# Agentic Operating Model

How your organization runs once Clones do real work on their own, covering who decides, who monitors, how work flows, and how the model evolves.

## What this is for

When Clones only draft and recommend, the existing org chart still works. When they act on their own within limits and coordinate workflows across departments, it does not. Someone has to own each Clone's work, set and review its limits, watch for exceptions and decide what changes. This document defines that operating model: roles, decision rights, work design, monitoring, and how it all evolves. It is the bridge from a delivery project to the way your organization works.

## When to use it

Draft it at the start of Month 7, once the first workflows reach Execute within limits. Finalize it by Month 9, alongside the CoE Charter. Review it at every quarterly business review after that.

## Who's involved

- **Executive sponsor** owns the model.
- **Business owners** confirm their accountabilities.
- **InfoSec lead** confirms control and monitoring roles.
- **CoE lead** takes over maintenance.
- **HR lead** advises where roles and job designs change.

## How to fill it in

1. Agree the principles (section 1). Keep the five; add your own if needed.
2. Fill in the accountability map (section 2) for every live Clone.
3. Set decision rights (section 3) using the RACI.
4. Redesign two or three priority workflows end to end (section 4): what the Clone does, where a person decides, what is orchestrated.
5. Define the monitoring rhythm (section 5).
6. Note role and capacity changes for people (section 6), with HR.
7. Agree how the model is reviewed and changed (section 7).

## The template

### 1. Operating principles

1. **People stay accountable.** Every Clone has a named business owner who answers for its work.
2. **Autonomy is earned and reversible.** Levels move up with evidence and dual sign-off, and down on agreed triggers.
3. **Limits are explicit.** Every Clone acting on its own has written thresholds and an escalation path.
4. **One governed platform.** Clones, skills, connections and adjacent AI tools are inventoried and governed together.
5. **Capacity goes somewhere.** Time given back is redeployed deliberately, and we say where.

### 2. Accountability map

| Clone | Department | Business owner | Escalation owner | Champion | Highest autonomy level | Workflows at that level | Review cadence |
|---|---|---|---|---|---|---|---|
| _Billing Clone_ | _Finance_ | _Controller_ | _AR Manager_ | _Priya N._ | _4 Execute within limits_ | _Credit memos within limits_ | _Monthly sample_ |
| | | | | | | | |

### 3. Decision rights

R = responsible, A = accountable, C = consulted, I = informed.

| Decision | Business owner | InfoSec lead | CoE lead | IT lead | Champion | Exec sponsor |
|---|---|---|---|---|---|---|
| New Clone in a department | A | C | R | C | C | I |
| New connection or data scope | C | A | R | R | I | I |
| Autonomy promotion | A (joint) | A (joint) | R | C | C | I |
| Change limits within a level | A | C | R | I | C | I |
| Emergency pause of a Clone | R | A | R | R | I | I |
| Skill version upgrade in production | C | C | A | R | I | – |
| New orchestrated workflow (Commander) | A | A | R | C | C | I |
| Retire a Clone | A | I | R | C | C | I |

### 4. Work design: priority workflows

Repeat this block for each redesigned workflow.

| Field | Entry |
|---|---|
| Workflow | _Order-to-cash dispute resolution_ |
| Trigger | _Customer disputes an invoice (email, portal or phone)_ |
| Clones involved | _Support Triage Clone → Billing Clone → Collections Clone_ |
| Orchestrated by | _Commander_ |
| Clone steps (on their own) | _Classify dispute; pull invoice, PO and delivery proof; draft resolution; issue credit memo within limit_ |
| Human decision points | _Credits above limit; any write-off; legal language from customer_ |
| Limits | _Credit ≤ agreed threshold; existing customers; one credit per invoice_ |
| Escalation | _AR Manager within 4 business hours_ |
| Measures | _Days to resolve; % resolved without escalation; override rate_ |
| Before | _11 days median; 3 teams; 6 handoffs_ |
| After (target) | _3 days median; 1 handoff to a person on exceptions only_ |

### 5. Monitoring rhythm

| Cadence | What | Who | Output |
|---|---|---|---|
| Daily | Exception digest per department | Escalation owners | Exceptions resolved or reassigned |
| Weekly | Clone Health and failed tasks | Org admin / CoE | Fixes logged |
| Monthly | Sample audit of within-limits actions; Power Rating and usage review | Business owner + CoE | Sample results; tuning actions |
| Monthly | Override and incident review | InfoSec lead | Risk register updates |
| Quarterly | Autonomy map, limits and operating model review | Steering / QBR | Promotions, demotions, model changes |

### 6. People: roles and capacity

| Team | Work moved to Clones | Hours given back / week | Redeployed to | New or changed responsibilities | HR action |
|---|---|---|---|---|---|
| _Accounts Receivable_ | _Dispute research, credit memo drafting_ | _38_ | _Customer reviews with top 50 accounts_ | _AR Manager becomes escalation owner and sample auditor_ | _Update job description_ |

### 7. Evolving the model

- [ ] Model owner named: __________ (CoE lead by default)
- [ ] Changes proposed through the CoE intake and approved at the quarterly business review
- [ ] Any change to decision rights for promotions or connections requires InfoSec sign-off
- [ ] Model reviewed after any incident above agreed severity

## Good looks like

- Every Clone has a person accountable for it. No "the AI did it".
- Decision rights are explicit for pause, promotion and limits, and InfoSec holds the pause.
- Work is redesigned end to end, not automated step by step inside the old process.
- Monitoring is proportionate: more autonomy means more sampling, not less attention.
- Redeployed capacity is named and visible to staff, which keeps the capacity-not-replacement promise credible.
- Common mistake: writing the model for Clones and forgetting the people roles that change around them.

## How Clone helps

The **Solutions Architect Clone** drafts the accountability map and workflow designs from the Integration Architecture, autonomy history and Systems & Workflow Map. The **GRC Clone** checks decision rights against the Governance Charter. The **Project Clone** tracks the monitoring rhythm in the portal and reports missed reviews at steering. Orchestrated workflows are run and monitored in Commander.

## Sign-off

| Role | Name | Date | Decision |
|---|---|---|---|
| Executive sponsor | | | Approved |
| InfoSec lead | | | Approved |
| CoE lead | | | Accepted ownership |
| HR lead | | | Consulted |

## Related

- [Clone Governance Charter](/playbook/templates/clone-governance-charter)
- [Autonomy Promotion Evidence Pack](/playbook/templates/autonomy-promotion-evidence-pack)
- [CoE Charter](/playbook/templates/coe-charter)
- [Integration Architecture](/playbook/templates/integration-architecture)
- [Getting started with Commander](https://support.clone.org/articles/getting-started-with-commander)

---
From the Clone Solutions Playbook: https://solutions.clone.org/playbook/templates/agentic-operating-model
