Back to all blogs
Blog Details

Subscribe to Punch! newsletter

Stay in the loop with updates, ideas, and resources.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
5 min read
Published
16 Nov 2026
5 min read
Published
16 Nov 2026

AI Made Outbound Ten Times Faster. That's Now Everyone's Problem

Last updated
16 Nov 2026
AI summary

‍

Contents

Every revenue leader we speak to has an agent now. LinkedIn is full of people offering a thousand leads for a comment, or claiming they've automated their entire sales team. Ask those same leaders what AI has done for pipeline, though, and the answers get thin. Their reps are doing more research, pulling more data and sending more emails. They aren't having more sales conversations, and they can't show the board a revenue number that moved.

We put that gap to three people who build agentic go-to-market (GTM) systems for a living at our recent discussion, Going Agentic: Inside the Shift in GTM: independent AI Growth Engineer Zeno Farhat, Principal GTM Strategist Sumair Shakir, and our Chief Commercial Officer Ben Robinson. They disagreed on plenty. They agreed on the uncomfortable part.

AI has made outbound dramatically faster to build. It has made it faster for everyone, so speed on its own no longer wins anything.

Speed is now the cheapest thing in outbound

A year ago, launching a campaign meant days inside enrichment tables that slowed down, errored out and needed refreshing every few hours. Sumair says he hasn't opened that kind of interface once this year. List building, persona selection, angle generation and scraping now run with far less hands-on work, and going from a client brief to a live campaign is about ten times faster. Zeno goes further: he builds workflows by talking, so his voice becomes the workflow.

That should be good news. In practice, it has created a new problem.

Everyone is reading the same signals

Data is so easy to get that everyone is using the same signals. A company announces an acquisition and receives the same "congratulations on the merger" email from a dozen vendors in a week. Same trigger, same data, same message, because it was all easy to automate.

"We're all creating our own problems here," Ben says. His fix is to go back to the questions most teams skipped once AI made sending free: what are we actually saying, what are we offering, and are we listening to the replies?

A Formula One car doesn't make you Schumacher

Sumair has a comparison for this. AI is a Formula One car. Anyone can buy one now, and owning it won't win you a single race.

The effect also depends on who is driving. A newer company still working out who its buyer is can use speed to test its way to a pitch that lands. A mature team that already knows its buyer and its pitch gets little from churning out fast variations, and often gets worse results. Speed rewards teams that are still learning. For everyone else, judgment is the scarce part.

Agents work when they have something real to read

With speed available to everyone, reliability is what separates teams. The systems the panel trusts with real prospects share one trait: the agent is never left to make things up.

Know whether you need a workflow or an agent

Zeno draws a line most vendors blur. A workflow runs fixed, predictable steps. An agent makes a judgment call, and judgment calls can go wrong. The setups that hold up combine the two, running fixed steps wherever the outcome is known and handing over to an agent only at the points where judgment is needed.

His inbound qualification flow shows how that looks. Every form submission goes to an agent that pulls its rules and context from a GitHub repository rather than from any single GTM platform. It enriches the lead through one API, scores it, and pushes qualified leads into HubSpot to trigger the follow-up. Two months in, lead quality is up roughly 50%, and he hasn't seen the agent do anything it wasn't built to do. The stack behind builds like this, five tools and one loop, is what Zeno shares in the GTM Field Kit, the free companion guide we built with the panel.

Grounded jobs are the safe jobs

Sumair trusts agents most with list building, list cleaning, signal sourcing and reply sorting. Each of those reads from real data, so there's very little room to hallucinate. The further an agent drifts from something concrete, the more you have to check its work.

Guardrails get written one failure at a time

None of these systems worked first time. Sumair's example is painfully ordinary: an enrichment API times out and returns "company not found", and a lazy quality check passes it because the company name field isn't empty. The fix is a rule that checks the name actually matches the domain. Small fixes like that stack up until the pieces click together like Lego.

The same logic applies to agents that talk to buyers. Sumair recommends grounding them in your own data, then role-playing hard conversations with them before they go anywhere near a prospect. He learned why the hard way, when he got a newly launched agent inside a popular GTM tool to curse him out. Ben's stress test in the Field Kit turns that kind of checking into a routine you can repeat before every launch.

Prospecting is going. The conversation isn't.

The "SDRs are dead" posts don't survive contact with people who run real sales teams. Zeno's view is that jobs won't disappear, but tasks inside them will. He doesn't expect any sales development rep to be manually prospecting in 2027, and a company that still lets them hasn't matured its systems. A well-built system can source, enrich and qualify a lead before a person touches it. What's left for the SDR is the call.

