A practical working resource
Compare AI ideas, expose missing evidence and choose a useful next investigation.
Use the guide, editable worksheet and fictional example together. Download the ZIP, open the HTML guide in a browser and work in a copy of the CSV template.
What is included
- Printable guide with an assessment framework
- Editable opportunity scorecard and fictional example
- Three contextual prompts and a decision record
Who it is for
Business leaders, operations teams and advisers turning an AI idea into a clear business decision.
How to use it
Complete the record with the people who perform and own the work. Keep evidence and unresolved questions visible. Use the prompts to challenge a draft, then have the relevant owners confirm the decision.
All worked examples are fictional. The material helps structure planning and review; it does not promise a particular outcome.
Using this resource
Choose a question worth investigating
Use this scorecard to compare a small set of AI opportunities before commissioning a project. A useful opportunity has a recognizable business problem, a workflow owner and evidence that a change would matter. Start with the work: delayed enquiries, inconsistent reporting, repetitive document checks or slow internal routing. Record the current process and the decision that an assessment should support. Avoid beginning with a preferred product and searching for something to apply it to.
Invite the workflow owner, someone who performs the work, a technology representative and the person who would sponsor a pilot. Review three to five opportunities in one session. Describe each in a sentence: “Help [person] do [specific task] using [available inputs], while [human decision] remains with [owner].” If the problem is a missing process or unclear ownership, fix that prerequisite before proposing automation.
Assess six dimensions
| Dimension | What to record |
|---|---|
| Business value | Name the delay, rework, service gap or capacity constraint. Identify a measure and the person who cares about it. |
| Feasibility | Describe the required action, available systems and exceptions. Distinguish drafting advice from taking an action. |
| Data availability | Identify the actual source, access owner and quality. Record whether representative examples can be used for evaluation. |
| Ownership | Name a sponsor, workflow owner and person who can resolve exceptions. Confirm capacity to participate. |
| Acceptable risk | Explain the consequence of a wrong answer, disclosure or failed action. Define where a person must review. |
| Testability | State how a bounded experiment could be compared with the current process and stopped or reversed. |
Use “promising”, “needs evidence” or “blocked” for each dimension. Add a reason and an evidence source, such as a process walkthrough or a reviewed sample. These are discussion labels, not a predictive score. A high potential value cannot cancel a missing access permission or an unacceptable failure consequence.
Fictional worked example
An example distributor routes internal service requests manually. The operations lead proposes a tool that suggests a category and destination to a coordinator. The coordinator still assigns the request. A walkthrough suggests repeated routing work, but nobody has measured the time spent correcting categories.
| Finding | Assessment and next step |
|---|---|
| Business value | Needs evidence. Establish handling time and correction frequency before estimating value. |
| Feasibility | Promising. Suggesting a category is a bounded task; changes to request ownership remain manual. |
| Data availability | Needs evidence. The systems owner must approve a representative, appropriately prepared sample. |
| Ownership | Promising. Operations owns the workflow; the coordinator agrees to review suggestions. |
| Acceptable risk | Needs evidence. Restricted requests must be excluded or routed without exposing their contents. |
| Testability | Promising. Compare suggestions with human decisions in a limited evaluation before live use. |
Decision: investigate, with a named owner for the baseline and data-access questions. The example supports a discovery task. It does not support claiming cost savings or approving deployment.
Your blank decision record
| Field | Complete before deciding |
|---|---|
| Opportunity and workflow owner | ________________________________________________ |
| Business problem and evidence | ________________________________________________ |
| Current measure and missing information | ________________________________________________ |
| Human decision retained | ________________________________________________ |
| Main dependency or stop condition | ________________________________________________ |
| Decision: investigate / defer / decline | ________________________________________________ |
| Next action, owner and review date | ________________________________________________ |
Open the CSV template and add one row per opportunity. Keep evidence and reasons beside the assessment. Use the example CSV to see the expected level of detail, then create your own copy. Review blocked items first, select the most useful next investigation and record why the other ideas can wait.
Use AI to challenge the shortlist
The accompanying prompt file can help identify vague scope, missing evidence or duplicated ideas. Supply only information approved for the AI tool you use. Ask it to identify gaps rather than invent ratings. Have the workflow owner check the output against real operations. The final prioritization remains a business decision supported by the recorded evidence.
More from Authentia
Explore Authentia resources or contact Authentia to discuss the next decision.