Three plays to grow submission volume.

What it does

Builds the broker universe from Apollo (construction and energy practice titles at the top retail brokerages), scores every broker for fit against Shepherd's appetite, sweeps construction and insurance trade press for intent signals (project awards, practice hires and moves, carrier appetite changes), and drafts a short first-touch email tied to the signal. An underwriter approves, edits if needed, and sends from their own name.

Why it grows submissions

Wider: it surfaces offices and brokers that have never sent Shepherd a risk, at firms where Shepherd is already appointed. Deeper: it flags a known broker the week new construction risk is about to cross their desk, so the underwriter is first in the door again. No SDR headcount; the agents do the finding and the drafting, the underwriter keeps the relationship.

What is live on this site

• A real broker universe pulled from Apollo (names stay masked until you click Reveal, which spends a credit)
• Claude scores each broker 0 to 100 with a one-line rationale you can argue with
• A signal sweep over ENR, Construction Dive, and Insurance Journal, matched to firms
• One-click drafts in Shepherd's voice with approve and reject

How we know it worked

• New submitting offices per month
• First submissions from brokers who had never submitted
• Reply rate on signal-triggered touches (target 20%+; my own engine runs 25% reply, 5% meeting)

Who runs it, and what it costs

About two hours a week. The sweep runs on a schedule; an underwriter or BD lead clears the draft queue Monday and Thursday. Voice and appetite rules are edited in one file.

Apollo seat and credits (~$200/mo at 2,000 brokers), Claude (~$50/mo), sending tool.

First 90 days

1. Days 0 to 30. Universe of 2,000 non-submitting brokers at appointed firms. Underwriters sign off on voice, appetite rules, and the do-not-contact list.
2. Days 31 to 60. 20 signal-triggered touches a day from three underwriters' names. Measure reply, meeting, and first-submission rate by signal type.
3. Days 61 to 90. Wire to the CRM. Add appointment triggers: a newly appointed office gets a first-submission sequence automatically. Report wider and deeper movement.

What it does

Type a broker and a firm. The agent researches them live (the firm's construction practice, recent placements and posts, the office, the book they likely carry) and returns a one-page brief: who they are, what their book probably holds, which Shepherd products and risk classes fit, three talking points that pair speed with differentiated pricing, the objection to pre-empt, and the first question to ask. Saved to a link you can forward before a lunch, a call, or a conference.

Why it grows submissions

Deeper is a relationship game, and the underwriter who shows up knowing the broker's book wins the next submission. This turns thirty minutes of prep into thirty seconds, so every broker touch gets it, not just the national accounts.

What is live on this site

• Real research through Claude with live web search, not a template
• A structured brief with the sources it used
• A shareable, printable URL for every brief

How we know it worked

• Briefs generated per underwriter per week
• Accounts per broker over two quarters, briefed versus not
• Time from first meeting to first submission

Who runs it, and what it costs

Near zero. It runs on demand. The Shepherd facts it reasons from live in one file and take five minutes to update when a product or partner changes.

About fifteen cents per brief.

First 90 days

1. Days 0 to 30. Underwriters use it before every new-broker meeting. Collect what they change by hand; that is the next prompt revision.
2. Days 31 to 60. Add CRM context (prior submissions, declines, binds) and a conference mode that briefs an entire attendee list the night before.
3. Days 61 to 90. Ship it as a Slack command so it lives where underwriters already are. Measure accounts per briefed broker.

What it does

A prompt set that mirrors the bottom-of-funnel question: comparison, shortlist, objection. Run the audit and each prompt is asked to Claude with live web search; the engine records whether Shepherd is mentioned, in what position, which carriers are, and which sources were cited. Share of voice per run, and a trend across runs. For every gap, it drafts the page that would close it: a comparison page with FAQ schema, written to be cited.

Why it grows submissions

Julian's read is right: this does not move the top of the funnel. It moves the last step, where two similar quotes sit on a desk and brand, track record, and what the internet says decide. It is a small test with a kill criterion, and the data decides whether it earns a bigger budget.

What is live on this site

• A live audit across 12 bottom-of-funnel prompts
• Share of voice, Shepherd's rank, and the sources each answer leaned on
• Page drafts for the gaps, with FAQ schema ready to paste

How we know it worked

• Share of voice on the prompt set (mentioned / total) and Shepherd's rank when mentioned
• Citations pointing at shepherdinsurance.com
• Kill rule: if share of voice has not moved 15 points 90 days after the first six pages ship, stop

Who runs it, and what it costs

One hour a week: run the audit Monday, publish one page. The prompt set is a table; add a prompt when a broker asks a new question.

About $30/mo in model calls. Profound or similar only if the test earns it.

First 90 days

1. Days 0 to 30. Baseline. Publish three comparison pages and add schema to the existing product pages.
2. Days 31 to 60. Three more pages, plus risk-class pages (data centers, renewables) and the Procore and DroneDeploy claims studies as citable sources.
3. Days 61 to 90. Re-audit. Keep, scale, or kill on the share-of-voice rule.