Keep AI work within defined limits.
Apply usage allowances and execution boundaries.
AI usage allowance · this periodIllustrative
82%of allowance used
Execution limits · every runStep limitPausesRepeated tool failuresStops
Revomo AI · Intelligence architecture
Scale commercial decisions.
Keep control.
Connect AI models, commercial context and decision engines with governed execution. Build on an intelligence architecture that can evolve with your business.
- Commercial understanding
- Model flexibility
- Governed execution
The architecture
One connected system for commercial intelligence.
Commercial context, AI models, decision engines and workflows work together under shared controls.
FIG 01 · System mapHow the architecture works
Entry points
Revomo Copilot
Solution Interfaces
Events
Schedules
APIs
Defined workflow
Agent reasoning
Identity · Permissions · Approvals · Audit
Execution
Workflow Orchestration
Defined workflows and agent steps
Agent Harness
Context · Reasoning · Tool use
Model Hub & Gateway
Model selection · Routing · BYOM
Authorized Tools & Actions
Read and update records
Execute decision models
Shared commercial foundation
Commercial Context
Shared entities, definitions and lineage
Decision Engines
Calculate, predict, simulate, optimize
Invoke agent
Return result
Model request
Response
Inputs & definitions
Entry points
Revomo Copilot
Solution Interfaces
Events
Schedules
APIs
Defined workflow
Agent reasoning
Identity · Permissions · Approvals · Audit
Execution
Workflow Orchestration
Defined workflows and agent steps
Agent Harness
Context · Reasoning · Tool use
Model Hub & Gateway
Model selection · Routing · BYOM
Authorized Tools & Actions
Read and update records
Execute decision models
Shared commercial foundation
Commercial Context
Shared entities, definitions and lineage
Decision Engines
Calculate, predict, simulate, optimize
Invoke agent
Return result
Model request
Response
Inputs & definitions
Entry points
Revomo Copilot
Solution Interfaces
Events
Schedules
APIs
Defined workflow or agent reasoning
Identity · Permissions · Approvals · Audit
Execution
Workflow Orchestration
Defined workflows and agent steps
Invoke agent · Return result
Agent Harness
Context · Reasoning · Tool use
Model request · Response
Model Hub & Gateway
Model selection · Routing · BYOM
Authorized Tools & Actions
Shared commercial foundation
Read and update records
Commercial Context
Shared entities, definitions and lineage
Inputs & definitions
Decision Engines
Calculate, predict, simulate, optimize
Execute decision models
Entry points
Revomo Copilot
Solution Interfaces
Events
Schedules
APIs
Defined workflow or agent reasoning
Identity · Permissions · Approvals · Audit
Execution
Workflow Orchestration
Defined workflows and agent steps
Invoke agent · Return result
Agent Harness
Context · Reasoning · Tool use
Model request · Response
Model Hub & Gateway
Model selection · Routing · BYOM
Authorized Tools & Actions
Shared commercial foundation
Read and update records
Commercial Context
Shared entities, definitions and lineage
Inputs & definitions
Decision Engines
Calculate, predict, simulate, optimize
Execute decision models
Commercial context & decision engines
Intelligence grounded in your commercial world.
Connect AI to shared business definitions, commercial relationships and decision engines, so reasoning draws on the context and calculations that matter.
Explore Commercial ContextCommercial Context Graph
CustomerAgreementProductCostPrice
Shared definitions & policies
Decision engines
CalculatePredictSimulateOptimize
Specialized engines, not the language model, produce the figures.Available to the agentBusiness meaning from the graph, computed results from the engines.
Model Hub, routing & AI Gateway
Evolve your models. Keep your commercial foundation.
Work with a collection of models through Revomo’s Model Hub, route requests to configured models, and bring your own model. Keep commercial context, tools and workflows connected as model choices evolve.
- Model Hub
- Work with a collection of supported models.
- Routing
- Direct requests according to supported configuration and routing policies.
- BYOM
- Connect a model you provide.
