Three ways to work with us
Start with a decision, build when you're ready, and keep it accurate after.
Offer 1 · Foundations Workshop
A fixed-scope week that ends with an approved roadmap, not a slide deck of options.
Two weeks of pre-work make the week productive: access setup, a stakeholder intake survey, five to eight interviews, and the diagnostic audit.
- Monday
- Audit readout: where the foundation breaks today, rebuild versus extend, and baselines for time to answer, contract change cycle time, and data team ticket volume
- Tuesday
- Use case and metric alignment: prioritize AI use cases, agree P0 metric definitions and owners, and draft them into the PRD template
- Wednesday
- Architecture: target design for the contract-generated model, knowledge graph, driver agent harness, routing, Genie spaces, and context file generation
- Thursday
- Trust and evals: accuracy tiers, thresholds per tier, refusal boundaries, golden question set, and governance for sensitive data
- Friday
- Roadmap and executive readout
Deliverables
Readiness audit
Lineage, usage, observability, and contract coverage findings, scored by domain.
Aligned P0 metrics
Definitions, owners, grain, sources, and acceptance criteria that engineering and the agent both build against.
Target architecture
Knowledge graph, harness, routing with Genie, and context file generation, plus a routing diagram.
Eval plan
Golden question set, accuracy tiers and thresholds, refusal tests, and leadership metrics.
Roadmap
Phased build plan with baselines, success metrics, cost forecast, staffing, and risks.
Executive readout
A decision memo your leadership team can approve the same week.
Workshop credit: 100% of the workshop fee is credited toward implementation if signed within 60 days.
Offer 2 · Implementation
We build the full pattern with your team, including evals, and hand it over running.
Fixed fee, priced by number of domains, with milestones tied to outcomes instead of hours. The same contracts that generate the metric views also generate the context files, so the agent and the data never drift apart.
Evals we build in
| Eval | What it measures | Example threshold |
|---|---|---|
| Answer accuracy | Correct answers on the golden question set, by metric tier | Set by leaders per tier |
| Refusal precision | Agent declines out-of-scope, uncertified, or restricted questions | Near 100% on restricted data |
| Routing accuracy | Questions reach the correct domain space | Agreed during build |
| Contract regression | Answers stay correct after a contract or definition change | No regression on certified tier |
| Traceability | Every answer links to a certified source and definition | 100% of certified answers |
| Cost and latency | Compute cost and time per answer | Within forecast budget |
Scope tiers
Foundation
1 to 2 domains, driver agent, evals, enablement
Scale
3 to 5 domains, knowledge graph, automated context generation
Enterprise
Custom scope, multiple business units or regulated data domains
Offer 3 · Advisory
Keep production AI analytics accurate, affordable, and owned by your team.
Definitions change, new domains appear, and costs creep. Advisory keeps the system on track.
Essentials
Monthly office hours, debugging support, and a quarterly eval and accuracy report for leadership.
Embedded
Everything in Essentials, plus cost optimization reviews, operating model design for context updates, context governance reviews, and roadmap support.
Both tiers carry a three-month minimum. Implementation clients get their first month of Essentials included.
Founding client terms
Three founding spots.
Founding clients receive a discounted workshop rate, full workshop credit toward implementation, and priority scheduling. In exchange, they give permission to publish a co-branded case study.
Next step
Book a 30-minute fit call.
If it's a fit, we send the intake survey and access checklist, and the workshop can start within three weeks.