# AI agent budget planning worksheet

Preparation tool. This document is not a quotation and contains no Binov prices.

## Need and scope

- Task to improve:
- User and expected outcome:
- Tools and sources to connect:
- Permitted actions / actions requiring approval:
- Excluded scope:

## Usage assumptions

| Variable | Starting use | Expected use | Higher demand | Source or measurement needed |
| --- | --- | --- | --- | --- |
| Monthly requests | | | | |
| Processing steps per request | | | | |
| Document size | | | | |
| Share requiring review | | | | |
| Minutes per review | | | | |
| Retries | | | | |

## Development and operation

| Item | Assumption | Effort or volume | Actual unit cost | Estimated total | Included / excluded |
| --- | --- | --- | --- | --- | --- |
| Scoping | | | | | |
| Integrations | | | | | |
| Development | | | | | |
| Evaluation and corrections | | | | | |
| Launch | | | | | |
| Monthly hosting and storage | | | | | |
| Monthly processing | | | | | |
| Monthly human review | | | | | |
| Maintenance / improvement | | | | | |

Keep development costs and monthly costs separate. Review hours = requests × review share × minutes / 60. Do not add quantities to monetary amounts or compare estimates covering different periods.

## Questions before deciding

- Which assumptions are verified and which need measurement during a pilot?
- Who handles failures, tool changes and new uses?
- What criteria support continuing, correcting or stopping?
- Which costs are included, excluded or variable?
- Which capacity or quality improvement will be measured, and how?