Sample report. Anonymised. Names, figures, and findings illustrative.
Lumen & Lever
AI Usage Control Baseline · Confidential
Leadership Briefing

AI Usage Control Baseline. Meridian Holdings.

Prepared for
The Board, Meridian Holdings
Engagement Sponsor
Group CFO
Engagement Lead
Lumen & Lever
Engagement Period
Two weeks · concluded
Document Version
Final · for Board distribution
Classification
Confidential · Board only
Section 1 · Executive summary

What we found.

Meridian Holdings is at a point most diversified family groups reach roughly twelve months into informal AI use. The pattern is consistent and the implications are now actionable.

The headline finding

AI is already operating across all six business units at Meridian, predominantly through personal accounts, mostly on sensitive material. The binding constraint is not whether to adopt AI. It is governance over what is already happening.

Across 218 staff surveyed, 74% report regular AI use for work tasks. Of those, 61% access AI tools through personal accounts, predominantly ChatGPT and the AI features built into consumer phone and laptop operating systems. 43% have entered information that meets the group's existing definition of confidential, including lease documents, tenant correspondence, employee performance notes, and supplier negotiations.

None of this use is currently logged, sanctioned, or governed. Staff have not been issued guidance. No AI policy exists. No central register of tools is maintained. The exposure is not from an absence of staff capability. It is from an absence of structure around the capability they have already built informally.

A single workflow in the Property division, lease abstraction at renewal, has emerged as the strongest first candidate for a controlled implementation. Roughly 80 leases per year, three weeks of senior paralegal time per cycle, a corpus of accessible documents. The economics are real and the exposure is bounded.

74%
of Meridian staff use AI tools at work regularly
Source · Meridian staff survey · n=218
61%
access AI through personal accounts
Source · Meridian staff survey · n=161
43%
have entered confidential information into AI tools
Source · Meridian staff survey · n=161
0
formal controls currently in place
Source · Meridian governance review
Section 2 · Sector context

The pattern at Meridian is not unusual.

Meridian's findings sit within an established industry pattern. The figures from independent research are consistent with what the Baseline surfaced internally, which gives the leadership team confidence that the controls recommended below are appropriate to the scale of exposure.

90%
of organisations report employees regularly using personal AI tools for work, while only 40% have purchased official subscriptions
Source · MIT Project NANDA, State of AI in Business 2025
68%
of enterprise employees who use generative AI access it through personal accounts
Source · TELUS Digital, January 2025
57%
of those have entered sensitive information into AI tools
Source · Menlo Security 2025 / TELUS Digital
$670K
additional cost when a data breach involves shadow AI, on average
Source · IBM Cost of a Data Breach Report 2025

"If employees don't feel safe admitting how they use AI, you fail to manage the risks you can't see and you fail to capture the value you don't know exists. The most effective policies start with psychological safety and transparency, not punishment."

Khullani Abdullahi · Managers at Work, 2025

Section 3 · AI Use Register

What is actually in use, and where.

The register names the AI tools observed in use across Meridian's six operating units. Tools listed have been confirmed through the staff survey, the targeted intake conversations with senior leaders, or both. The register is current as at the close of the engagement and is intended to be maintained by the group's nominated AI governance owner going forward.

Tool
Primary Use
Account Type
Units
Frequency
ChatGPT (OpenAI)
Drafting, summarising, research
Personal · 87%
All 6
Daily
Microsoft Copilot
Email, document drafting
Work · partial rollout
Property · Finance
Weekly
Otter.ai · Fireflies
Meeting transcription
Mixed · 60% personal
All 6
Daily
Claude (Anthropic)
Long document analysis
Personal · 100%
Property · Hospitality
Weekly
Gemini (Google)
Search-augmented research
Personal · 100%
Marketing · Construction
Weekly
Canva AI features
Marketing imagery, social
Work · sanctioned
Marketing
Daily
Grammarly
Proofreading, tone
Mixed
Property · Hospitality · HR
Daily
Built-in OS AI features
Photo, transcription, summary
Personal devices
All 6
Daily
Perplexity
Research with citations
Personal · 100%
Property · Construction
Weekly
DALL-E · Midjourney
Marketing image generation
Personal
Marketing
Monthly
Where AI is being used · share of staff per business unit reporting regular use
Property
89%
89%
Marketing
92%
92%
Hospitality
71%
71%
Childcare
54%
54%
Beverages
67%
67%
Construction
78%
78%
Section 4 · Exposure Classification

Proceed. Conditions. Stop.

