NetSuite
We reimplemented NetSuite and brought it into active reporting use within 90 days. Before that work, the system had been live for two years without being used for reporting.
BINANCE
Accounting and Finance Systems
BINANCEAccounting and Finance Systems
Empowering Teams with Modular AI Workflows
Congratulations, finance. You’re engineering now. We just added guardrails. A new way to build enterprise software. Purpose-built for teams who know the problem, not necessarily the code. AI turns discussion into controlled, deterministic workflows that actually run.
03 / Systems
Four systems. One coordination layer.
Workday
The people and workflow layer. Workday listed Patrick for finance-track sessions.
Workday Planning
The planning layer. Tridant described a governed Binance Adaptive Planning program for revenue and margin modelling.
Trintech
The close and reconciliation layer.
05 / Approach
Make every change understandable before making it fast.
- Frame the requestRecord the problem, owner, source, affected system, and completion condition.
- Map the boundaryDefine which platform owns the record, decision, approval, and exception.
- Build in modulesKeep actions bounded, deterministic, reviewable, and easy to replace.
- Review the resultRead back the authoritative record instead of treating activity as proof.
- Improve from useUse support-room observations to sharpen the tools without weakening control.
Workflow modes
Match the control model to the uncertainty.
The supplied workshop frames predictable accounting flows differently from AI-assisted document flows. Both converge on visible outputs; the probabilistic path adds an explicit approval gate. Business partners can combine modular components like building blocks, assembling the process they need while keeping every step, control, and handoff visible.
06 / Validation
Evidence before outcomes.
Every change closes with a readback from the authoritative system, a recorded approval, and a clear recovery path.
SourceIdentify the authoritative record before work begins.
ApprovalKeep decision points visible and assigned to the right owner.
ReadbackConfirm the final state in the system of record.
RecoveryRecord the limitation, exception route, and safe rollback.
07 / Delivery
A repeatable path from request to controlled change.
Evidence-first intake
Problem, source, owner, system, controls, and expected behavior in one brief.
System boundaries
Plain-language ownership across records, integrations, approvals, and exceptions.
Controlled prototypes
Assumptions stay visible while production behavior stays explicit.
Validation receipts
Tests, readbacks, sign-off, limitations, and recovery tied to the exact change.
Role-aware enablement
Training connected to the task and to a clear support route when it fails.
From conversation to controlled build
Turn observed needs into small, auditable systems.
The workshop material frames enablement as a sequence: preserve context, expose the request, explore a solution, then turn it into a governed build.
FIG 0.1Start with discussion
Work through the problem together. Decisions, requirements and context become part of an auditable trail—not lost in meetings.
FIG 0.2Describe the need
Ask for what you need in plain language. The agent uses the discussion to fill the gaps, challenge assumptions and shape the solution.
FIG 0.3Let AI find the path
Controlled agentic reasoning explores the messy middle—turning intent into something structured, testable and surprisingly effective.
FIG 0.4Build it deterministic
The result becomes a low-temperature, repeatable pipeline of small products—scheduled, governed and built directly on your platform.
Preserve Core Controls
Maintain existing approvals, evidence retention, and financial governance processes.
SOC-Certifiable Automation
Build toward SOC verification for zero-temperature processes, centralising execution when appropriate.
Human Oversight for AI
Introduce additional review and validation as workflows become more probabilistic.