Operational roles change most. Sumair expects standalone tool operators and customer relationship management (CRM) admins to fold into a broader GTM engineering role. He pictures a GTM engineer running a pod, with SDRs inside it, some human and some agent.

AI can plan. It can't take the punch.

What agents can't do yet is sell. Sumair ran an AI voice agent pilot this year for a European client that didn't want to hire SDRs. The technology couldn't hear where an objection was coming from or the emotion behind it, so it couldn't answer it. He borrowed Mike Tyson's line to explain why: everybody has a plan until they get punched in the face. AI will hand you plenty of plans. The live conversation is the punch, and it still needs a person to absorb it.

Chris Muldoon, our CEO, sees buyers moving the same way. They're heading back to events and phone calls, looking for the context and emotional intelligence he doesn't think AI can offer.

The panel splits on how much reps should steer their agents

Ben wants trusted reps to have creative room to adapt their agents, inside clear guardrails, because tomorrow's sales teams will be smaller and need people who think for themselves. Sumair wants the agent's job tightly defined so reps spend their time on the phone, where they add value, rather than tinkering. Which one fits your team depends on how much you trust the people on your floor.

LinkedIn is working, and that makes it fragile

For Sumair, LinkedIn drives most results, in some campaigns up to 80% of them. Ben puts much of that down to the sender: a well-optimised profile, matched in seniority and discipline to the person on the other end, so an engineering manager hears from someone who speaks their language. The sender gets the message read. What earns the reply is the offer, which takes Ben back to his earlier question: what are we actually offering?

The risk is dependence. The most common job titles are being hit by every outbound team, and buyers are filtering them out. One policy or feature change could reset everyone's playbook overnight. Chris noticed LinkedIn has already started letting users flag posts and comments that seem like AI slop.

Email still earns its place too. Ben points to a robotics campaign getting strong results from email, while contact-centre buyers are hard to reach even on LinkedIn. Channel choice is a test you run for each market, and the answer won’t stay the same.

The next edge is a central brain

If everyone has the same tools and the same signals, the advantage moves to what only you know. Sumair and Ben both see companies building a central brain: one repository where sales, marketing, ops and HR context lives, so any team or model can draw on it without seeing every raw record.

Sumair says this was mostly theory six months ago. Ben sees smaller, tech-savvy companies already mapping it out, and the bigger platforms are catching up. You don't need anything exotic to start. Sumair suggests Obsidian, which is free to run locally, or Notion, which he thinks is underrated for this.

There's a practical upside too. Keep your logic and context somewhere any language model can connect to, and you can swap in a better model the week it launches without rebuilding a thing. Zeno's GitHub-based qualification flow already works this way.

Where to start

Sumair's answer is ideal customer profile (ICP) and persona selection. Take a title like "programme manager": it could sit in HR, IT or operations, and getting that right with manual rules means writing an almost endless list of exceptions. An agent working from real data handles it well, which makes it a low-risk place to learn. Once your ICP is set, his scoring rubric in the Field Kit takes the next step: who to contact first, and on which channel.

Ben's advice for anyone new is to understand why the agentic GTM movement exists, and what it's trying to achieve, before learning any platform. When you do pick one up, choose a tool that shows you every step of a build, so you can see how each piece connects.

What this means for your team

Three things change when every competitor has the same agents:

  1. Speed stops being an advantage. What you say, what you offer and how well you listen to replies decide whether outbound works.
  2. The SDR role narrows to the conversation. Prospecting moves to systems, and operational roles fold into GTM engineering.
  3. Your own context becomes the asset. Data your competitors can’t buy, stored where any model can reach it, is the part that compounds.

Buyers now choose who they hear from. Teams that keep using AI to send more will get faster at being ignored. Teams that use it to know more, test harder and free their people for real conversations will be the ones with a number to show the board.

Build your first agentic workflow with the GTM Field Kit

We built the free Going Agentic GTM Field Kit with the panel so you can put this into practice. It holds three tools for your first agentic GTM workflow, one from each panellist:

  • The stack, from Zeno Farhat: five tools and one loop you can set up in a weekend, from about $280 a month
  • The scoring rubric, from Sumair Shakir: decide who to contact first, and on which channel
  • The stress test, from Ben Robinson: check a workflow is ready before it reaches a buyer

Enter your work email below and the download starts straight away. If you'd rather hear the full debate, the conversation is on YouTube.

