How to scope an AI pilot in Canada | Lab Super Intelligence

Contact: andres@opcelerateneural.com

Business / website:
Project contact / city:
Audience / reviewer:
Task / current process:
Intended output:
Authorised inputs / revision status:
Formats / languages / deadline:
Review criteria / budget constraints:

1. Pick a task and a decision.
Choose a repeated task that someone already performs. Write down who uses the output and what the pilot will help you decide: continue, revise, compare another setup or stop. A broad request to “add AI” is difficult to evaluate.

2. Establish a baseline.
Describe the current process, the staff effort it takes and the errors reviewers encounter. Use the same inputs when comparing the AI-assisted version. Record observations rather than assume a saving.

3. Prepare representative samples.
Include routine examples, incomplete information, conflicting sources and questions that should receive no answer. Use only material you are authorised to share. Agree on redaction, storage, transfer and deletion before access.

4. Define the system boundaries.
List approved documents, connected tools, read and write permissions, cloud services and local hardware. Decide which actions need human approval and what happens when a source or tool fails.

5. Write the evaluation before the build.
Select the checks that matter to the task: source accuracy, unsupported statements, refusal quality, review effort, latency and operating cost. Name the reviewer. A convincing demonstration is one input to the decision, not the evaluation itself.

6. Scope the handoff.
The proposal should identify the prototype, evaluation summary, configuration notes and next-step recommendation. Ask what deployment, monitoring, support and wider integrations would require if the pilot earns a next phase.

Arrange secure transfer before sharing confidential files. Project price, timing, deliverables and review are agreed before paid work.