The Retainer.
The build, running and tuned, in your boardroom every month.
Where the operation actually lives. Once the install is live, the same tiger team that built it stays on retainer to run it, tune it, defend it, and expand it. The team that shipped the system stays accountable for it.
Deliverables.
- →Monthly KPI review on every workflow in scope
- →Quarterly business review with executive sponsor
- →Vendor management across the AI stack
- →Governance updates as models, regulations, and data perimeter evolve
- →Incremental builds as new workflows enter scope
- →Tiger team on call during business hours for incidents
How the retainer actually runs.
Operational watch
We watch the KPI dashboard daily. Anomalies (latency, refusal rate, agent error rate, conversion drops) get caught before you'd have noticed. Most weeks pass without incident — the goal is for nothing to surprise you.
Monthly KPI review
60-min review with operational leads. Every workflow in scope, its current KPI vs. baseline, and any tuning we did or recommend. We don't ship slide decks; you get a one-pager.
Quarterly business review (QBR)
90-min review with the executive sponsor. Strategic posture: what's working, what's changing in the AI landscape that affects your operation, what new workflows are candidates for the next Install expansion. We're operator-to-operator here, not vendor-to-customer.
Vendor management
When OpenAI ships a new model, when Anthropic changes pricing, when a vertical SaaS adds AI features that overlap with our agents — we evaluate, recommend, and execute the change. You don't have to track this.
Incremental builds
Workflows you didn't include in the original Install can be added later — typically scoped as 2–4 week mini-Installs inside the retainer envelope, or as new Beachhead Builds depending on size.
Tiger team on call
Business-hours pager. If an agent misfires — a wrong answer, a customer complaint, a regulatory question — we're on it within an hour. Severity-1 incidents get a written post-mortem.
What you get on paper.
ANONYMIZED SAMPLES. CLIENT-SPECIFIC ARTIFACTS STAY UNDER NDA.
Monthly KPI one-pager
Every workflow's current KPI vs. baseline, tuning notes, and what we recommend next. Shared in Notion.
QBR summary (quarterly)
Strategic-posture review, run as an operator conversation — 90 minutes, with a written summary, not a slide show.
Vendor stack changelog
Every model swap, every pricing change, every new vendor evaluated and decided against. Hosted in Notion.
Incident post-mortems
For any Sev-1 event. Public to your team, retained for governance audit.
Annual security review
Once a year, formal walk-through of your data perimeter, audit logs, and model-routing posture. Defendable in a regulatory exam.
Inside this phase.
The toolchain is selected per engagement. This is the typical stack.
- PagerDuty (severity routing)
- Notion (deliverable layer)
- Sentry / Datadog (observability)
- OpsCheck cadence
- Vendor Stack Rationalization Framework
Your team's role during this phase.
- →Operational leads: 60 min/month for the KPI review.
- →Executive sponsor: 90 min/quarter for the QBR.
- →IT lead: 60 min/year for the security review (more if regulated).
- →On Sev-1 incidents: as much of your team as needed, on our coordination.
- →Outside those touchpoints, the system runs without you in the room. That's the design.
Where this phase ends.
The Retainer is open-ended by default. 12-month initial term. Adjusts annually with expansion scope. Cancellable with 60 days' notice after year one. Most clients don't cancel — they expand. The point of the install is that the operation now runs without you having to think about it, and most operators then want the next domain installed rather than going back to running this one by hand.
Book an Ops Call.
30 minutes, operator-to-operator. We talk through your operation and the numbers. If we're not the right fit, we say so on the call.
Book an Ops Call →