Free decision tool

Define one AI job before choosing a model.

Use this two-page card to record today's process, identify the data and consequences, and decide whether the next step is an approved service, a no-purchase local experiment, a structured assessment, or no test yet.

Download the two-page PDF PDF · US Letter · No email required

How the online version behaves. The online version works in your browser. Do not transmit or store answers. Printing or saving is controlled by the reader's browser.

Almiga Tech / Local AI for Small Business · Page 1 of 2 · Define

AI Workload Readiness Card

Start with one measurable job — not a model, server, or subscription.

Use this page to describe the work as it happens today. Unknown answers are useful: they show what must be resolved before a safe test.

1. Write the workflow in one sentence

When [trigger], [person or role] uses [approved inputs] to produce [specific output] for [recipient or system], reviewed by [qualified person] before [deadline or action].

Your workflow

2. Record today's process

Business owner

Qualified reviewer

Frequency

Tasks per week/month

Active staff minutes/task

Typical waiting time

Rework/correction time

Acceptable downtime

Safe fallback if AI is unavailable

3. Identify the data and consequences

Check every input class involved

Public Internal business Client/customer Employee Financial Regulated/contract-limited Unknown

May the business use these inputs for this purpose?

Yes No Not sure

Can a qualified person detect a bad answer before harm?

Yes No Not sure

Consequence if wrong or disclosed

Low Medium High

Must AI send, pay, delete, sign, publish, or alter records?

No Yes

Stop condition. If rights are unclear, consequences are high, or an action cannot be safely reviewed, stop and obtain qualified help before testing with real data.

4. Define a useful result

A useful output must contain

It must not

Improvement to measure

Accepted tasks Staff time Waiting time Rework Other

Current baseline

Test target

Page 1 checkpointIf the job, owner, reviewer, permitted data, fallback, or success measure is still unknown, mark NOT READY and resolve the gap before choosing technology.

Page 2 of 2 · Decide

Choose the Next Safe Test

Five stop/go questions turn the job definition into one bounded next action.

1 Can the current process or conventional automation meet the need?

Compare the job before adding AI. A template, search, rule, formula, or workflow may be cheaper and more dependable.

Answer Yes — test or improve that first; AI may be unnecessary. No / unclear — continue to question 2.
2 Are the data rights and approved route known?

Name what may be used and where it may travel or remain. Do not infer permission from convenience.

Answer Yes — record which data may use which route. No / not sure — stop; do not use real sensitive data.
3 Can a qualified human review the result before harm?

The reviewer must have enough expertise and time to detect material errors before the output is used.

Answer Yes — name the reviewer and include review time in cost. No — narrow the job, add controls, seek assessment, or stop.
4 Is there a named operator for the proposed route?

Where applicable, someone must own access, updates, monitoring, backup, recovery, support, and shutdown.

Operator or provider

Answer Yes — continue. No — do not self-host a business dependency; use a managed route or qualified help.
5 What evidence is required before purchase or rollout?
- Start with five safe cases: ordinary, difficult, exception, should-refuse, and bad/missing input.
- Compare the same job with today's process and a plausible conventional alternative.
- For a formal pilot, expand to at least thirty representative cases.
- Record the exact service/plan or model/software/machine/OS and test date.
- Test simultaneous use, outage behavior, fallback, recovery, and full operating cost.
Evidence you will collect

Select today's result

Not readyA required answer is unknown, data is prohibited, or consequences exceed available controls. Test an approved business serviceThe job is bounded and reviewable; managed operation matters more than controlling the model host. Run a no-purchase local experimentOne person, existing business-controlled hardware, safe data, human review, and no production dependency. Seek a structured assessmentData paths, permissions, multiple users, integrations, continuity, or operation need deeper work.

Record the next safe test

Candidate route or simpler alternative

Test owner and reviewer

Decision date

Safe cases/data

Maximum time and cash

Safe fallback

Stop immediately if

Do not buy hardware or widen access based on this card alone. Test the exact job and route. An unresolved answer means NOT TESTED, not a conditional pass. This card does not certify security, privacy, compliance, or business value.

Printing opens your browser's print dialog only. Clearing asks for confirmation first and removes only answers you typed in this browser.

Want dated updates?

Optionally receive corrections, publication news, workshop dates, and new test results from Almiga Tech. The Readiness Card, sources, corrections, and essential safety updates remain available without subscribing.

Email address

First name (optional)

Consent statement naming the selected email provider and linking to the privacy notice appears here. [provider not yet selected]

Is your result “seek a structured assessment”?

Almiga Tech can apply the framework to one bounded workflow and return a written, route-neutral recommendation. The result may support a service, local experiment, hybrid design, conventional alternative, narrower scope, deferment, or stop.

Request an AI Fit Assessment

Define one AI job before choosing a model.

AI Workload Readiness Card

Choose the Next Safe Test

1 Can the current process or conventional automation meet the need?

2 Are the data rights and approved route known?

3 Can a qualified human review the result before harm?

4 Is there a named operator for the proposed route?

5 What evidence is required before purchase or rollout?

Want dated updates?

Is your result “seek a structured assessment”?