Expansion of AI use was left to local operational judgment.

After

Boundary Governance controls change requests, approvals, and periodic reviews.