GVGanivra

Customer Support Platforms

Resolve the ticket. Protect the margin.

A fixed-fee support contract earns the same revenue whether an agent resolves a request in two calls or twenty. See where repeated attempts, growing conversation context, and tool calls make customers expensive to serve.

The margin risk

More attempts. More context. The same contract fee.

An agent searches again, generates another answer, and passes an even longer history into the next call. Each step may succeed technically while the customer problem remains unresolved. Aggregate provider bills hide which accounts and workflows drive that spend.

Resolution attemptLookup & generationRepeated stepsCustomer cost

Attribute

Group the work behind a resolution

Send a stable customer_id for the account and one execution_id for the resolution attempt. Link model and MCP steps with step_id and parent_step_id so their costs can be inspected together.

Investigate

Find where the agent keeps spending

Use the margin-risk queue to prioritize recorded executions, then inspect repeated steps, input and output token usage, and tool costs. Compare expensive attempts with typical runs of the same workflow.

The execution trail helps you investigate possible loops; your application supplies the step relationships and resolution outcome.

Connect

Compare account costs with contract revenue

Configure customer pricing or import confirmed revenue against the same customer ID. Review costs and revenue over the same contract period to identify accounts whose support usage puts contribution at risk.

Keep a monthly contract fee at customer level instead of repeating the full fee on every ticket.

Act

Test a cheaper path to resolution

Use the findings to test shorter context, fewer repeated lookups, a different model, or an earlier human handoff. Compare tracked costs after the change alongside resolution quality in your support system.

Set retry budgets and escalation rules in your application. Ganivra provides the economics to evaluate those decisions.

Illustrative contract economics

A busy account can erase a fixed-fee margin.

Monthly contractRevenueTracked model + tool costContribution after tracked costs
Efficient support workflows$500$150+$350
Repeated, context-heavy attempts$500$650−$150

Invented figures illustrate how usage affects contribution. Human support, hosting, and other operating costs are excluded, so this is not the complete contract gross margin.

Integration essentials

Send usage facts with support context.

Which account and workflow?customer_id + workflow

Use internal account IDs and consistent workflow labels, such as ticket_resolution. Keep names and ticket content in your support platform.

Which steps belong together?execution_id + parent_step_id

Give each step its own step_id and link it to its parent. Send separate MCP events for tool usage and model events for downstream generation.

What drove the bill?input_tokens + output_tokens

Report actual token usage and known tool costs. Avoid counting a downstream model charge again inside an MCP tool cost.

Start with visibility, then tune the workflow

Review the margin-risk queue as telemetry arrives and the dashboard refreshes. Use execution details to investigate the cause, then apply limits or handoffs in your support agent.

Connect contract revenue →

Start with one support workflow

See what a resolution costs your business.

Connect model and tool usage, verify customer attribution, and add contract revenue to understand the economics.

Open integration guide →