What to Ask Before Paying Someone to Automate Your Business
Before paying for an automation, ask the provider to explain what happens when the ordinary sequence breaks.

Quick answer
Ask how an automation handles duplicates, missing fields, disconnected accounts and changes halfway through the work. Establish data ownership, permissions, approval points and recovery before purchase. A successful demonstration should lead to a bounded test of ordinary work and failure handling, with a clear person responsible.
In this article
A demonstration may show a form creating a record, a notification appearing, and a task completing. Your business also needs an answer for a duplicate request, a missing field, a disconnected account, or a customer changing their mind halfway through.
The quality of an automation proposal depends on how clearly it defines the job, the limits, the people responsible, and the evidence that it works. These questions help an owner evaluate that proposal without having to become the person who builds it.
What exactly starts and finishes the job?
Ask for the sequence in plain language. What event starts it? What information does it use? What action will it take? What proves the action completed?
“We automate your leads” leaves too much open. A useful description might say that a new estimate request creates a record, assigns a staff member, and alerts them to missing information. Whether the customer receives a message or a quote should be stated separately.
Ask which decisions remain with a person. Sending a reminder, offering an appointment, confirming a booking, approving a price, and charging a customer carry different responsibilities.
If the provider cannot describe the finish precisely, it will be difficult to agree that the work is complete. Mapping one job through your existing tools is a useful starting conversation.
Which systems and permissions are required?
Have the provider list the accounts, subscriptions, integrations, and access needed. Ask which capabilities are available in your actual plan and account, rather than in a demonstration environment.
Separate each action. A connection that can publish a social post may not be able to change business hours. A connection that can read a customer record may not be able to update it. “Integrated with” is too broad to establish either capability.
Ask who authorizes access, how access can be revoked, and what happens when the connection expires. Confirm where information will be stored and which people and services can access it. Share only the information needed for the agreed job through an appropriate access method.
These are implementation questions for the provider to answer in a way you can assess. A recognizable platform name does not answer them on its own.
What happens when the job changes or fails?
Use a realistic sample to walk through the normal case and several relevant exceptions. What if the same request arrives twice? What if a required detail is absent? What if an external service is unavailable after part of the job has completed?
Ask how anyone will notice a failure and who can resolve it. A dashboard nobody checks is a weak notification plan. A retry that creates another booking is a new problem.
If AI makes an interpretation or recommendation, ask how uncertainty is handled and where human review belongs. Use our guide to planning for better AI to distinguish useful flexibility from depending on a capability that has not been demonstrated.
Agree on a way to pause the automation and a manual path for continuing necessary work. The fallback should be understood by the person who may actually need it.
What will it cost to operate and maintain?
Separate setup work from ongoing subscriptions, usage charges, monitoring, support, and later changes. Ask what the quoted amount includes and which changes trigger additional work.
Clarify ownership of accounts, workflow configuration, documentation, and business records. If you change provider, what can you access, export, or transfer? Which parts depend on a platform or managed service that cannot simply be handed over?
A useful handoff includes a plain-language description of the process, the responsible people, the relevant access arrangements, and how to recognize and recover from common problems. A video of the successful demo is only part of that picture.
What evidence will we use to accept the work?
Agree on the acceptance check before implementation. It should cover a representative complete job, relevant exceptions, and confirmation in the destination systems. Seeing a success label in one application does not prove the final action occurred elsewhere.
If the proposal includes time savings, establish the current effort, frequency, and ongoing supervision needed. Keep a projected saving separate from a measured result after the system is in use.
The most useful provider can explain these boundaries without burying you in terminology. Ask for a defined first improvement with evidence you can inspect. Marketomic's systems work starts with the job the business needs done and the people who remain responsible for it.