Outreach your company can see, check and run
What changes when finding new buyers moves from one person's AI to a system the company runs: the machine finds and drafts, your people decide who counts and who hears from you, and you make one decision a day.
Ready now. 6 October 2026
Build thisHow outreach runs now
If your outreach starts with a list, someone picks names by job title, AI drafts the emails, and they go out in volume. Each email is faster to make than it used to be. But nobody decided who on that list actually needs what you sell, nobody checks what went out, and the replies keep falling.
1.4%
the median reply rate to cold email across 2.1 million emails from 480 business-to-business teams, down from 2.1% in 2024. Warm introductions in the same data got 27%. (KnowledgeNet, May 2026, from teams using its own tools, Q3 2025 to Q1 2026)
What sales leaders are seeing
- More agents won't fix it. Gartner expects AI agents to outnumber sellers 10 to 1 by 2028, yet fewer than 40% of sellers to say agents made them more productive. Its warning: "If those systems are fragmented, the agents will scale the fragmentation." Gartner, July 2026
- The difference is the system, not the tool. A quarter of sales organizations get a return of 50% or more on AI; one in five lose 50% or more. And 72% don't put the time AI saves back into real selling. Gartner, survey of 210 sales leaders, January to February 2026, published May 2026
- A person in the path still wins. In the same 2.1 million emails, a warm introduction got 27% replies against 1.4% for a cold one. KnowledgeNet, May 2026
If your outreach starts with a bought list, the Before below is the old way. If AI already writes your emails, the chart shows what has to sit around it.
What it takes
Once it runs: what you see each week
Here is the whole revenue side of a worked-example company, seen from its head of revenue's seat. Each area of work has one number, a target and a person who answers for it. You see which are furthest from target, how much of each runs without a person, and one decision to make today.
Cobalt Air, a worked example. Tuesday 6 October. Your seat: Dana Ruiz, head of revenue.
Orange edge: furthest from target.
Decision for today
Outreach replies have fallen to 6 a week
Outreach wrote to 1,400 builders this week and 6 replied. Three weeks ago the same emails got 17. Most now go to contacts more than eight months old, and the agent keeps sending because nothing tells it the list has gone stale.
- Gap
- Far
- Trend
- Falling 3 weeks
- Also affects
- Attract
- Answers for it
- Sam Okafor
- Refresh the builder list this week. About 2 hours of Sam's time.
- Pause outreach until it is fresh. A week with no replies.
- Leave it running. Free now; replies likely keep falling.
Decision needed from you: which of the three.
Your eight areas of work, furthest from target first
What it runs on
Warning
Priya Shah answers for customer records, but technicians still write paper tickets she cannot see until someone types them in. She answers for something she cannot see.
Cobalt Air is invented: a heating and cooling contractor that sells to builders. Its people, numbers and tools are made up to show what the executive sees once the work runs.
Before: how outreach runs the old way
The old way buys a list and pays people to work it. A sales development rep (the junior seller whose job is first contact) pulls names by job title, writes from a template, now often with AI's help, and sends to everyone on the list. Nobody watches for the buyers who are saying, in public, that they need what you sell. And nobody writes down what a real buyer looks like, so it stays in the sales lead's head.
The same eight squares as the chart below, the old way, each marked by who does it. Squares nobody does are drawn empty and listed first.
What it takes
One sales development rep: a median base of $60,000 and median on-target pay of $86,000 a year. Plus the list, the contact data and the sending tool.
Where it breaks
Only 57% of sales development reps reach quota. A list bought by job title can't tell who needs you this month.
What nobody does
Watch for buyers who raised their hand, measure each step, and write down who a real buyer is.
What your people are now for
Your customer now
Raises a hand in public
The machine now
Finds, qualifies with reasons, drafts
Your people now
Decide who counts as a real buyer and what is worth saying, approve every message before it goes, and take the conversation. The machine works from their judgment; it does not replace it.
Before you build, check that you can
- See what it did
- Stop it
- Change its rules
- Give it time every week
- Do the work by hand if it fails
- Answer for catching mistakes, not just for speed
Two ways it goes wrong
- Approval turns into a rubber stamp when one person reviews too much at once.
- People who only watch lose the skill to step in. Keep the expert doing some of the work by hand.
When not to do this
- Nobody can own it every week. A system with no owner looks broken, even when it works.
- You can't yet say who a real buyer is, or what is worth saying to one. Write that down first; until then, the machine has nothing to run on.
