The Four Jobs of Outbound

The replies stopped coming back, and the founder was the last person in the building to notice.

He'd spent the quarter celebrating his outbound numbers. Sends were up 9x. Open rates looked respectable. The pipeline dashboard was loud with green. Then he asked why his sourced-pipeline-to-meeting conversion had quietly cratered, and learned his AI SDR had spent 90 days addressing every prospect by their first name plus their LinkedIn headline, congratulating them on jobs they'd left in 2022, and pitching use cases that didn't apply to their business.

“Aw, shit.”

The replies still trickling in were almost all some version of "please remove me." His sender domain was sitting on three blocklists. The agency that had set the tool up was now charging him $4,000 a month to fix it.

He sent me a Slack message that read, in full: "I think I broke our outbound."

He didn't break it. The tool he bought broke it. He just signed the check.

A note before we go further. This issue runs longer than the usual 10-minute Wednesday read, and there's a reason. I run a consulting and coaching practice alongside this newsletter, and right now my inbox is doing something I haven't seen in 18 months of writing The Next Gear. It's almost entirely founders asking me to help fix broken go-to-market.

Not product. Not engineering. Not fundraising. Go-to-market.

Building the app isn't the bottleneck anymore. You can ship a working product in a weekend with the right tools. Vibe-coded code that needs cleanup? It's not hard to find a senior engineer contractor who will sort that in a sprint or two. The hard problem in 2026, the one quietly breaking otherwise-promising companies between $1M and $5M ARR, is figuring out how to find customers and convert them at a sustainable cost. So this week I'm going to spend more words than usual on the topic and walk you through the framework I take paid clients through in their first session. Use it however you want.

OK. Back to the founder who broke his outbound.

Before the framework, let me be honest about how he ended up in that hole. He bought the tool that promised the most automation, because the demo looked like magic. The pitch was that the AI SDR would handle every step. Find the prospect. Research them. Write the email. Send it. Track the reply. Repeat.

The pitch was correct in describing what the tool did. It was wrong in describing what the tool should be doing.

Outbound is four jobs. AI is exceptional at three of them, mediocre at the fourth, and catastrophic if you let it run that fourth job at volume. The companies still scaling outbound effectively in 2026 understand the split. The companies setting their sender reputation on fire have not figured it out yet. The tools won't tell you, because the tools are sold on the premise that more automation is more value. And that is simply not true.

Here are the four jobs.

When your business needs capital, timing matters.

Pinnacle Funding makes it simple to access fast, flexible capital without the delays of traditional financing. Apply in minutes, get approved quickly, and receive funding in as little as 24 hours. Real speed. Real results.

No fees. No credit impact. No obligation.

Get Funded Today.

There are no fees, charges, or obligations associated with obtaining a pre-approval. Pre-approval does not constitute a funding commitment.

The job AI does better than any sales team in history, and the one founders use it for the least.

AI is built for high-volume pattern recognition over messy text. That describes every outbound dataset you own. Your last 200 sent emails. Your last 200 replies. Your last 50 closed-won deals and 200 closed-lost. Your CRM notes. Your sales-call transcripts. Each of those contains the answer to a question your team can't find by eye, because no human is going to read that volume on a Tuesday afternoon and pull out a pattern.

I tell every client to start here. The work takes about 90 minutes and shows you exactly which parts of your outbound are working and which to rip out. The audits live in their own section below.

For now, the rule. Analyze is where the majority of your AI spend should go. If your tooling isn't helping you understand what already happened, it isn't helping you sell.

Strong fit. Not quite as strong as analyze, but close.

Research is everything you do before you write the message. Who is this person? What does their company actually do? What changed for them in the last 90 days that might make this conversation relevant? Have they raised money, hired aggressively, lost a key executive, shipped a competitor's integration, or posted a job spec that signals a problem your product solves?

AI is excellent at monitoring those signals at scale and feeding you a brief per account. The teams I work with that are still winning at outbound use AI to generate a one-paragraph research note per active prospect, refreshed weekly, with the trigger events highlighted. The model writes the brief. A human writes the email.

The mistake most founders make is letting the same agent that pulls the signals also generate the message. That collapses the discipline. A trigger becomes a line in a template instead of an insight in a conversation. You end up with "I noticed you raised your Series B in March, congratulations, and wanted to share how our platform can help you scale." That's a name-drop with a CTA dressed up as a personal note. The prospect knows.

Use AI to research. Have a human read the brief. Have a human write the email.

Mediocre fit. Use with strict guardrails.

This is where most teams overestimate AI. The model produces a passable first draft of a cold email. It cannot produce a good one without a human in the loop, and it cannot do so consistently across prospect types. Good outbound writing is judgment work. You have to choose which of the prospect's problems to lean on, what tone, what ask, and do it in three short paragraphs without sounding like every other vendor in their inbox.

Buyers in 2026 spot an AI email in under two seconds. The "I came across your work" opener. The "Quick question for you" subject line. The off-by-one personalization where the model pulled a stale company name from a bad data source. Every one is a flag in the buyer's brain that says "automated, ignore."

The rule for drafts: AI is the starting point, never the finish line. Every email gets a human edit, and that edit is the "would I send this to a friend who asked me for a favor" test. Typos are the least of what you're catching. If the answer to the test is no, the email doesn't go out.

This scales further than you think. A good SDR with AI assistance can edit and personalize 40 to 60 drafts a day. Aim for 50 good ones. That's plenty of pipeline at $2M ARR. The thousand-a-day target was always a vanity metric.

Catastrophic fit. The job AI should not be doing at volume, in any tool, ever.

Sending only looks easy. It's where domain reputation lives, where deliverability gets decided, where the buyer's filter rules and inbox placement and one-click-spam-report behavior happen. None of that lives in your CRM dashboard. All of it lives in silent infrastructure that decides whether your emails arrive at all.

