The AI-enabled workforce.
A closed-room session series on what it actually takes to get AI past the pilot. One speaker who has done it, no vendor presentations, one seat per organisation.
Stone & Chalk, Sydney Tech Central
From Banned to Board Priority
Building an AI-enabled operating model, told by Oliver Sebastian, Director of IT and Business Transformation at Landcom. Two years ago Landcom banned generative AI through policy and network controls. It is now the top strategic priority for its executive team and board.
Written into policy. Blocked at the network level.
Approval to continue the work, with reservations.
The board relaxed the enterprise AI risk rating.
report real productivity or efficiency gains from AI.
use it to redesign how they work, rather than just go faster.
see material EBIT impact, despite 88 in 100 using AI somewhere in the business.
Figures presented in the session, 24 September 2026.
The leadership challenge
Weak results usually reflect gaps in how the organisation works: no baseline, no clear owner, poor integration, unchanged workflows. Closing those gaps needs decisions from the leadership team, not only from IT.
Four reasons for the ban
One decision, four different fears. None of them were really about the technology.
What could it reach
It was unclear what would happen if a model reached data it should not.
Personal liability
AICD warnings put directors’ personal liability in front of them.
Unknown cost
The cost of responsible implementation was unclear, so funding was withheld.
Jobs
People were concerned that AI would lead to redundancies.
Underneath all four was really one thing. Nobody had a clear picture of what generative AI actually was. And you cannot govern what you do not understand.
Oliver Sebastian, on why the ban was the rational position at the time
Five steps before deployment
Security, data access, governance, skills and delivery capacity, settled before a single licence was issued.
Strengthen cyber security
Harden the environment before deploying AI tools.
Scan every unstructured data environment
Check what a copilot could reach before giving people licences.
Establish AI governance
Strategy, policy and an AI register, with AI recorded as an enterprise risk.
Develop internal skills
Build capability in the technology team and in an early adopters group.
Bring in a delivery partner
DeepWeaver provided delivery capacity, with direction retained in-house.
After a few months most of the foundational work was complete and Landcom moved to a curious but cautious position. The team could keep going, although the organisation had not yet endorsed AI adoption.
Six pillars, governance underneath
The CEO and COO helped develop the AI strategy and the board formally endorsed it. The four delivery pillars only stand because the two underneath them were built first.
Capability and training
Foundational AI training for every person in the organisation.
Business, operations, optimisation
Process, decisions, efficiency.
Customer and stakeholder experience
Engagement, personalised services, value to partners.
Workforce and partnerships
Planning for a workforce that uses AI agents.
AI security and governance
Controls over data, risk and compliance.
Responsible AI, ethics and ESG
Ethical standards, responsibility, sustainability.
No one left behind. Every single person at Landcom gets foundational AI training. Not just the early adopters, not just the technical teams. Everyone.
Oliver Sebastian, on capability. Executives were trained separately so they could lead the implementation
Know your board
Put the same proposal through each director’s area of expertise. They assess the risk they know best, and they find the issues that need attention.
Risk appetite, controls, the audit committee’s oversight.
Contract terms, vendor liability, third-party exposure.
Ethical use, community impact, environmental footprint.
With the strategy and governance in place, the board changed the enterprise AI risk rating to flexible. Bold but bounded opened the way to investment in experimentation, learning and wider deployment.
Why adoption has not become return
Around 6 in 100 organisations report material EBIT impact, while 88 in 100 use AI somewhere in the business. The gap is not the technology.
of generative AI pilots produced no measurable profit and loss return.
MIT, across more than three hundred disclosed deployments.
Applied at Landcom
The next phase, with the lessons from each process carried into the one after it.
Pick two or three
Focus on a small number of high value processes.
Measure first
Cycle time and cost recorded before anything is implemented.
Name a sponsor
Agree a success threshold and a date to scale or stop.
Keep productivity separate
Separate ownership, metrics and funding.
Training, usage and governance
Three things Landcom would tell anyone starting the same work.
01Managing AI usage
02Training is a condition of continued access
03Use the existing governance framework
It would have been very easy to stand up a parallel AI programme with its own governance, its own reporting line, its own everything. It feels responsible. And within a year it becomes invisible to the business it is meant to serve.
Oliver Sebastian, on the risk of creating a separate AI programme
Six actions for technology leaders
Tick what you could already answer today. The gaps are the agenda for your next leadership meeting.
Identify specific concerns
Give the board a clear account of each risk.
Scan your unstructured data
Understand what AI tools could reach.
Pick one core process
Focus the initial effort.
Set the baseline first
Measure current performance.
Require training for licences
Make training a condition of continued access.
Understand your board’s priorities
Explain risks in terms each director already knows.
Four questions for the executive team
What are the specific risks?
Can each member name the risks that concern them, and say why?
What is in our unstructured data?
If we handed a copilot to everyone tomorrow, what would it surface?
Which process, and why that one?
Can we name it, and state its current cycle time and cost?
Who owns it, and when do we stop?
Agree a sponsor, success criteria and a review date before work starts.
Two specialisms, one engagement
DeepWeaver brings the AI. Digital61 secures and operates the environment it runs in. The sessions exist because most organisations do not have an AI problem, they have a governance, security and operationalisation problem.
An Australian AI consultancy and an Anthropic Claude partner. The practice is built around adoption rather than procurement: get the technology to the people doing the work, prove it on real processes, then keep going until agents are running in production. Certified under ISO 42001, the AI management standard.
A Canberra-headquartered managed security and managed services provider working with Australian government, critical infrastructure and enterprise. Identity, endpoints, cloud platform and a 24/7 security operations centre, run inside the Microsoft tenancy you already own. DISP member with an IRAP PROTECTED-assessed managed SOC.
Who to talk to
Get the session summary and the slides
Four reasons for the ban, the five steps Landcom took before deployment, the six-pillar strategy the board endorsed, the eighteen-month evidence on adoption against return, and the lessons on training, usage and governance.
Sent in about a minute
Straight to the address you give.
Keynote content only
The discussion that followed was Chatham House and is not recorded anywhere.
A seat if you want one
Tick the box and we hold you one before the next session is announced.