How we work. Step by step.
Each program has a defined process. These playbooks walk through how Signal, Forge, and Run actually work — what happens in each phase, what we need from you, and what you get at the end.
The Signal Playbook
Signal runs in three phases over four to six weeks. Every phase has a defined output that the next phase builds on. You always know where we are and what comes next.
Operational Diagnostic
We map every major workflow in your business — not just the obvious ones. Stakeholder interviews, tooling audit, data infrastructure review. Output: a clear picture of where you are today and where the friction lives.
Opportunity Analysis
We score every identified opportunity against four variables: impact, effort, speed-to-value, and strategic fit. You get a ranked list with the rationale visible — not a black box. Output: prioritized opportunity map with sequencing logic.
Roadmap Delivery
A 12-to-24 month roadmap with defined milestones, budget ranges, and decision gates at each stage. Board-ready presentation included. Output: everything you need to align leadership and move to execution.
The Forge Playbook
Forge is a 90-day path to production. The scope is locked before day one — no scope creep, no open-ended retainer. What gets built is agreed in writing. Go-live occurs only when the production-readiness gates are met. Training is embedded across the deployment — not a separate academy after the fact. We work alongside your team through the defined stabilization period and transition.
Qualify and Blueprint (Days 1–15)
We define the outcome in measurable terms, name the internal owner, map the workflow, and lock the risk boundaries before anything gets built. Output: scoped build plan, technical design, and explicit success metric.
Build the Controlled Prototype (Days 16–30)
Agent construction starts in a controlled environment. Retrieval, reasoning, escalation, and workflow integration have to prove themselves before any write access is allowed. Output: working prototype with guardrails and traceability.
Shadow Test (Days 31–45)
We compare the agent against historical and live-shadow work, study disagreements, and hold the system at the eval threshold before it touches customer-facing production. Output: evidence that performance is real, not anecdotal.
Human-Approved Production (Days 46–60)
The system goes live on real work with approval-gated actions, measured execution reliability, and operators already training on the workflow. Output: live production with humans still approving the critical moves.
Bounded Autonomy (Days 61–75)
Selected low-risk actions earn limited autonomy inside explicit rules, while the team samples decisions and tunes the edge cases that only volume reveals. Output: measured autonomy without losing control.
Stabilize and Transition (Days 76–90)
We finish the stabilization period with runbooks, ownership transfer, operating cadence, and a clear choice about whether Run adds value next. Output: production system your team owns, complete documentation, and an optional path into managed assurance.
The Run Playbook
Run is an ongoing monthly engagement for managed governance and operational assurance — not a fixed-length project. We onboard once, then settle into a steady rhythm: continuous assurance, an Optimize phase each month (evals, change control, expansion), and quarterly strategy resets. Training continues as operator enablement inside Run. It is managed assurance for teams who already own the system — not a dependency manufactured at handoff.
Onboarding and Baseline
We instrument your deployed systems, connect to the usage and health signals that matter, and establish baselines so we know what good looks like. Output: a monitored AI estate with documented baselines and an agreed operating cadence.
Optimize Phase (inside Run)
Continuous monitoring catches issues before they reach your team. The Optimize phase each month tunes prompts and agents against real usage — under the Agent Change Control Standard: versioned prompts, models, tools, policies, and permissions; propose/approve; regression eval; staged rollout; auto-rollback on threshold breaches; production agents cannot change their own permissions or governing instructions; feedback is never deployment authority. We also deliver a usage analytics report and ranked recommendations for the next automations worth building. Output: systems that stay reliable and keep improving without silent self-modification.
Quarterly Strategy Review
Every quarter we step back, review what AI is actually moving in the numbers, and reprioritize the roadmap — scoping the next expansion so the investment compounds instead of plateauing. Output: a refreshed roadmap and a clear plan for the quarter ahead.
Ready to run the play?
Every engagement starts with a discovery call. We'll figure out which program fits where you are and what a scoped engagement looks like before anything is signed.
Book a Discovery Call