A pilot has one job: to tell you, with evidence from your own team, whether one workflow is worth keeping. Not whether AI is worth it in general. Whether this workflow, run by these people, releases time without an unacceptable loss in quality.
The AI Workflow Pilot costs $5,000 and runs for four weeks. It happens only if you approve it, after an AI Opportunity Review has recommended a workflow and given you a pilot brief with a baseline, success measures, responsibilities, scope, and a fixed implementation quote. Here is what happens inside it, using proposal preparation as the example.
The boundaries come first
The standard pilot is deliberately small:
- One workflow
- One team of up to five users
- Up to two supported application connections
- Existing commercial tools, not a custom build
- Four weeks, measured against your own baseline
Custom applications, substantial data cleanup, complex integrations, sensitive workflows, and company-wide deployment need separate scoping. Software subscriptions and third-party usage charges are additional and disclosed before you commit. Connections, permissions, and software costs are confirmed before you commit too.
Small is the point. One workflow with a clear before and after gives you an answer you can trust. Five workflows at once give you a feeling.
The worked example: proposal preparation
This is the illustrative workflow from our agency page. It is not a client result. It runs from discovery call to reviewable draft in three steps.
01 · Approved inputs
Discovery notes, your proposal template, service descriptions, and approved examples. The draft is built from material you already trust, not an AI's idea of your offer.
02 · Structured first draft
Client objectives, proposed approach, deliverables, and the questions still needing an answer. That last part matters. If the discovery notes have no budget or timeline, the draft flags the gap for the owner to resolve. It does not fill it in.
03 · Your team's approval
Your team confirms scope, pricing, evidence, and delivery commitments, and resolves the gaps before sending. The review checkpoint confirms that a missing budget or timeline was flagged rather than invented. In the sample review, the human checkpoint is written as one line: the proposal owner checks scope, pricing, factual claims, and client details before sending.
Nothing in this workflow writes and sends a proposal on its own. It prepares a first draft from approved inputs only. Your team reviews scope, fees, claims, and commitments before anything goes to a client.
What gets built and handed to your team
- One configured workflow using existing commercial tools, with up to two supported application connections
- Reusable prompts or templates, workflow instructions, and human review checkpoints
- One team training session and a recorded walkthrough
Review is made explicit. Who checks accuracy, scope, pricing, and client commitments before a proposal leaves the agency is written down, so the workflow does not depend on one person remembering how it works.
How it is measured
The four weeks include at least two weeks of observed use, subject to access and team participation. Those weeks of real use are what the results report is built on.
Three things are measured against your current process: preparation time, corrections, and team adoption. Corrections matter as much as time. A draft that arrives faster but needs heavy rework has moved the hours, not released them. Adoption matters because a workflow two of your five people ignore is not the workflow you tested.
The decision rule is agreed before the pilot starts, not after. In the sample review it reads: agree a time-saving target before the pilot, and expand only if the target is met without unacceptable quality loss. Setting the rule first stops anyone from grading the result on a curve.
An illustrative example, not a client result: the review's baseline shows about five hours per proposal, and the pilot brief sets a target of 90 minutes released on each. Over the observed weeks, either the team's real proposals come in near that target with corrections the proposal owner accepts, or they do not. Both answers are useful. Only one of them justifies spending more.
At the end you get a results report comparing baseline and pilot performance, including quality and adoption.
What the pilot does not promise
Faster proposals and more wins are not guaranteed, and we won't promise them. Proposal win rates depend on more than drafting speed. We agree deliverables and measurement upfront; financial returns are not guaranteed.
What you are buying is a fair test with a written result. The decision then rests on evidence from your own team rather than on a vendor's estimate.
Expand, revise, or stop
The pilot ends with a handover guide and one of three recommendations.
- Expand. The target was met without unacceptable quality loss. You keep the workflow. Additional workflows are scoped and quoted separately.
- Revise. The result came close enough that changing the approach is worth more than walking away.
- Stop. The target was not met, or quality slipped too far to accept. You stop with the baseline and pilot figures side by side, not a hunch.
If you keep the workflow, AI Performance Support is available at $750 a month, optional and billed monthly. It covers a monthly performance review with usage and software cost checks, documentation updates as the workflow changes, and up to three hours a month of minor configuration changes or troubleshooting.
How you get to a pilot
Through the review. The AI Opportunity Review costs $1,500 and takes 10 business days, and it can recommend deferring AI instead of a pilot. If you approve a pilot after it, the review and pilot together cost $6,500. Each is a separate decision, and scoping is free. For the review step by step, see What You Get From a $1,500 AI Opportunity Review.
GradeCircle builds and runs six products of its own: AfroSynergy, CallerBridge, HealthSquire, Heyzl, Lexfora, and LukeAPI, two of them in beta. The assumptions in our models come from operating those systems, not from a framework. See what we run.
If your agency writes proposals every month, scope your proposal workflow: five questions, and you see a preliminary fit summary before sharing contact details.



