Where should a business start with AI automation?

Look for a workflow, not a tool

A useful candidate happens regularly, follows recognisable rules and consumes time moving or preparing information. Enquiry triage, onboarding checklists and document intake can fit this pattern. A rare task with changing judgement calls is usually a harder starting point.

Separate preparation from authority

Preparing a draft is different from sending it. Extracting a date is different from changing a contractual deadline. Define which steps AI may prepare, which it may execute and which always require a person.

Check the data and the connection

Ask where information comes from, whether it is reliable, who is allowed to access it and whether the software provides a supported API. A product having an API does not prove that a particular integration is available on your plan or suitable for your workflow.

Measure a baseline

Record the actual work: how often the task occurs, the time spent, rework, exceptions and the people involved. Use those observations to agree acceptance criteria. Do not substitute a generic industry savings claim for your own baseline.

Choose the appropriate level of help

• Essentials: AI tool selection, setup and training
• Operations: connected recurring workflows
• Telospath OS: bespoke business infrastructure

Questions to ask before committing

• What is included, and what counts as a scope change?
• Which software, usage and support costs recur?
• What happens when information is missing or AI is uncertain?
• Who reviews actions and owns ongoing operation?
• How will the workflow be tested before it touches real work?