01
Operator-led, AI-augmented
ScaleCadence is operator-delivered. The founder makes every recommendation that reaches the client. AI accelerates the analytical work: approximately 60% of the routine labor (document intake, parallel research, schema reconciliation, renewal analysis, deliverable assembly), so the founder’s time is reserved for the 40% that determines outcomes (executive interviews, problem framing, sense-check judgment, synthesis, and the final recommendation).
02
Variant A deployment
ScaleCadence accelerator IP runs in the ScaleCadence environment, not at the client. Clients receive insight and deliverables (reports, registers, playbooks, briefs), they do not install or operate any ScaleCadence software. This keeps engagements consulting-shaped, keeps the moat on ScaleCadence’s side, and keeps clients out of vendor-management overhead.
03
Sole AI sub-processor: Anthropic
All AI calls route to Anthropic’s Claude API. Commercial API inputs and outputs are not used to train Anthropic models. No other LLM providers are invoked. No third-party embedding services. No third-party agentic platforms. This is the simplest defensible AI sub-processor disclosure in fractional SPO advisory.
04
Optional structural-context inputs
For Offering 2 and Offering 4 engagements, ScaleCadence asks the client’s IT team for structural context (data dictionary, master data, entitlement state, transaction event taxonomy) alongside the data extracts. The framing: send what you have, not a blocker for kickoff. Engagements run end-to-end regardless of how much context exists, but the more documentation the client has, the faster the engagement compresses.
05
Three-tier data triage — Green / Amber / Red
Every input is classified before any AI call. – Green-tier (interview notes ScaleCadence takes; public strategy documents; metric definitions; aggregate summaries), freely AI-processed. – Amber-tier (internal docs; redacted board packs; tokenized account lists; financial summaries), AI-permissible only after redaction or aggregation. – Red-tier (raw customer records, PII, PHI, payment-card, government IDs), never enters AI workflows; analyzed manually or in the client’s own environment.
The triage is enforced at the plugin level, not as an external policy.
06
Opt-out right
Every client may elect no-AI delivery. ScaleCadence reverts to the manual baseline; timelines extend by 30–50% and fees adjust accordingly. The client retains every deliverable produced under no-AI delivery just as they would under AI-augmented delivery.
Where does my data live during the engagement?
ScaleCadence environment for the engagement window; deleted 30 days after final deliverable acceptance.
Will my data train any AI models?
No. Anthropic’s commercial API does not train on inputs.
Can I see what AI is doing under the hood?
Yes. The AI Delivery Brief at kickoff lists tools, sub-processors, data tiers, and retention policy.
What if our IT team can’t share the structural-context documentation you ask for?
The engagement runs anyway. We surface the documentation gap as a finding and lead the IT remediation recommendations with “build a data dictionary.”
What if I want to own the model and run it myself after the engagement?
Available via the Predictive Renewal Risk Model. Build & Transfer variant of Offering 4. Custom-trained model + complete handoff package + workshops with your data and CS teams. You run the model on your infrastructure post-handoff. No subscription. Year-2+ cost = $0.