When you let a tool send autonomously at volume, you are letting a system that doesn't know your domain history make decisions that will affect your domain's future. Most autonomous tools skip basic hygiene. They send too many emails to one recipient, or too many to one corporate domain in a short window, or to people who never engaged with prior outreach. Each of those choices trips filter logic somewhere.

What good looks like: hard caps on sends per rep per day. No autonomous send. Every email goes out from a real person's address with a real signature. Domain warming on any new subdomain you spin up. SPF, DKIM, and DMARC configured properly. A bounce-rate monitor that pages somebody when bounces cross 2%. None of it is glamorous. All of it is the difference between an outbound program that compounds over a year and one that lights itself on fire in a quarter.

If you remember nothing else from this issue: humans push the send button. Slowly.

Talk through your reply and get polished, professional text ready to paste. Wispr Flow strips filler, fixes grammar, and formats everything. 89% sent with zero edits. Works everywhere.

Try Wispr Flow free

Time for the Monday-morning move. This is the audit I run with new clients in their first week, and you can run it on yourself.

Pull a CSV of your last 200 sent outbound emails and your last 200 replies. Most platforms export it in two clicks. Then run four AI passes. Each one answers a different question. None of them require you to read the emails yourself.

A practical note before we go in. Nobody is pasting 200 emails by hand. The workflow is:

Your AI tool reads structured data without any formatting. The prompts below are written for Claude but work in any modern LLM. You don't need a fancy plan. Claude Pro at $20 a month accepts file uploads natively and handles 200 emails in a single pass. The free tier works too; you'll just want to batch in groups of 30 or 40. Paste each prompt into a fresh conversation so the model isn't biased by the previous audit.

What it tells you: the percentage of your outbound that reads as bot-written, with the worst offenders pulled verbatim. Most founders are shocked by the number. That's the point.

What it tells you: whether your bigger problem is targeting (wrong people) or messaging (right people, wrong pitch), and the exact words you should be writing against next quarter.

What it tells you: how much of your "personalized" outbound is actually fake personalization, and what real personalization would have looked like for the worst offenders.

This one is the most labor-intensive of the four because it requires pairing each email with the recipient's context. If your CRM exports the LinkedIn URL alongside the email send record, you're already most of the way there. Twenty pairs is enough to get the signal; you don't need 200.

What it tells you: what the few emails that worked have in common, what the losers are doing differently, and a cheat sheet you can hand your SDRs tomorrow.

Total time across the four: about 90 minutes including the cleanup. Total findings: enough to rebuild your outbound in the right order.

Audits give you the diagnosis. Here's the sequence to act on it.

Weeks 1 to 4: Triage. Kill autonomous send across every tool you own. Cap volume per rep at a number that hurts; a team sending 200 a day gets capped at 60. Use AI to clean the prospect list: feed your CRM in, have it identify dead leads (no engagement in 90 days), wrong-fit leads (industry or size mismatch), and warm leads with new triggers worth re-approaching. That last bucket is your starter pipeline for the rebuild.

Month 2: Rebuild research. Set up AI to monitor signals at scale: funding events, leadership moves, hiring patterns, product launches, new job specs that hint at problems you solve. Have it generate a one-paragraph brief per active account, refreshed weekly. The brief feeds humans who write the actual emails. Measure reply rate and meaningful-reply rate. Send rate is a vanity metric. Open rate is barely an honesty metric. Reply rate is the only number that maps to revenue.

Month 3: Scale cautiously. Now, and only now, expand volume. The tell that you've earned the right is sustained reply quality across at least four weeks. If reply quality drops as volume rises, stop. The math broke and you need to find out why before pushing further. Use AI to monitor reply-rate drift and flag it before your dashboard does.

The sequence is unsexy. That is on purpose. The founders winning at outbound right now are running disciplined operations. Discipline is the only secret.

A few things I tell clients that they always push back on, and that always work.

Apologize to your TAM. Email the prospects you've already spammed. Own it. Something like: "We were running an automated outbound program that wasn't matching the quality of conversation you deserve. We've shut it down. If we reach out again, it'll be because there's something specific you'd want to hear about. Sorry for the noise." A small number of those people will become customers, specifically because no vendor has ever done that for them. You will absolutely stand out. I've watched it work three times in the past 10 months.

Cut your prospect list by 90%. Most outbound failures are targeting failures dressed up as messaging failures. Smaller TAM. Deeper context. Higher hit rate. The companies winning between $2M and $10M ARR are sending less email to better-chosen prospects.

Make the founder write five cold emails a week by hand. Call it a recalibration exercise. The day you hire your first SDR is the day you start to forget what good looks like. Five hand-written emails a week keeps the muscle alive, and surfaces customer insights nobody else on the team is positioned to hear.

The founder I opened with rebuilt his outbound on this framework over the spring. He cut send volume 80%. He killed his AI SDR's autonomous mode and put a human in the loop on every email. He ran the four audits and found that 71% of his outbound had read as AI-generated to a third-party model trained to spot it. He apologized to the cohort he'd spammed. He cut his target list from 12,000 accounts to 800.

His reply rate went up sixfold. His meaningful-reply rate almost tenfold. The pipeline he's building now is smaller in count and dramatically larger in value, because he's actually talking to people who can buy.

I'm telling you this because the inbox patterns I'm seeing right now suggest a lot of you are about three months from where he was when he sent me the "I think I broke our outbound" message. The fix is the one he made. Decide which jobs you actually want AI doing for you, and protect the one you keep.

There are four jobs in outbound. Three of them belong to AI now and probably forever. The fourth one is yours. Don't give it away.