Fund the workflow. Measure the business result.
Nodes is enterprise AI software that uses company records, human judgment and outcomes to investigate an owned business question. The customer-specific knowledge graph belongs to your company as defined in the agreement. Finance evaluates the estimated business impact, full cost and work the team retains. A product walkthrough requires no dataset. Decision Replay is a separate historical test.
Define the business process and its owner
Start with one owner, one decision or workflow and an agreed measure. An illustrative question is whether to renew an underused software contract, reduce its scope or gather more usage evidence. Finance supplies costs; the process owner explains local dependencies. This example needs its own validation and is separate from the insurance production evidence.
Before access and analysis begin, the customer and Nodes define:
- the decision and population in scope
- the outcome finance accepts as measurable
- the agreed sources, checking frequency and permitted evidence
- the starting business measure, current cost, team effort and comparison period
- access, recommendation and action authority, with approval and exception owners
Agree acceptable task quality, its supporting evidence, exception limits and total AI and retained human cost, including correction, rework and adoption. Disclose missing baseline evidence. The owner uses these checks and later results to expand, hold or narrow delegated work; prior approval does not expand current authority.
The commercial unit is the named workflow from trigger through approved action and evidence, scoped as a business responsibility. Nodes does not charge per seat, user, token, model call, action, or credit. Additional agents, branches, and improvements inside the agreed process do not create separate usage charges. A separate process or material expansion is scoped and agreed before it is added. See the workflow pricing model.
Include the full cost of running the workflow
Ask for one scope that identifies included and separately funded work. Compare acting, waiting, gathering evidence and leaving the process unchanged over the same period. Operating costs and pricing units are different; the no-token-meter policy does not remove infrastructure or human work from the business case.
| Cost to review | What finance should establish |
|---|---|
| Workflow license | The agreed process, systems, operating envelope, support, and scope-change rules |
| Implementation | Integration, configuration, acceptance checks, change management and adoption; scope and responsibilities agreed separately from the brief |
| Infrastructure | AI use, hosting, model endpoints, storage and network costs for the selected Nodes Cloud, customer VPC or on-premises configuration |
| Retained customer effort | Review, exceptions, correction, rework, governance, monitoring, data stewardship and ongoing maintenance |
Show supported inputs, labeled assumptions, expected risks and sensitivity. Keep time freed, cash saved, revenue and margin distinct. After separately agreed implementation, compare actual business results, operating costs and team effort with the baseline over the agreed period. An estimated opportunity remains separate from a measured result.
What one production record shows
The public evidence comes from one Fortune 500 insurance carrier. It demonstrates the method in enterprise talent and cannot forecast another company or decision type.
| Finance question | Reference evidence | Limit |
|---|---|---|
| What changed in the live operating record? | $1.58M in Q1 2025 net savings, validated by the customer's CFO | Customer-attributed result that does not assign every operational change to Nodes |
| How much did the hiring loop compress? | 127 days to 38 days, requisition to hire | One deployment, with no promise that another program will match it |
| How much sooner did producing hires reach the first production milestone? | 109 days to 62 days to the first production milestone, a 47-day reduction | Median comparison within the reference evidence |
| What did one rigid filter put at risk? | $17.7M in annual production across 2,863 producing agents the filter would have rejected | Retrospective counterfactual from one carrier |
The Decision Traces paper documents the research method. The case study separates observed results from modeled counterfactuals.
Build the unit economics from your own baseline
A regression on the reference carrier's production records associated each day faster to the first production milestone with $54.35 more in annual production per producing hire. The same evidence set includes an approved 30-day planning reference of $1,357 in projected annual production per producing agent. Both figures are modeled references from one role at one company, not causal estimates or realized savings. A Decision Replay tests your own volume, outcome rate, delay cost, and value per outcome before anyone uses either figure in a budget case.
For a hiring workflow, finance can inspect this equation:
eligible decisions × outcome rate × value per outcome × measured improvement
Show how the case changes when volume, outcome rates, or attribution assumptions change. The hiring value estimator illustrates potential gross value using your inputs and scoped carrier references. It excludes investment costs and does not calculate net ROI or payback.
Finance workflows to investigate
The production evidence above comes from hiring at one insurance carrier. The following are illustrative workflow scopes for a discovery conversation, with their own evidence and validation requirements. They are not validated production use cases today.
- Software renewals: connect contract dates, application usage, and payments; prepare renewal actions for approval and track whether expected savings appear.
- Collections coordination: connect receivables, customer commitments, and account context; propose permitted follow-up work and measure actual receipts.
- Contractor outcome review: connect purchase orders, delivery milestones, and invoices; prepare exception workflows and track the contracted result.
Each program needs its own outcome, permitted evidence, baseline, policy boundaries, and named human owner. No hiring result transfers automatically.
Frequently asked questions
How do you separate contribution from coincidence?
Finance agrees the baseline, period and attribution rules before work begins. Inspect the Decision Trace linking evidence, human judgment, action and later result. Compare actual costs and retained effort as well as the business measure. Completed execution, observed improvement and causal value are separate claims; disclose unsupported attribution.
What if the carrier result does not transfer?
It is not assumed to transfer. Each new workflow needs an agreed evaluation, baseline, and acceptance criteria. Decision Replay can test repeated historical decisions when suitable linked outcomes exist. A product walkthrough does not require that dataset or commit you to a Replay.
Who controls the assumptions?
Your finance team reviews the calculation and its assumptions with the business owner. Nodes shows sources, gaps and sensitivity. The owner can question an estimate, compare another option, defer a decision or decline a change. Agreed approvals remain separate from permission to read data.
See what Nodes would take on
Bring one question and its accountable owner. Request a product walkthrough without a dataset. The first decision brief returns findings, gaps, options, estimated costs and next steps within 72 hours of complete agreed data access, after typical setup of 1-2 weeks. Implementation is separately scoped and scheduled.
When you are ready to test your history
Bring one repeated high-stakes decision and at least two years of linked decisions and measured outcomes. Before the clock starts, Nodes checks the volume, join keys, outcome coverage, and data quality. Your finance team defines the $1M evidence bar and calculation in writing.
The 72 hours start only after Nodes accepts the agreed deidentified dataset delivered under a mutual NDA. The Replay is read-only. No live decision or production workflow changes. The result is historical evidence, not realized savings. If the historical record falls short of the written bar, there is no production fee and you keep the analysis. Review the separate Decision Replay offer.