Each pattern of observed AI use has been classified against the group's existing risk appetite. The classification model is intended to be operational rather than aspirational. Staff can be told what they may do, what requires sign-off, and what must stop.

Pattern of Use
Risk Profile
Classification
Volume
Action
General drafting on public information
Low. No data exposure.
Proceed
Daily, all units
Issue staff guidance
Marketing copy and image generation
Low. Brand voice drift.
Proceed
Daily, Marketing
Brand-voice guideline
Internal note summarisation
Medium. Internal data drift.
Conditions
Daily, all units
Sanctioned tool only
Meeting transcription via personal accounts
Medium-high. Confidential meeting content stored externally.
Conditions
Daily, all units
Move to enterprise tool
Lease and contract analysis
High. Commercial obligations.
Conditions
Weekly, Property
Workflow under design
Tenant correspondence drafting
High. PII exposure to consumer tools.
Stop
Weekly, Property
Halt pending sanctioned alternative
Employee performance note drafting
High. HR confidentiality, regulatory.
Stop
Monthly, HR and managers
Halt immediately
Childcare incident note drafting
Severe. Child PII, regulatory.
Stop
Weekly, Childcare
Halt immediately. Audit trail.
Supplier pricing and negotiation
High. Commercial confidentiality.
Stop
Monthly, all units
Halt pending review
Section 5 · Workflow Candidate Shortlist

Where AI would land cleanly, ranked.

Five candidate workflows have been identified through the staff survey, the targeted intake, and the analysis of where current shadow AI use clusters. Each has been scored on three dimensions: business value, exposure profile, and feasibility within twelve weeks. The recommended anchor is named. The remaining candidates are held in backlog.

Workflow
Value
Exposure
Feasibility
Confidence
Outgoings reconciliationBacklog · Property
High
Medium
Medium
3 of 5
Marketing copy productionSanctioned, light governance
Medium
Low
High
4 of 5
Meeting summary and action captureGroup-wide, enterprise tool
Medium
High
High
4 of 5
Rent review supportBacklog · pending Property data layer
High
High
Low
2 of 5

Lease abstraction is named as the recommended anchor not because it is the largest opportunity, but because it is the most defensible first move. The volume is real, the documents are accessible, the cost of error is manageable with appropriate review, and the work that staff are already attempting through personal AI tools can be moved into a sanctioned environment with measurable improvement.

Lumen & Lever · engagement note

Section 6 · Control Baseline

Six controls to install in the next ninety days.

The following six controls move Meridian from no formal posture to a defensible operating baseline. Each is named with an owner, a measurable outcome, and an implementation horizon. None require new technology purchases. All are within the scope of the existing leadership team.

i

AI Use Register, maintained quarterly

The register established during this engagement becomes a living document. Updated quarterly through a five-minute staff pulse and a review by the nominated owner.

Owner · Group CIO Live · 30 days
ii

Staff Use Guidance, single page

One-page document distributed to all staff stating what is sanctioned, what requires sign-off, and what must stop. Published on the group intranet. Reviewed annually.

Owner · Group HR Live · 14 days
iii

Sanctioned Tool List

A short list of sanctioned AI tools with sanctioned use cases. Personal-account access to sanctioned tools is acceptable for low-risk patterns. Sensitive workflows require enterprise access.