Every revenue leader we speak to has an agent now. LinkedIn is full of people offering a thousand leads for a comment, or claiming they've automated their entire sales team. Ask those same leaders what AI has done for pipeline, though, and the answers get thin. Their reps are doing more research, pulling more data and sending more emails. They aren't having more sales conversations, and they can't show the board a revenue number that moved.

We put that gap to three people who build agentic go-to-market (GTM) systems for a living at our recent discussion, Going Agentic: Inside the Shift in GTM: independent AI Growth Engineer Zeno Farhat, Principal GTM Strategist Sumair Shakir, and our Chief Commercial Officer Ben Robinson. They disagreed on plenty. They agreed on the uncomfortable part.

AI has made outbound dramatically faster to build. It has made it faster for everyone, so speed on its own no longer wins anything.

Speed is now the cheapest thing in outbound

A year ago, launching a campaign meant days inside enrichment tables that slowed down, errored out and needed refreshing every few hours. Sumair says he hasn't opened that kind of interface once this year. List building, persona selection, angle generation and scraping now run with far less hands-on work, and going from a client brief to a live campaign is about ten times faster. Zeno goes further: he builds workflows by talking, so his voice becomes the workflow.

That should be good news. In practice, it has created a new problem.

Everyone is reading the same signals

Data is so easy to get that everyone is using the same signals. A company announces an acquisition and receives the same "congratulations on the merger" email from a dozen vendors in a week. Same trigger, same data, same message, because it was all easy to automate.

"We're all creating our own problems here," Ben says. His fix is to go back to the questions most teams skipped once AI made sending free: what are we actually saying, what are we offering, and are we listening to the replies?

A Formula One car doesn't make you Schumacher

Sumair has a comparison for this. AI is a Formula One car. Anyone can buy one now, and owning it won't win you a single race.

The effect also depends on who is driving. A newer company still working out who its buyer is can use speed to test its way to a pitch that lands. A mature team that already knows its buyer and its pitch gets little from churning out fast variations, and often gets worse results. Speed rewards teams that are still learning. For everyone else, judgment is the scarce part.

Agents work when they have something real to read

With speed available to everyone, reliability is what separates teams. The systems the panel trusts with real prospects share one trait: the agent is never left to make things up.

Know whether you need a workflow or an agent

Zeno draws a line most vendors blur. A workflow runs fixed, predictable steps. An agent makes a judgment call, and judgment calls can go wrong. The setups that hold up combine the two, running fixed steps wherever the outcome is known and handing over to an agent only at the points where judgment is needed.

His inbound qualification flow shows how that looks. Every form submission goes to an agent that pulls its rules and context from a GitHub repository rather than from any single GTM platform. It enriches the lead through one API, scores it, and pushes qualified leads into HubSpot to trigger the follow-up. Two months in, lead quality is up roughly 50%, and he hasn't seen the agent do anything it wasn't built to do. The stack behind builds like this, five tools and one loop, is what Zeno shares in the GTM Field Kit, the free companion guide we built with the panel.

Grounded jobs are the safe jobs

Sumair trusts agents most with list building, list cleaning, signal sourcing and reply sorting. Each of those reads from real data, so there's very little room to hallucinate. The further an agent drifts from something concrete, the more you have to check its work.

Guardrails get written one failure at a time

None of these systems worked first time. Sumair's example is painfully ordinary: an enrichment API times out and returns "company not found", and a lazy quality check passes it because the company name field isn't empty. The fix is a rule that checks the name actually matches the domain. Small fixes like that stack up until the pieces click together like Lego.

The same logic applies to agents that talk to buyers. Sumair recommends grounding them in your own data, then role-playing hard conversations with them before they go anywhere near a prospect. He learned why the hard way, when he got a newly launched agent inside a popular GTM tool to curse him out. Ben's stress test in the Field Kit turns that kind of checking into a routine you can repeat before every launch.

Prospecting is going. The conversation isn't.

The "SDRs are dead" posts don't survive contact with people who run real sales teams. Zeno's view is that jobs won't disappear, but tasks inside them will. He doesn't expect any sales development rep to be manually prospecting in 2027, and a company that still lets them hasn't matured its systems. A well-built system can source, enrich and qualify a lead before a person touches it. What's left for the SDR is the call.

Operational roles change most. Sumair expects standalone tool operators and customer relationship management (CRM) admins to fold into a broader GTM engineering role. He pictures a GTM engineer running a pod, with SDRs inside it, some human and some agent.

AI can plan. It can't take the punch.

What agents can't do yet is sell. Sumair ran an AI voice agent pilot this year for a European client that didn't want to hire SDRs. The technology couldn't hear where an objection was coming from or the emotion behind it, so it couldn't answer it. He borrowed Mike Tyson's line to explain why: everybody has a plan until they get punched in the face. AI will hand you plenty of plans. The live conversation is the punch, and it still needs a person to absorb it.

