Creative agencies · New business intake
AgencyFlow
Every inbound inquiry scored against the agency's own rubric, with the reasoning shown line by line and a tailored reply waiting on a human to approve.
See it run
The problem
Inbound inquiries arrive at wildly uneven quality, and the criteria that separate a real project from a time sink live in one person's head. When that person is busy, strong leads wait and weak ones absorb the same careful reply.
A form stores the inquiry. A CRM stores it more expensively. Neither helps the reviewer make a consistent, explainable call, so the first response is slow, the standard drifts, and nobody can say afterwards why a lead was chased or dropped.
The insight
The instinct is to hand the inquiry to a model and let it decide. That is the wrong split. Whether the work is a service you actually offer, whether the budget clears your floor, whether the timeline is physically possible: those are the agency's rules, not judgment calls, and a model that gets them wrong is worse than a spreadsheet.
So the scoring is deliberately hybrid. Sixty-five of the hundred points are deterministic and the model cannot touch them. The remaining thirty-five cover what genuinely needs reading comprehension: how clear the scope is, whether the work is strategically worth having, where the risk sits. Every point carries its reason, which is what makes a score arguable instead of oracular.
What it does
A ranked queue of inquiries where the score explains itself, and no reply leaves the building until a person has approved the exact words.
A score you can explain
One hundred points with every component and its reason on screen. Sixty-five come from the agency's own deterministic rules; thirty-five from the language judgment the model is allowed to make.
Rules the agency owns
Offered services, budget bands, and tier thresholds live in a settings screen rather than in the code. Change the rubric and the queue re-sorts against it.
Bounded AI
Model output is validated against a schema and score bounds before it is trusted. If generation fails, the lead survives intact and a structured fallback keeps the reviewer moving.
Approval that actually holds
Sending stays locked until the current draft is approved, and editing a message returns it to draft. The outbox is simulated on purpose: this build never sends external email.
How it works
Submit
A public inquiry form validates the brief at the boundary and preserves exactly what the client wrote.
Score
Service match, budget fit, and timeline feasibility are scored by rule. Scope clarity, strategic fit, and risk are assessed by the model inside a schema.
Review
The lead screen shows the tier, every scoring reason, the risk flags, and what the brief failed to say.
Approve
The reviewer edits the draft, overrides the recommended status if they disagree, and approves the exact copy that goes out.
Send
Simulated delivery writes a timestamp and an activity event, and records plainly that no external mail was sent.
Who this is for
A fit if
- Agencies of roughly 5–25 people where one founder or new-business lead reads every inbound inquiry by hand.
- Teams whose fit criteria are real but undocumented, living in one person's judgment rather than anywhere a second person could apply them.
- Shops with enough inbound that a slow first response has a measurable cost in leads gone cold.
Probably not for you if
It isn't for agencies whose work arrives entirely through referral and relationship, and it isn't a CRM. It owns one stretch, inquiry to approved response, and stops there on purpose: no proposals, no invoicing, no project management.
What this would take
AgencyFlow is Build-shaped: four to six weeks to a deployed system carrying your services, your budget bands, and your thresholds. If you are not certain intake is the leak worth closing first, that is the question an Audit answers.
Next project
PitchIQ →© 2026 Myles Smith. All rights reserved.
Designed & built by Myles Smith