Buying an AI licence is easy. The real value comes when it helps finish a task: an enquiry is ready to answer, a document is filed correctly, a report is complete. That is the outcome a proposal should cover.
So what does AI automation cost? There are two parts: the initial implementation and ongoing operation. Both depend on your process. These five questions make a proposal much easier to assess than a flat price per “AI agent”.
1. What should be finished at the end?
“Use AI to handle emails” leaves a lot open. A concrete goal is: incoming enquiries are identified, enriched with information from the CRM and prepared as draft replies for approval.
That defines what needs to be built and tested. It also makes the result measurable: how much manual work remains per enquiry? How often are corrections needed? Which cases still go to a person?
2. How much integration is involved?
A workflow within one system differs from a process spanning email, document storage, CRM and accounting. Existing interfaces, access permissions and data quality all affect the effort required.
Ask which connections the proposal includes. Who sets up access? What happens when information is missing or contradictory? And who makes adjustments when a connected system changes?
3. Which costs are one-off?
Implementation includes understanding the process, technical planning, configuration or development, integration and testing with your team. It also includes getting people started: employees need to know how the workflow works and when to step in.
At AI Værk, we divide implementation into clear phases. You receive a technical assessment and an agreed scope before implementation starts. We use existing tools where they suit the task and add custom development where it is needed.
4. What does ongoing operation cost?
Beyond software and model usage, costs can include hosting, monitoring, support and further development. Ask about expected usage and how any excess is charged. Clarify what ongoing support includes.
A low starting price tells you little if nobody is clearly responsible for keeping the process running. Compare proposals over the same period and with the same scope of service.
5. When does automation pay off?
Illustrative calculation: neither a price offer nor a customer result. All amounts and time savings below are hypothetical.
Take a process with 500 cases a month. If three minutes of manual work could be removed per case, that would free up 25 hours. At an assumed internal cost of EUR 40 per hour, that represents EUR 1,000 of available capacity a month.
Suppose implementation costs EUR 8,000 in this example and operation costs EUR 200 a month. The calculated monthly benefit is EUR 800. The initial investment would be recovered after ten months.
This calculation depends entirely on your inputs. Available capacity does not automatically mean a lower payroll bill. The benefit comes when your team uses that time to handle more work, respond faster or clear a backlog. Errors, rework and quality also belong in the assessment.
What a comparable proposal should include
- The specific process and expected outcome.
- The systems, interfaces and data sources included.
- Testing, human approvals and exception handling.
- One-off and ongoing costs, with their assumptions.
- Responsibility for operation, support and later changes.
With these details, you can compare the solution your team will actually use. Our AI process automation service explains how we take a first workflow into everyday operation.
Start with a process that matters
At Kirsch & Drechsler, the first workshops grew into an ongoing partnership covering knowledge management, a chatbot, automation and joint product development. That first step created a foundation the company continues to use. See our work with Kirsch & Drechsler.
Which process regularly takes up your team's time? Send us a description, the systems involved and a rough monthly volume. We will help you assess a worthwhile implementation and the costs to consider. Discuss your automation project.