Chris Muldoon, our CEO, sees buyers moving the same way. They're heading back to events and phone calls, looking for the context and emotional intelligence he doesn't think AI can offer.

The panel splits on how much reps should steer their agents

Ben wants trusted reps to have creative room to adapt their agents, inside clear guardrails, because tomorrow's sales teams will be smaller and need people who think for themselves. Sumair wants the agent's job tightly defined so reps spend their time on the phone, where they add value, rather than tinkering. Which one fits your team depends on how much you trust the people on your floor.

LinkedIn is working, and that makes it fragile

For Sumair, LinkedIn drives most results, in some campaigns up to 80% of them. Ben puts much of that down to the sender: a well-optimised profile, matched in seniority and discipline to the person on the other end, so an engineering manager hears from someone who speaks their language. The sender gets the message read. What earns the reply is the offer, which takes Ben back to his earlier question: what are we actually offering?

The risk is dependence. The most common job titles are being hit by every outbound team, and buyers are filtering them out. One policy or feature change could reset everyone's playbook overnight. Chris noticed LinkedIn has already started letting users flag posts and comments that seem like AI slop.

Email still earns its place too. Ben points to a robotics campaign getting strong results from email, while contact-centre buyers are hard to reach even on LinkedIn. Channel choice is a test you run for each market, and the answer won’t stay the same.

The next edge is a central brain

If everyone has the same tools and the same signals, the advantage moves to what only you know. Sumair and Ben both see companies building a central brain: one repository where sales, marketing, ops and HR context lives, so any team or model can draw on it without seeing every raw record.

Sumair says this was mostly theory six months ago. Ben sees smaller, tech-savvy companies already mapping it out, and the bigger platforms are catching up. You don't need anything exotic to start. Sumair suggests Obsidian, which is free to run locally, or Notion, which he thinks is underrated for this.

There's a practical upside too. Keep your logic and context somewhere any language model can connect to, and you can swap in a better model the week it launches without rebuilding a thing. Zeno's GitHub-based qualification flow already works this way.

Where to start

Sumair's answer is ideal customer profile (ICP) and persona selection. Take a title like "programme manager": it could sit in HR, IT or operations, and getting that right with manual rules means writing an almost endless list of exceptions. An agent working from real data handles it well, which makes it a low-risk place to learn. Once your ICP is set, his scoring rubric in the Field Kit takes the next step: who to contact first, and on which channel.

Ben's advice for anyone new is to understand why the agentic GTM movement exists, and what it's trying to achieve, before learning any platform. When you do pick one up, choose a tool that shows you every step of a build, so you can see how each piece connects.

What this means for your team

Three things change when every competitor has the same agents:

  1. Speed stops being an advantage. What you say, what you offer and how well you listen to replies decide whether outbound works.
  2. The SDR role narrows to the conversation. Prospecting moves to systems, and operational roles fold into GTM engineering.
  3. Your own context becomes the asset. Data your competitors can’t buy, stored where any model can reach it, is the part that compounds.

Buyers now choose who they hear from. Teams that keep using AI to send more will get faster at being ignored. Teams that use it to know more, test harder and free their people for real conversations will be the ones with a number to show the board.

Build your first agentic workflow with the GTM Field Kit

We built the free Going Agentic GTM Field Kit with the panel so you can put this into practice. It holds three tools for your first agentic GTM workflow, one from each panellist:

  • The stack, from Zeno Farhat: five tools and one loop you can set up in a weekend, from about $280 a month
  • The scoring rubric, from Sumair Shakir: decide who to contact first, and on which channel
  • The stress test, from Ben Robinson: check a workflow is ready before it reaches a buyer

Enter your work email below and the download starts straight away. If you'd rather hear the full debate, the conversation is on YouTube.

contributors
Expertise
No items found.
No items found.
No items found.
No items found.
other Articles

Continue reading

7
min read
Published
13 Nov 2026
How Punch! Combines Multi Channel Into One Motion
No items found.

How Punch! Combines Multi Channel Into One Motion

Last updated
13 Nov 2026
5
min read
Published
06 Nov 2026
What Is the Difference Between Martech and Sales Tech?
No items found.

What Is the Difference Between Martech and Sales Tech?

Marita van der Merwe

Marketing Manager
Last updated
06 Nov 2026

Subscribe to Punch! newsletter

A monthly dose of Sales Intelligence, delivered straight to your inbox

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.