AI Automation & Workflow Discovery
Current process, automation goals, existing software, data sources, volume, and guardrails.
Reach for it when: Automating a manual workflow, or scoping where a model actually helps.
- sections
- 9
- questions
- 27
- required
- 5
- AI assisted
- 27
Every section, every question
This is the template as it ships. Edit the wording, reorder the sections, or delete what does not apply to your studio.
About Your Business
- Business name*Short text
- IndustryShort text
- Company sizeSingle choice
- Geographic marketSingle choice
- Primary products or servicesShort text
Current Process
- Describe your current workflow*Long text
- Which department is this for?Multi choice
- What is the biggest challenge today?*Long text
- What tasks are currently performed manually?*Long text
Automation Goals
- What would you like to automate?Multi choice
- Primary objectiveMulti choice
- What does success look like?*Long text
Existing Software
- Which tools do you currently use?Multi choice
- Additional software detailsLong text
- Do you currently use AI tools?Multi choice
Data & Integrations
- Where does the automation need to get data from?Multi choice
- Which systems need to be connected?Long text
- Security or compliance requirementsMulti choice
Volume & Scale
- Monthly transaction or request volumeSingle choice
- How many users will interact with the solution?Single choice
- Is this automation business-critical?Single choice
AI & Automation Preferences
- What type of solution are you interested in?Multi choice
- Preferred deployment modelSingle choice
- Human approval requirementsSingle choice
Budget & Timeline
- Desired timelineSingle choice
- Budget rangeSingle choice
Additional Context
- Anything else we should know?Long text
* marks a required question. Every question in this template has AI suggestions available while the client is writing.
How to write an AI automation brief
Automation projects fail for a boring reason: nobody wrote down what the process actually does today. Not what it's supposed to do — what it does, including the exception everyone handles manually and nobody documented.
This is the AI automation and workflow discovery brief in Coligna — nine sections, twenty-seven questions — and it's built around that one gap.
The nine sections, and what each one is for
About Your Business. Industry, size, market, services. Context for everything below, and a first signal of regulatory weight.
Current Process. The workflow end to end, the departments involved, the biggest challenge, and the manual tasks. This is the section the whole brief turns on — see the next heading.
Automation Goals. What to automate, the primary objective, and the definition of success. Success stated as a measurable outcome, not as "less manual work."
Existing Software. Current tools, additional software detail, and any AI tools already in use. What's already there constrains what you can build more than any requirement will.
Data & Integrations. Where the automation reads and writes, which systems it must connect to, and security or compliance requirements. Automation is mostly a data problem wearing a workflow costume.
Volume & Scale. Transaction volume, user count, and business criticality. The same automation at ten a day and ten thousand a day is two different builds — and criticality decides how much failure handling you're really quoting for.
AI & Automation Preferences. Solution type, deployment model, and human-in-the-loop expectations. The human-in-the-loop answer separates a serious brief from a naive one.
Budget & Timeline. Range and horizon.
Additional Context. The catch-all, and in automation projects it's where the undocumented exception finally surfaces.
Two questions clients answer badly — and two the brief doesn't ask
"Walk through the current process end to end." — insist on the real one. Clients describe the official process. The one that runs includes a spreadsheet someone maintains by hand and a step that gets skipped when it's busy. The automation has to handle the real one, and the helper text on this field says so because it's the most important answer in the brief.
"Where does a human stay in the loop?" Clients either want everything automated or haven't considered the question. Both answers need a conversation. An approval step that a person must keep is a requirement, not a limitation — and in regulated work it's often the whole design.
Two more worth asking in the kickoff call, even though the template doesn't ask them:
"What happens when it gets something wrong?" Every automation is eventually wrong. Who notices, and what do they do?
"Who owns this process today?" Automating someone's job without their involvement is how a technically successful project gets quietly abandoned.
From brief to project
A submitted brief arrives in your inbound queue, sectioned and typed. It becomes an estimate, and the accepted proposal becomes a live project still carrying the process walkthrough the client wrote — which is the document you'll be checking against in build. See how the full lifecycle works →
A PDF gets filled in once and filed. A brief-capture link needs no account from the client and starts the work at submission.
Frequently asked questions
What should an AI automation brief include?
The current process in detail, what specifically should be automated and why, which systems hold the data, the volume and criticality, and where humans stay in the loop.
How detailed does the current-process description need to be?
More detailed than feels necessary. This is the one section where thoroughness pays directly — an automation designed against an idealised process breaks on contact with the real one.
Do I need to know which AI tools I want?
No. The brief asks about tools already in use and preferences, not a technology decision. Choosing the approach is the agency's job, and it should follow from the process, not precede it.
What happens after the client submits it?
It arrives as structured answers, ready to review, estimate and convert into a project without re-typing anything. See pricing for what each plan includes.