Owner · Group CIO Live · 30 days
iv

Stop List, with audit trail

Three categories of use are stopped with immediate effect. Tenant correspondence, employee performance notes, and childcare incident notes are not to enter any AI tool until a sanctioned alternative is in place.

Owner · Group Legal Live · 7 days
v

Enterprise meeting transcription

Existing personal-account use of Otter and Fireflies is migrated to the group's Microsoft 365 environment, where transcription inherits the existing data residency and access controls.

Owner · Group CIO Live · 60 days
vi

AI Governance Forum, quarterly

A standing 60-minute quarterly forum chaired by the Group CFO. Reviews the register, surfaces new patterns, ratifies tool list changes, and signs off any new workflow commencing.

Owner · Group CFO First sitting · 60 days
Section 7 · Summary Position

The single page for the Board.

The position below is the Board-grade single-page summary of the Baseline. It is intended for inclusion in the next Board paper without further translation.

For Board Distribution

Meridian Holdings has an undeclared but material AI footprint, and the right next move is to make it declared.

  • Current state. AI is in regular use by 74% of staff across all six business units. 61% of that use is through personal accounts. 43% has touched material the group considers confidential. No formal controls are in place.
  • Risk position. The exposure is consistent with sector pattern but currently uncontained. Three patterns of use are recommended for immediate halt. Five further patterns require conditions. Two are appropriate to proceed with light governance.
  • Recommended controls. Six controls to install over 90 days. None require new technology. All sit within the existing leadership team's scope. Total estimated internal effort: roughly 12 person-days across six owners.
  • Recommended next investment. A bounded Implementation Control Brief on lease abstraction at renewal, the strongest workflow candidate identified. Not an AI build. A controlled brief that defines what an internal team or external partner must satisfy before any system is commissioned.
  • What this engagement does not do. This Baseline does not select tools, build software, or commit the group to any specific vendor. It establishes the position from which subsequent decisions can be made with evidence.
Section 8 · Recommended Next Steps

What to do next, in order.

Three options for the Board, ranked by structural priority. The recommended path is the first. The remaining two are listed for completeness.

Recommended

Commission an Implementation Control Brief on lease abstraction at renewal.

The Property division's lease abstraction workflow has emerged as the strongest first candidate. Approximately 80 leases per year. Three weeks of senior paralegal time per cycle. A defined corpus of accessible documents. Staff are already attempting this work informally through ChatGPT, with predictable variability in output quality.

An Implementation Control Brief defines what an internal team, current IT provider, software vendor, or specialist AI implementation partner must satisfy before the workflow is trusted in business use. It does not build the system. It produces the document the implementer can build against and the buyer can hold them to.

Engagement
Implementation Control Brief
Duration
2 weeks
Investment
AUD 15,000 to 18,000
Output
Implementer-ready brief, acceptance criteria, evaluation framework, production gates
Alternative

Resolve the Property data layer first via a Document Structure Review.

If the Board's preference is to resolve technical readiness before commissioning an implementation brief, a Document Structure Review of the lease corpus would test whether the source material can support reliable AI use. This is the right move only if there is concern about the extraction reliability of the existing lease documents. Based on the engagement findings, this concern is moderate but not pressing.

Engagement
Document Structure Review
Duration
1 week
Investment
AUD 9,500 to 12,500
Output
Document-ingestion risk assessment, remediation priorities
If circumstances change

Defer to a Structural AI Architecture Sprint when the engagement scope expands.

If the conversation moves beyond a single workflow into broader AI capital allocation, vendor governance, lifecycle control, and Board-level scale readiness, a four-week Structural AI Architecture Sprint becomes the appropriate engagement. This is not the immediate next step. It is the appropriate step in the medium term if AI becomes a material capital question for the group.

Engagement
Structural AI Architecture Sprint
Duration
4 weeks
Investment
AUD 40,000 to 55,000
Output
Board-grade structural briefing, capital gate recommendation
← Return to Lumen & Lever Download PDF