<img alt="" src="https://www.smart-7-innovative.com/802971.png" style="display:none;">
Skip to main content
← All case studies
Case Study · Applied AI

A revenue role filled in 32 days, with 4.5 hours of executive time

FullFunnel ran its own hiring process on Perplexity Computer: 267 applicants, 249 graded with written rationale by an agent, six final working sessions, and one hire.
Client profile
FullFunnel (internal)
RoleChannel & Alliances Manager
ServiceAgentic hiring workflow
TimelineJune 26 to July 28, 2026
PlatformPerplexity Computer
WorkspaceGoogle Workspace, one Drive folder
Applicants267
Results
Requisition open to accepted offer
32 days

Requisition open to accepted offer

Against a 39-day SHRM median

267

Applications received and processed

249 graded with written rationale

4.5 hrs

Total executive time consumed

Second decision-maker met only the final six

~$1

Est. platform cost per applicant evaluated

About $250 in total

The Challenge

High applicant volume, scarce senior time

The role was a Channel & Alliances Manager: a new seat owning technology partnerships and a set of private equity portfolios, against a first-year target of $2M in closed channel revenue. It is the kind of role that attracts enormous applicant volume, and where resume keyword overlap is nearly useless as a filter.

The most expensive line item in hiring is not the recruiter fee. It is senior people sitting in interviews with candidates who were never going to get the job. As an applied AI GTM firm and certified Perplexity Computer implementation partner, FullFunnel decided to prove on itself that AI belongs in the operating core of a revenue function.

What FullFunnel Did

An agent that builds the foundation, so people can judge

In late June 2026, FullFunnel ran intake, grading, research, outreach, interview synthesis, decision support, and candidate communications inside Perplexity Computer. Scheduled agentic workflows, browser automation, and native Google Workspace integration meant the entire system lived in one Drive folder. No applicant tracking system, no integration build, and no engineering time.

The agent never made a hiring decision. It made every decision reviewable. Two decision-makers made the final call and overruled the system twice: the only 9-graded applicant was passed after one conversation, and the ranked slate was reordered on a standard no rubric captured.

Sound Familiar?

This is not a FullFunnel-specific problem

Hiring teams of any size run into the same signs:
01 Every open role brings more applicants than anyone can read with care.
02 Keyword screens filter resumes without explaining why a candidate passed or failed.
03 Senior leaders sit in early interviews with candidates who were never going to get the job.
04 Automation fails silently, and nobody knows which candidates were dropped.
What FullFunnel Delivered

Four components, assembled in a single afternoon

Then left to run every weekday at noon Eastern without supervision.
COMPONENT 1

A single source of truth

The job description was written to a Drive folder. The applicant tracker (identity, grade, written rationale, outreach status, errors, and interview outcome) was created thirty minutes later in the same folder. Everything read and wrote there.

COMPONENT 2

A scheduled grading agent

Every weekday at noon Eastern, the agent swept for new resumes, parsed them, researched each candidate beyond the page, scored fit from 1 to 10 against the job description, wrote a paragraph of rationale, and upserted the row keyed on email so re-applicants never duplicated.

COMPONENT 3

Qualified-only outreach

Candidates who cleared the threshold were queued for LinkedIn connection requests via browser automation, with send status and every failure written back to the tracker. When 3 of 17 outreach attempts failed on browser timeouts and connection errors, each failure was logged to a dedicated errors column with the profile URL and an instruction to send manually. Nothing was silently dropped.

COMPONENT 4

An interview intelligence layer

Recordings landed in the same folder as transcripts, then were read against the job description and an internal rubric to produce per-candidate assessments and a ranked slate with cited evidence.

249 graded

Each received a researched, role-specific written evaluation averaging 55 words.

32 scored 8 or higher

Only three applicants earned a 9, and none earned a 10.

14 first-round interviews

Conducted July 6 to 27 and transcribed into the same folder.

6 final working sessions

Finalists built and presented a real 90-day channel plan.
The funnel 267 applicants in, one hire out, and every one of them read.
Outcomes

What changed

Filled in 32 days

SHRM puts the median time-to-fill at 39 days for nonexecutive roles. This search closed in 32.

70% less executive time

The second decision-maker spent ~4.4 hours in final sessions, against a conventional 12 to 17, and read no resumes.

Interviews expanded, not compressed

21 days on interviews against an 8 to 9 day benchmark, plus a working-session round most processes would skip.

About 1.4% of a recruiter fee

Roughly $250 in platform cost, against a contingency fee of about $18,000 for a role at this profile.
SHRM average FullFunnel
Job posted to screening started: 8 days Same day
Screen applicants: 8 days Continuous, every weekday at noon
Conduct interviews: 9 days 21 days, deliberately longer
Final decision to offer: 6 days 1 day
Total time-to-fill: 39-day median 32 days
What Is Next

Ready for the next requisition.

Unlike a recruiter fee, none of this disappeared when the role closed. The workflow, the rubric, and the tracker are still in place, ready for the next requisition at near-zero marginal cost.

The automation bought back time at the top and bottom of the funnel, and every hour of it was reinvested in human judgment. That is the motion FullFunnel builds and runs for clients.

Ready to put AI in the operating core?

Thirty minutes to understand where you are, what you are trying to build, and whether we are the right partner to implement and run it, the way we ran this on ourselves.
Request a Consultation