Sydney SMBs running their first Claude pilot in 2026 ask the same question: how do we get from a working pilot to production? Most pilots stall at exactly this transition. The 90-day playbook that does not stall has structure, named owners, and explicit measurement at every step.
For a Sydney SMB at $5M to $50M revenue running a focused pilot, the right 90-day playbook produces a production-ready deployment for $80,000 to $300,000 AUD. The wrong playbook produces an extended pilot that costs the same and never ships.
The number 90 is not arbitrary. Shorter timelines skip operational design and change management, the two places where production deployments actually fail. Longer timelines let momentum die while the pilot team drifts back to their day jobs. Ninety days is long enough to build properly and short enough that the senior sponsor stays engaged from kickoff to handover.
Days 1 to 14: pilot evaluation
The first two weeks evaluate the pilot. The questions to answer:
Did the pilot meet the measurable outcomes set at kickoff
What did the pilot teach about user behaviour and edge cases
What needs to change before real customers or staff depend on the system
What does production-ready mean for this specific use case
Who will own the system in production after the consultant leaves
The output is a production-readiness document with explicit gaps and the remediation step for each. Skipping this document is the single most common cause of the never-ending pilot. It also gives the senior sponsor a clear go or no-go decision point backed by evidence rather than enthusiasm.
Days 15 to 35: production design
The next three weeks design the production system around what the pilot taught:
Architecture for production: capacity, redundancy, observability
Operational design: who responds when something breaks, and which runbooks exist
An eval suite that catches behaviour drift once the system is live
Audit and compliance design aligned to the regulatory profile, including Privacy Act obligations
A change management plan for affected staff and customers
The output is a production design document the SMB's own team understands and can operate. If the design only makes sense to the consultant, it has failed.
Days 36 to 70: build and test
The build phase turns pilot work into a production-grade system:
Code and configuration hardening from pilot quality to production standard
Integration with existing systems, from the CRM to the accounting stack
Eval suite implementation with a measured baseline before launch
Operational tooling: alerting, logging, dashboards
User acceptance testing with a pilot user group
Daily standups with the SMB's nominated owner. Weekly demos to the senior sponsor. Issues surface as they arise, not at the end. With Claude-based builds, most of the engineering effort sits in integration and evals rather than the model itself, which is why the build window is five weeks and not five months.
Days 71 to 85: soft launch
The soft launch is a controlled rollout, not a celebration:
10 to 25 percent of the eventual user base
Active monitoring against explicit success and failure criteria
Rapid iteration on observed issues
Structured user feedback collection
Operational runbooks tested in real conditions
The output is confirmation that the system handles real production traffic from real Sydney users, not just synthetic tests. Sydney businesses that skip this stage usually discover their edge cases in front of the full customer base, which is the most expensive possible classroom.
Days 86 to 90: full launch and handover
The final week completes the rollout:
Rollout to the full user base, with confidence built on the soft launch data
Operational handover to the SMB's nominated team
Documentation transfer and training completion
A post-launch review with lessons captured in writing
30, 60, and 90-day check-ins booked before the consultant walks out the door
The consultant's exit is clean. The SMB owns the system, the runbooks, and the evals.
Where pilots stall
Five patterns stall Sydney pilots indefinitely:
No nominated production owner from day one
Vague success criteria that get reinterpreted later
Operational design left until launch is imminent
Underinvested change management for the people the system affects
Consultant scope creep that quietly resists handover
Each one is preventable with playbook discipline, and each one is cheaper to prevent in week one than to fix in week ten.
What it costs
A working 90-day pilot-to-production engagement for a Sydney SMB typically runs $80,000 to $300,000 AUD across consultant time, infrastructure, and internal staff time. The wide band reflects integration depth and regulatory profile: a $45,000 internal Claude tool graduating to production sits at the bottom of the band, while a customer-facing deployment in a regulated industry sits at the top.
If you are sizing a pilot-to-production move for your Sydney business, book a brainstorming call and we will map your 90 days against this playbook.
