How it works · Flow F18
Every rupee invested in an agent, shown against what it returned
Each month the Investment & ROI Ledger totals every cost, records observed and assisted returns, computes named ratios with confidence, applies guardrails and sends a statement with evidence links.
What the Monthly Agent Investment Statement flow does
The Investment & ROI Ledger answers three questions for every rented agent and squad: how much did I invest, what happened and how much can be credited to the agent, and should I continue, change or stop.
Costs include agent rent, usage credits, pass-through messaging and calls, ad spend controlled by agents, avatar and video generation, human supervision time and setup or migration. Returns include observed and assisted revenue, weighted pipeline, collections accelerated and cost saved. Time saved is shown separately and never inside financial return.
The ledger computes named ratios, payback, cost per outcome and — only where an experiment supports it — lift versus baseline or holdout, with confidence. If a guardrail metric is below threshold after the minimum sample, external-facing work auto-pauses or the budget is cut, the owner is alerted and a review task is created. Otherwise work continues within caps. A monthly statement goes to the client and agency with evidence links.
Ledger v0 (spend meter and caps) is planned for W1; v1 with monthly statements for W2; experiments and outcome contracts in W3.
Monthly Agent Investment Statement, step by step
Seven steps every month.
- 1
1. Capture all costs
**AI Workforce · Investment & ROI Ledger · AI FinOps · Metering.** Rent, credits, pass-through, agent-controlled ad spend, avatar and video, supervision time, setup.
- 2
2. Capture returns
**CRM & Sales · Finance · Marketing.** Observed and assisted revenue, weighted pipeline, collections accelerated, cost saved. Time saved is kept separate.
- 3
3. Compute named metrics
**Ledger.** `net_roi_ratio` (break-even 0.0), `return_on_spend_ratio` (break-even 1.0), payback, cost per accepted outcome, lift with confidence where an experiment exists.
- 4
4. Guardrail check
**Ledger.** Below threshold after the minimum sample?
- 5
5a. Auto-pause or budget cut
**Ledger · Human gate.** External-facing Actions pause, internal work continues, the owner is alerted and a review task is created.
- 6
5b. Continue within caps
**Ledger.** Work continues within daily and monthly caps.
- 7
6. Issue the statement
**Ledger · Notifications.** Day 1–3 of the month: PDF, portal and WhatsApp summary, signed with a hash and stored immutably, with evidence links. Agency clients see the agency's price.
Honesty rules
Per-outcome rows are observed or assisted only — never "caused". Incremental lift is a cohort-level fact from a final, valid experiment and is never attached to an individual sale. Ratios are named by metric id, never as a bare "ROI". Low-confidence numbers are labelled, and financial ROI reads unavailable when you supply no margin basis.
KPIs
| Item | Detail |
|---|---|
| KPIs | Spend accuracy vs invoices, statement delivery on time, attribution dispute rate, share of agents auto-paused |
Apps involved
Each step runs as a typed, permissioned Action inside one of these apps. Nothing happens outside the Action contract.
AI Workforce
Investment & ROI Ledger, AI FinOps and Squads produce the statement and enforce guardrails.
Learn morePlatform
Metering & Subscriptions accrues rent; Model Hub supplies inference cost records.
Learn moreCRM & Sales
Deals and stages provide observed and assisted revenue and pipeline.
Learn moreFinance
Billing and bank reconciliation provide payments and collections.
Learn moreMarketing
Ads connectors provide agent-managed spend and ROAS.
Learn moreQuestions about this flow
Is supervision time counted as a cost?
Yes, always. Leaving it out would overstate the return. You set the hourly rate.
Will the statement say the agent caused a sale?
No. Individual outcomes are observed or assisted. Incremental value appears only from a final, reviewed experiment, at cohort level.
What if there is no experiment?
The statement shows observed outcomes, assisted participation and full cost, and says incremental value is "not established".
What happens when an agent underperforms?
After the minimum sample, external-facing Actions auto-pause or the budget is cut, and the owner gets a diagnosis and a review task.
Can I dispute a line?
Yes. Disputes raised in the window get an auto-assembled evidence pack and a review; upheld disputes produce a credit note.
See a statement design
Ask us to walk through the statement layout and metric definitions.
Apps involved
Where this happens in One3
Platform
One governed foundation for every human, AI agent, integration and channel
ExploreCRM & Sales
A WhatsApp-first CRM where sellers and Digital Employees work the same pipeline
ExploreMarketing
Consented, measurable marketing across WhatsApp, email, search, ads and social
ExploreFinance
Bill, collect, pay, account, plan and comply on the records your business already keeps
ExploreAI Workforce
Rent ready-made AI Digital Employees that work for your team, with a monthly ROI statement
ExploreMore in How it works
- Tenant onboarding
- Lead to cash
- Campaign loop
- Issue to resolution
- Hire to retire
- Procure to pay
- Order to delivery
- Record to report
- Idea to release
- Goal to outcome
- Agent lifecycle
- Model improvement
- Pack upgrade
- DPDP data-principal requests
- AI incident
- Rent a Digital Employee
- Switch to One3 in a weekend
- Goal to verified statement