Shiv Brodie does not start every search at the one-yard line. She starts at the fifty. The difference is a stack of custom GPTs, a Perplexity workflow, and a personal rule that the system prompt gets sharper every week. Anyone with two hours and a willingness to iterate can copy the move.

Shiv Brodie (GTM recruiter at Metaview) joined Nolan Church and Siadhal Magos on 10x Recruiting (more episodes on the 10x Recruiting hub) for the second instalment of the "How I AI" series. The episode walks through the practical AI workflows Shiv runs daily: custom GPTs for JD analysis and Boolean strings, Perplexity for company intel, Lavender for outreach polish, Metaview snippets for multi-touch candidate engagement, and structured assessment templates that compound team calibration over time.

This recap is a copyable playbook. The mechanics of each tool, the prompts that made the difference, and the discipline that turns AI from a toy into infrastructure.

Starting at the fifty-yard line

Shiv's metaphor for the AI-enabled recruiter caught Nolan immediately. The legacy recruiter starts every search at the one-yard line, marching the field through manual research, screening, and calibration before they get to the meaningful work. The AI-enabled recruiter starts at the fifty. The first hour is already on insights, not on staging.

The difference is not the talent of the recruiter. It is the system the recruiter has built around themselves: a JD-analysis GPT, a Perplexity workspace for company intel, an outreach assistant for messaging, a rejection-email helper for the bottom of the funnel. Each piece compresses what used to be an hour of human work into a minute of AI work the recruiter reviews and ships.

The compounding benefit is that the saved hours go back into the part of the job that compounds: talking to candidates, talking to hiring managers, learning the market. The recruiter gets more reps where reps matter.

Custom GPTs as the unfair advantage

Shiv's first weapon is a tailored GPT built inside ChatGPT with a long, specific system prompt. The system prompt is the asset; the model is just the runtime.

The GPT takes a JD as input and outputs three things in 30 seconds. First, gaps the JD did not address (the unstated salary band, the unmentioned travel expectations, the missing tooling preferences). Second, candidate questions the hiring manager should be ready for. Third, Boolean search strings and alternative job titles for LinkedIn sourcing.

The output is not perfect. Shiv's discipline is to treat every imperfect output as a system-prompt improvement opportunity. If the GPT included salary in the search-strings output and Shiv does not want salary there, she adds "do not include salary or location" to the prompt that night. The next run is sharper. Six months in, the GPT operates closer to Shiv's actual taste than any junior recruiter ever would.

Perplexity for real-time company intel

For company intelligence, ChatGPT's training-data lag hurts. Shiv switches to Perplexity for real-time search. The query "give me a competitor list for [target company], including funding stage, headcount, and growth metrics" returns a cleanly cited list in under a minute.

Shiv organizes Perplexity into "spaces", topical workspaces with persistent custom prompts. The space for early-stage GTM searches has different defaults than the space for enterprise hires. The space remembers her preferences and surfaces the right level of detail without manual re-prompting on every search.

The combined motion: custom GPT for the JD-level work, Perplexity for the market-level work, then both outputs feed her sourcing campaign. The full prep that used to take half a day takes 30 minutes.

Multi-touch outreach with video snippets

The top of the candidate funnel is brutally noisy. Senior candidates report four-plus recruiter touches per week. The math says one outreach message does not get a reply; three to four touches with real personalization sometimes do.

Shiv's stack here: Lavender to coach each outreach in real time (open-rate prediction, subject-line guidance, best-practice nudges), and Metaview snippets for video-based touches. The Metaview snippet workflow: she records intake calls with the hiring manager, then clips a 30-second segment where the manager pitches the role in their own voice. That clip becomes touch #2 or touch #3.

A senior candidate who ignores the first cold email opens the second one when they see "watch the hiring manager pitch the role in 30 seconds." Conversion goes up because the candidate gets to evaluate the hiring manager before saying yes to a call. Inversion of the usual dynamic, and it works.

Structured assessment templates

Inside the interview loop, Shiv leans on Metaview's structured assessment templates. The template captures hard skills, soft skills, motivation, and potential flags in a unified format that all interviewers fill out the same way.

The discipline she models: disable the AI on the parts of the assessment where human judgment is non-substitutable. Things like raw intelligence, taste, executive presence are still recruiter-judgement calls. Things like "did the candidate describe their last project " are AI-extractable from the transcript. The template is built to lean on each appropriately.

The compounding payoff is team calibration. When every recruiter uses the same template, the team can compare candidates across loops and across recruiters with apples-to-apples scoring. The data on why candidates advance (or do not) builds up over time. That dataset is what turns a recruiting team into a recruiting function.

Rejection emails with warmth at scale

The bottom of the funnel is where most recruiters cut corners. The rejection email is templated, impersonal, and forgotten in 30 seconds. Shiv built a "Rejection Assistant" GPT that takes a few notes ("strong on X, gap on Y, was kind in the loop") and outputs a warm, personalized rejection in 60 seconds.

The candidate gets a thoughtful message that references the actual conversation, not a form letter. The candidate's experience of being rejected at Metaview is meaningfully better than at any competitor that templated the work away. Six months later, those candidates remember and recommend Metaview to friends.

The cost is two minutes per rejection. The payoff is brand and word of mouth that no other line item in the recruiting budget can buy.

Where AI gives recruiting teams use

Shiv's stack is illustrative. The pattern across teams getting real use is the same: capture every interview signal, build agents on top of the captured signal, iterate the system prompts every week, treat AI as infrastructure not a side experiment.

Metaview Notetaker captures the intake call and the candidate interviews so the conversation becomes structured input for downstream agents. AI Sourcing generates targeted candidate lists. AI Outreach handles the sequence-and-personalization layer that Lavender solves for at the per-message level. Reports tracks which assessments led to which hires so the templates compound. For the AI-augmented recruiter pattern in depth, see claude-for-recruiters, and for the sourcing-specific angle, see most-accurate-sourcing-coworker.

Numbers from Metaview's 2026 AI & Hiring Alignment Report, surveying 505 recruiting leaders and hiring managers across North America and EMEA. Shiv's playbook is what the optimistic 79% is actually doing with that optimism. The recruiters who build the stack and iterate the prompts are the ones converting AI from theoretical use into measurable throughput.

The operating shift

Three concrete moves any recruiter can run this week:

One: build the JD-analysis GPT. Start with a system prompt that asks for JD gaps, candidate questions, and Boolean strings. Spend 20 minutes. Iterate the prompt every Friday with the lessons of the week. Six months in, you have the most valuable proprietary asset in your workflow.

Two: clip the intake call. Take a 30-second segment of the hiring manager pitching the role and add it to your second outreach touch. Senior candidates open it. The conversion lift pays for the workflow on the first hire.

Three: standardize on a structured assessment template. Even if you are a team of one. The template compounds because the data compounds. Three months of clean structured notes is a calibration asset no spreadsheet can replicate.

The recruiters who internalize these three moves operate from the fifty-yard line, every search. That is the operating shift.