Every GTM plan I've looked at this year has the same line in it: put agents on everything. Outbound, inbound, paid, content, events. It's the right ambition. But it skips the question that decides whether the money pays back.
Which job should the agent do in this channel?
Gartner expects AI agents to outnumber sellers ten to one by 2028, and its own advice to sales leaders is to ask "not simply where agents can be deployed" but what friction they remove and what capacity they create (Gartner, Jul 2026). Most frameworks I've read stop at the workflow. Budgets, though, get planned by channel.
So I map channels instead. Every channel gets an agent. What changes is how much of the job it owns, and where the person sits.
Two questions for every channel.
- Question 01How much leverage is there?The upside an agent can unlock.
- How many similar actions the channel runs
- How repeatable each action is
- How fast you learn whether it worked
- Question 02What does one miss cost?The price of a single bad action.
- Value at stake per touch
- Relationship and brand exposure
- Whether it can be undone
- Platform rules you could break
Put the two answers together and every channel lands in one of four zones.
The leverage map · starting placements
Go all in
Agents run it end to end. People set the rules and review weekly.
Agents draft, people approve
Agents do the research and the first draft. People approve what goes out.
Automate simply
Rules-based automation does the job. Save agent budget for the zones above.
People lead, agents prepare
People own the moment. Agents do everything before and after it.
Zone 01: Go all in.
High volume, fast feedback, and a single miss costs little. This is where agents pay back first, so give them the whole job.
Inbound is the clearest case. The classic Harvard Business Review study of 1.25 million leads found that companies contacting a lead within an hour were about seven times more likely to qualify it than those who waited even one more hour (HBR, 2011). An agent answers in seconds, at 2am, every time. Paid search execution (bids, budgets, ad variants) and content work (versioning, localisation, keeping pages fresh for AI search) follow the same logic: lots of small, reversible decisions with quick feedback.
- The agent's jobRespond, qualify, book. Adjust bids and budgets. Produce and refresh variants.
- The person's jobSet the rules, budgets and brand guardrails. Review results weekly.
Zone 02: Agents draft, people approve.
The volume is just as high, but one bad batch can cost you a domain, an account or a first impression with a buyer you've been chasing for months.
Cold email is the textbook case. The average reply rate across billions of cold emails is 3.43%, while the top 10% of campaigns clear 10.7% (Instantly, 2026). The gap is quality, not volume. And the platform sets a hard limit: Google asks bulk senders to keep spam complaints below 0.1% and never reach 0.3%, and Gmail has been rejecting non-compliant mail since late 2025 (Google). LinkedIn outreach and named-account ABM sit here for the same reason.
- The agent's jobResearch the account, pick the signal, write the draft, line up the sequence.
- The person's jobOwn the brief, approve batches, review a sample every week.
Zone 03: Automate simply.
Some work is repetitive but small: scheduling social posts, event logistics, webinar reminders. A simple rule or workflow does it well. There's no judgment for an agent to add, so this is where you save agent budget and spend it in Zones 01 and 02.
- The automation's jobRun the same steps on schedule.
- The person's jobSet it up once and check it monthly.
Zone 04: People lead, agents prepare.
Few actions, high stakes: executive meetings, late-stage deals, field events, partners, community. These are the moments buyers want a person in the room.
Gartner found that 69% of B2B buyers turn to sales reps to validate what AI tools told them, and that reps matter most when buyers define the problem, choose a supplier and finalise the purchase (Gartner, May 2026). That makes agents more useful here, not less. They make sure the person walks into every one of those moments fully prepared.
- The agent's jobPick who to meet, build the account brief, map the buying group, send the follow-up within the hour, update the CRM.
- The person's jobThe conversation, the judgment, the relationship.
Same channel, different zone.
Here's the part most maps miss. The zone depends on who you're selling to, not only the channel. McKinsey's research on high-growth B2B teams describes exactly this split: agents run outreach end to end for lower-tier leads, while for high-value accounts they work as copilots and sellers make the calls (McKinsey, Jul 2026).
So score channel by segment. Cold email to small businesses and cold email to enterprise accounts are two different channels on the map.
The map moves every quarter.
A placement isn't a verdict. As your brief gets sharper, your data cleaner and your checks tighter, channels move toward Zone 01. Sometimes one moves back for a quarter while you fix something. Both are progress. Two simple rules:
- Move it forward whenThe agent has earned more of the job
- Weekly sample reviews pass for a month straight
- Results hold steady as volume grows
- The brief hasn't needed a big fix in weeks
- Move it back whenSomething needs fixing first
- Spam complaints climb toward 0.1%
- Negative replies per 100 touches rise
- A miss reaches an account that matters
In practice it sounds like this: webinar follow-up moves from Zone 02 to Zone 01 because a month of sample reviews came back clean, while LinkedIn outreach steps back to Zone 02 for a quarter because negative replies started creeping up. One channel earned more of the job; the other needs a sharper brief first.
The goal isn't the most agents. It's the right job for every one of them.
Draw your own map this week. List your channels, split the big ones by segment, answer the two questions for each, and place them. You'll see where your next agent dollar should go, and what the person in each channel should spend their time on.
Call it the leverage map.
Sources (6)
- Gartner, "Gartner Predicts AI Agents Will Outnumber Sellers 10 to 1 by 2028". 28 Jul 2026.
- Oldroyd, McElheran & Elkington, "The Short Life of Online Sales Leads". Harvard Business Review, Mar 2011.
- Instantly, "Cold Email Benchmark Report 2026". Platform data for 2025, published Jan 2026.
- Google, "Email sender guidelines FAQ". Google Workspace Admin Help.
- Gartner, "Gartner Survey Finds 69% of B2B Buyers Turn to Sales Reps to Validate AI-Generated Insights". 20 May 2026.
- McKinsey, "The future of B2B sales: How growth champions rewire their playbooks with AI". 16 Jul 2026.
Navneet Kumar
Building Neolism in the open.