Explore model access and routing
Routing. Routing policies are configured in Revomo; the router applies them to each request, and a workflow or agent can be pinned to a specific model. The model interprets and drafts; commercial figures come from the decision engines.
AI Gateway. The gateway connects and mediates model access for the agent harness. Models run outside Revomo; permissions and approval policies govern commercial actions on either path.
Inference. Revomo-provided inference runs on Amazon Bedrock, a listed subprocessor in the Data Processing Addendum. A model you bring runs under your arrangement with its provider. Model support is confirmed for each configuration.
Illustrative · generic model labels
Task requestRoutingConfigured policy
Model Hub
Model AModel BSelectedModel CModel D
AI GatewayControlled connection to inference
Revomo-provided inferenceAmazon BedrockBring your own modelBYOM · your provider
Shared foundationContextToolsPolicies
The selected model runs on Revomo-provided inference.
The request runs on the model you bring.
Agent harness & orchestrationExplore Agents & Automations
Connect reasoning to governed execution.
The agent harness brings context, model interaction and authorized tools into an execution loop. Workflows coordinate defined steps and agent-assisted work under your policies.
Explore how agents execute
Context. The harness assembles the relevant records, rules and recent changes before the model plans.
Authorized tools. The tool list is filtered to the caller's role before the model sees it; each tool is marked read, draft or live.
Orchestration. A workflow follows its defined path and can invoke an agent step, then continue with its result. A live change is a lifecycle transition the approval policy adjudicates.
Limits. 30 steps, then the agent pauses and asks; two consecutive tool failures stop the run; abort is checked before each tool call. The loop is implemented on LangGraph.
Workflow orchestration · defined path
TriggerConditionAgent stepApprovalcheckpointPublish
Agent harness · chooses its next step
PermissioncheckContext→Model interaction↑↓Result←Tool action
Authorized toolsRun margin analysisreadSave draftdraft
Control, observe & improve
Operate with control. Improve with evidence.
Set boundaries for AI work, inspect execution and evaluate changes. Use reviewed feedback to guide improvement.
Inspect every stage of execution.
Follow the steps, tool calls and outcomes behind a run.
- ContextCommercial records and definitions loadedDone
- ModelNext step plannedDone
- ToolAuthorized tool callDone
- Action
- Read records
- Authorization
- Allowed for the caller’s role
- Result
- Records returned to the model
- ResultOutcome written to the run recordDone
Check changes before release.
Compare observed behavior with defined expectations.
Candidate changeNew agent version
Illustrative evaluation set · commercial-rule validation is a separate scope, in the details belowEvaluation casesAuthorized tool usePassPermission handlingReview
Release reviewWaits on the flagged caseImprove guidance through reviewed feedback.
Turn execution evidence into changes people approve.
Execution evidenceRun records, outcomesHuman reviewPeople approveFuture runsUse the new guidanceGuidancev3 → v4
Guidance changes only after approval · no automatic model retrainingPermissions and approval policies remain in force throughout execution. See security and governance
Explore controls and evaluation details
Usage and execution limits
- Usage quotas
- Configured quotas bound AI consumption. With your own model, model charges are billed by that provider.
- Execution limits
- 30 steps, then the agent pauses and asks. Two consecutive tool failures stop the run with an explanation. Abort is checked before each tool call.
Run records and audit
- Run records
- One per execution, with every effect it produced. An agent step keeps its prompt, reply and the tools it used.
- Audit
- Each agent tool call is written to the audit trail with the resource it touched. Transitions are logged against the person who signed.
Evaluations
- Agents
- Agent evaluations check an agent's behavior on defined tasks against expected results.
- Commercial rules
- A separate scope: rules, formulas and price lists are validated against cases with a known outcome. Passing a rule check does not establish overall agent quality.
Continuous learning
- Today
- Feedback and run records are reviewed, and an approved change updates the instructions agents use in future runs, versioned like any other change. Models are not retrained automatically, and nothing is learned across customers.