- Your buyers don't say what they need in public. This finds people by what they post, comment and change; if your market is silent, it finds no one.
Where the work lives
Every decision lands in one place your team can open: who was found, why each person passed or stopped, who approved each message, and every reply, on the buyer's own record. The rules it works from are written down, with the date and reason for every change. Nothing lives only in one person's chat window, so it keeps running when someone leaves.
Build this in your business
Starter gets you every redraw and its playbooks. Pro adds the tools and folders behind it. Expert teaches you to build it for your whole business.
Build thisFor whoever builds it
Everything below is for the person who will set this up: the case, the chart, the tools and the steps.
The redraw
Buying a list and blasting it now gets your mail turned away. Gmail, Yahoo and Microsoft hold bulk senders to a spam complaint rate under 0.3%, and mail from senders above it is filtered or rejected. Meanwhile buyers say what they need in public every day: they comment on posts about the problem, post jobs, change roles. A firm can now collect those people, check each one against what its sales lead knows about a real buyer, write down why each one passed or stopped, and draft a first message that quotes the buyer's own website. A person decides who gets it.
- The business
- A business-to-business service firm of 10 to 100 people that already runs a CRM
- The domain
- Revenue operations: finding new buyers
- What it's worth
- Replies, times the share that become meetings, times the share you close, times your deal size. Your own numbers.
- Who else offers it
- List vendors and outsourced sales-development agencies
The chart
Pick a seat to light the squares it works in. Squares marked along the bottom include a person's step.
What the RevOps expert writes in the middle
- Who counts as a real buyer
- Every reason to stop someone
- What may be sent, when, and who approves it
- What you measure, and the target
- Every decision, with its date and reason
Before you build, check that the accountable lead can
- See what it did
- Stop it
- Change its rules
- Give it time every week
- Do the work by hand if it fails
- Answer for catching mistakes, not just for speed
Two ways it goes wrong
- Approval turns into a rubber stamp when one person reviews too much at once.
- People who only watch lose the skill to step in. Keep the expert doing some of the work by hand.
0.3%
Why now
The spam complaint rate Gmail, Yahoo and Microsoft hold bulk senders under. Microsoft began enforcing its rules on 5 May 2025.
Put it in your business
Sources
- Bulk sender requirements for Gmail, Yahoo and Microsoft, and Microsoft's enforcement date (Red Sift)
- How Gmail filters and rejects bulk senders by complaint rate (GMass)
- "Gartner Survey Finds AI Saves Sellers Nearly 5 Hours Per Week, Yet 72% of Sales Organizations Fail to Reinvest Time in High-Value Activities" (Gartner press release, 19 May 2026; survey of 210 CSOs and senior sales leaders, January to February 2026)
- Median cold email reply rate 1.4%, down from 2.1% in 2024; warm introductions 27.0% (KnowledgeNet, "2026 B2B outbound email benchmarks", May 2026; 2.1 million outbound touches from 480 B2B teams using KnowledgeNet and partner workflows, Q3 2025 to Q1 2026)
- "Gartner Predicts AI Agents Will Outnumber Sellers 10 to 1 by 2028, Yet Fewer Than 40% of Sellers Will Say Agents Improved Productivity" (Gartner press release, 28 July 2026; a prediction, with a survey of 210 CSOs and senior sales executives, January to February 2026)
- Sales development rep pay and quota: median base $60,000, median on-target pay $86,000, 57.0% attain quota (RepVue, Sales Development Representative salary, 8,276 verified submissions; data updated 8 October 2026)
- Running-cost estimate, from public prices read 8 October 2026, for about 1,000 people a month (one message each, three mailboxes, 5,000 tokens read and 400 written per message): finding people, PhantomBuster Start, $56 a month billed annually
- Running-cost estimate: work emails and checks, Hunter Starter, $49 a month for 2,000 credits (1.5 credits to find and check one email)
- Running-cost estimate: AI writing, Claude API, Sonnet 5.5, $2 per million input tokens and $10 per million output, about $0.014 a message
- Running-cost estimate: sending, Instantly Growth, $47 a month for up to 5,000 emails
- Running-cost estimate: mailboxes, Google Workspace Business Starter, $7 per mailbox a month, three mailboxes
- Running-cost estimate: one data platform in place of finding people and work emails, Clay Launch, $167 a month (the same page also lists $185)