The lesson: people want to help. Reach out.
The situation
I spent months doing what most people do: find a role, tailor a CV, apply through the portal.
The real problem
More than 100 applications produced nothing. Better applications were not the fix. Every conversation that went anywhere had started with a person, not a portal.
The decision
Run the search like a product. One system, ideas tested against evidence, reviewed monthly, and effort moved from applications to people.
How to work it
Part 1: set up the project
One home for everything.
A Claude project holds the instructions and memory; one numbered folder holds every file. The instructions are a short index pointing to the rules, so no rule lives in two places.
Write down the facts that never change.
Dates, numbers and titles in one list. Every CV and letter is checked against it.
Put approval gates in the workflow.
Gate 1 is a one-page brief on the role and the framing; nothing is built until I approve it. Gate 2 is a screening check for AI-sounding writing after the CV and letter are built, with changes shown as a table. Nothing is ever sent automatically.
TIPThe gates are where you add judgement. Everything else can be automated.
Ban the phrases that make every application sound the same.
An audit found the same template sentences in most applications. They went on a banned list, and a script checks sentence rhythm and dash use before anything is saved.
Part 2: run strategy reviews like product reviews
List your hypotheses for why it isn't working, and test them.
Mine were writing quality, CV readability, pay expectations, the market, LinkedIn, execution discipline and the channel.
Review monthly, or sooner when rejections stack up.
Close each hypothesis with evidence. Writing quality was ruled out after an audit. CV readability was ruled out after about 60 applications on a rebuilt CV. The channel was confirmed: the conversations that moved came through people.
Turn findings into rules.
I searched two levels of role. An application only counts as a priority if a named person has agreed to act before it goes in.
TIPIf a review doesn't change a rule, it wasn't a review.
Part 3: get feedback from leaders
Ask people who read CVs for a living.
Recruiters and senior product leaders rebuilt my CV around accountability first, a career summary up front, and a professional summary that works as a two-minute pitch.
Fix the summary once.
It describes who you are and how you work. Tailoring happens in the role detail, never in the summary.
Part 4: reach out to people
State your value proposition in two sentences.
Map the market in rings.
Your core sector first, then adjacent sectors, then sectors where the skills transfer.
Aim for 6 conversations a week.
Meet people inside organisations, not only recruiters. Ask for advice, not a job.
TIPAsk two specific questions. A general "career advice" request is easy to decline. A specific question is easy to answer.
Leave every conversation with two new names.
That is how the network grows past the people you already know.
When the week is short, cut applications, not conversations.
Part 5: build the CRM in Notion
Model the data like a product.
Seven databases: Roles, Companies, People, Interactions, Interviews, Reports and a Weekly Scorecard. Companies are the parent of roles; people link to the company they work for, not the role.
Track conversations as stages.
Named, Asked, Booked, Held. Count conversations held, not messages sent.
Feed it automatically, decide manually.
Gmail, LinkedIn messages (exported every two weeks) and the calendar flow in. A weekly run searches for new roles. A Friday run lists the follow-ups due, and I choose which to send. Nudge after 14 days; recruiters and network contacts every 2 months.
What changed
More than 120 people reached. A CV rebuilt from leaders' feedback. A clearer view of my market: which sectors, which roles, which companies need what I do. One CRM in place of four spreadsheets and a dashboard. Almost everyone I asked made time.
Use this when
- You've sent a lot of applications and heard nothing back.
- You want AI to do the admin and research while you keep every decision and every message.