200 people murdered in Benue, 50 killed in Plateau

Just because you pay for an AI stack doesn't mean it'll build you pipeline. The tools you pay for and the system you've built are 2 different things. Each level needs its own build. 1/ The Tool Operator: run 1 campaign manually. In the pipeline audits we run, I keep seeing teams run 1 campaign across separate tabs. Some still update the CRM by hand. → Find 25 companies that fit and 1 buyer at each in Clay. It has to work by hand for 25 people before it goes to 5,000. → Get each email from Prospeo, send from 1 inbox in Instantly and read every reply yourself. → Write down every step: the tool, what goes in and what comes out. → List every tool, inbox and seat you pay for, and cancel the idle ones. A head of marketing told us they had about 50 cold email mailboxes set up, and nobody was using them. → Connect what's left by MCP, API or webhook. 2/ The Workflow Builder: make every step run the same way. We track 70 signals and 37 triggers in Clay. → Give each fit rule and each signal points, add them up in a formula column and write the cutoff down. → Stop paying for email and phone data before a company passes that cutoff. Set every email and phone column in Clay to run at or above it. → Tag the rest "held" and re-run their signals each month. A company failing today can pass next quarter. → Run the email waterfall: Prospeo 1st, LeadMagic 2nd, FullEnrich 3rd, every miss going to the next. Across 1,000+ lists, it finds 85 to 90% of emails. → Have someone read a sample of the rejects and change the rule that cut any good account. 3/ The Context Engineer: put your playbook in 1 file. A Claude skill works like an onboarding doc for a new hire. On our team, 12 Claude Code skills run $200K+ a month in client ad spend. → Save your ideal customer, buyers, offers and rules in 1 context file. → Write a skill for every decision your team repeats. → Connect Clay, Prospeo and lemlist by MCP. → Run every skill on 10/80/10: a person decides the task, the skill does 80% of the work, and a person checks the last 10%. → Stop rewriting the copy when replies drop. It might be the last thing that's broken. 4/ The GTM Architect: build an engine that gets better every round. For AirOps, the ads and the outbound ran on 1 account list. That came to $7.83M in qualified pipeline in 10 months. → Put outbound, content and ads on 1 account list and 1 CRM. → Let Ergo write up every call. → Take people at target accounts who engage with your LinkedIn posts, enrich them in Clay and send them to lemlist. → Feed every reply, every call and every deal you win back in. The people who just bought tell you who to go after next. → Test 1 thing per round: the company, the buyer or the copy. Judge it on pipeline. So find your level first, and pick your next build: → Level 1: score every company and buy contact data at or above the cutoff. → Level 2: write the context file and your 1st skill. → Level 3: put ads and outbound on 1 list in 1 CRM. → Level 4: test 1 thing per round and log the result. Stop counting meetings that never turn into pipeline. Any tool can inflate that number. The scoring, the context file, the skill and the MCP commands are in the article (bookmark it for later).