- Product direction
- Learning from decision history through a trace-derived context graph is a concept, not shipping functionality. It is drawn below.
Product directionLearning from decision historyConcept
Copilot · tool · engine · Copilot · publish
Relationships between
decisions and outcomes
Probabilistic
decision paths
A pattern qualifies
after evaluation and review
Approved guidance for the next run
review → evaluate → draft
tool family →
Raw values and identities
are removed before patterns form
Cost pass-through
Margin recovery
Renewal pricing
Customers
Agreements
Prices
Products
Costs
Transactions
Reconciles identities and entities · lineage from any figure to its record
Clusters events into tasks and workflows
Permission boundaries
Review cost update
Evaluate levers
Draft price action
Quote request
Price guidance
Approve
ERP
CRM
Price lists
Cost files
Agreements
Quotes
Events: cost updates, price changes, quotes, orders, renewals, approvals, publishes
Reinforces paths
Flags paths
that held
that failed
Observed pattern
Evaluation
Review
Approved guidance
inside your organization
Copilot · tool · engine · Copilot · publish
Relationships between
decisions and outcomes
Probabilistic
decision paths
A pattern qualifies
after evaluation and review
Approved guidance for the next run
review → evaluate → draft
tool family →
Raw values and identities
are removed before patterns form
Cost pass-through
Margin recovery
Renewal pricing
Customers
Agreements
Prices
Products
Costs
Transactions
Reconciles identities and entities · lineage from any figure to its record
Clusters events into tasks and workflows
Permission boundaries
Review cost update
Evaluate levers
Draft price action
Quote request
Price guidance
Approve
ERP
CRM
Price lists
Cost files
Agreements
Quotes
Events: cost updates, price changes, quotes, orders, renewals, approvals, publishes
Reinforces paths
Flags paths
that held
that failed
Observed pattern
Evaluation
Review
Approved guidance
inside your organization
Copilot · tool · engine · Copilot · publish
Approved guidance shapes the next run; each run feeds back which paths held and which failed
Relationships between decisions and outcomes · probabilistic decision paths · a pattern qualifies after evaluation and review, inside your organization
review → evaluate → draft
tool family →
Cost pass-through
Margin recovery
Renewal pricing
Raw values and identities are removed before patterns form
Customers
Agreements
Prices
Transactions
Products
Costs
Reconciles identities and entities · lineage from any figure to its record
Review cost update
Draft price action
Quote request
Approve
Clusters events into tasks and workflows · permission boundaries
ERP
CRM
Price lists
Cost files
Agreements
Quotes
Events: cost updates, price changes, quotes, orders, renewals, approvals, publishes
Observed pattern
Evaluation
Review
Approved guidance
Copilot · tool · engine · Copilot · publish
Approved guidance shapes the next run; each run feeds back which paths held and which failed
Relationships between decisions and outcomes · probabilistic decision paths · a pattern qualifies after evaluation and review, inside your organization
review → evaluate → draft
tool family →
Cost pass-through
Margin recovery
Renewal pricing
Raw values and identities are removed before patterns form
Customers
Agreements
Prices
Transactions
Products
Costs
Reconciles identities and entities · lineage from any figure to its record
Review cost update
Draft price action
Quote request
Approve
Clusters events into tasks and workflows · permission boundaries
ERP
CRM
Price lists
Cost files
Agreements
Quotes
Events: cost updates, price changes, quotes, orders, renewals, approvals, publishes
Observed pattern
Evaluation
Review
Approved guidance
Explore the intelligence behind your commercial operations.
Discuss how Revomo’s architecture fits your data, model choices and governance requirements.
How your data is used
Your data supports analysis, predictions and recommendations in your workspace and is not used to train third-party foundation models. Revomo-provided inference reaches those models through Amazon Bedrock. Where you bring your own model, it runs under your arrangement with its provider. Aggregated, de-identified data may improve the service, as the Data Processing Addendum sets out.