Selected work in progress

What enterprises are preparing to prove.

An anonymized view of active problem areas at early stages of definition. Early client engagements.

Active Engagements

Problem areas and client evidence in progress.

TokenOps · Public-company AI capital baseline

Establish finance-ready evidence across AI spend.

Enterprise situation: A CFO and board need a consistent view of cross-vendor token and coding-assistant spend.
Workflow: Map usage and invoices to projects, accounting treatment, budgets, and reporting periods.
Product arm: TokenOps
Current stage: Baseline in preparation · proof-of-concept and baseline design.
Intended proof: A project-level ledger, capitalization evidence, budget variance, and board reporting can be produced from traceable inputs.
Next decision: Confirm the baseline boundary, source systems, and proof-of-concept acceptance criteria.

Open TokenOps Zero Integration Sprint demo ↗

RunTime · Customer-support autonomy

Bound routine support actions before a pilot.

Enterprise situation: A support organization is evaluating where routine actions may be safely delegated.
Workflow: Define cost, speed, policy reliability, escalation, and auditability requirements for routine support actions.
Product arm: RunTime
Current stage: Pilot scoping · workflow scoped.
Intended proof: A bounded pilot can show whether permitted actions remain policy-grounded, reviewable, and appropriately escalated.
Next decision: Choose the first action set, policy sources, escalation rules, and pilot evaluation plan.

TokenOps · PE portfolio compute governance

Frame a portfolio-level compute control model.

Enterprise situation: An operating partner and portfolio CFO need comparable visibility without flattening company-level context.
Workflow: Structure questions around inference spend, cost per outcome, margin exposure, forecasting, and escalation thresholds.
Product arm: TokenOps
Current stage: Discovery · pilot conversation.
Intended proof: A common evidence model can make compute economics reviewable across portfolio companies.
Next decision: Select a representative workflow and agree which economic signals warrant escalation.

RunTime · Product and retail workflow RunTimes

Design governed action paths for two distinct workflows.

Enterprise situation: Product-data teams and, separately, a retail operator are exploring controlled autonomous workflows.
Workflow: Governed product-data actions over MCP; separately, a retail-demand pilot with its own scope and decision path.
Product arm: RunTime
Current stage: In design · solution design and commercial scoping.
Intended proof: Each workflow can be modeled with explicit authority, evidence, exception handling, and auditability.
Next decision: Approve the product-data solution boundary and independently determine whether to advance the retail-demand pilot.

Where the pattern applies

A catalog of candidate workflows.

Beyond the selected work above, the same control pattern can be evaluated anywhere a workflow needs bounded authority, economic evidence, and an audit trail.

RunTime

Customer support

Order status, returns, exchanges, cancellations, refunds, address edits, damage claims, and product questions—with policy checks, recorded actions, and human escalation designed into the workflow.

order exceptions

RunTime

Order & fulfillment exceptions

Backorders, split shipments, failed payments, address mismatches, and fraud holds — handled against live order and customer state instead of a static rulebook, with every action logged and reversible.

ecommerce ops

RunTime

Promotion execution

Applying, validating, and reconciling promotions and discounts against policy in real time — with guardrails on dollar impact and escalation when a case falls outside approved bounds.

pricing & promotions

TokenOps + RunTime

Close & reconciliation

Structured, repeatable close and reconciliation steps encoded as a compiled workflow — with TokenOps providing the finance-ready cost and value evidence auditors ask for, and RunTime executing the approved steps.

finance ops

RunTime

Product data & syndication

Product data, retailer rules, and syndication workflows compiled into a deterministic graph — readiness validation, blocked-syndication diagnosis, and governed product actions exposed over MCP.

product experience

RunTime

Relationship & demand RunTimes

Durable customer-relationship processes that act against live state instead of reconstructing context from CRM on every request; and demand RunTimes that turn owned channels and purchase truth into a verified demand marketplace.

enterprise / advanced

Ask about your workflow