No Dependency Trap: How to Build GTM Systems That Compound, Not Just Noise with CheeTung Leong

GTM systems builder CheeTung Leong on agentic AI, automation’s hidden bottleneck, and why sales teams are betting on machines that actually work

Most conversations about AI and go-to-market are stuck in the strategy layer—full of promise, empty of production-ready systems. The real unlock isn't more theory; it's building the infrastructure that turns theory into repeatable, scalable execution.

That’s exactly where CheeTung Leong lives. As Founder and GTM Engineer at The GTM Architects, he doesn’t sell strategy decks. He builds working infrastructure inside his clients’ existing stacks—using tools like Claude Code, modern GTM platforms, and a 15-year B2B pedigree that includes two SaaS companies and eight figures in pipeline influenced.

CheeTung’s lane is the missing layer between strategy and execution: attribution that actually clarifies decisions, content systems that compound, outbound tied to real signals, and workflows teams can run without guesswork. No black box. No dependency trap. Just production-ready systems that real companies can run.

We spoke about the state of automation, where agentic AI is actually delivering, and which parts of the sales process still belong to humans.

Interview with CheeTung Leong

How fast is the automation industry growing and what is driving that growth?

Industrial automation is sitting at about $236B right now, on track to hit $450B by 2033. Warehouse automation is growing at 14% year-over-year. There are 4.6 million industrial robots deployed globally, and that number's growing 6% annually.

But the more interesting story isn't the hardware side. It's what's driving it.

Three things are converging: labor shortages that aren't going away, e-commerce economics that demand faster fulfillment, and AI that's now capable of automating knowledge work, not just physical work. That last one is the real unlock. The first two were already real ten years ago.

The economics have also shifted. Autonomous mobile robots used to be a 3-5 year ROI conversation. Now they pay back in 12 months. When the payback period compresses that much, the objection disappears.

I see it directly in the signal data I run for clients. When a $5M ARR company posts a "Revenue Ops Manager" role, that's usually a company that's been running GTM manually and finally hit the ceiling where it doesn't scale anymore. They're not hiring strategy, they're hiring someone to build the machine. That signal shows up in hiring data months before any market report picks it up.

What are the biggest challenges the automation industry is facing right now?

The one that surprises people: 54% of enterprises struggle to map their processes clearly enough to automate them. The AI isn't the hard part. Knowing what you're actually trying to automate is the hard part.

From there it breaks into three real categories:

1. Skills gap. 75% of IT automation projects hit skill deficits. The tools exist. The people who can build and maintain them — especially at the intersection of AI and business operations — don't exist in enough volume yet.
   

2. Integration complexity. Modern GTM stacks have 8-15 tools talking to each other. One broken connection and the whole pipeline silently fails. Most companies don't find out until someone checks the dashboard and the numbers look wrong.
   

3. Process opacity. This one's underrated. When you automate, you're forced to document every step that a human was doing informally. Tribal knowledge surfaces. The system breaks on edge cases nobody thought to mention. Automation doesn't create this problem, it reveals it.

The economic pressure is real too. Manufacturing PMI has been below 50% since mid-2022. Companies need to automate to defend margins, but they're doing it with thin teams and constrained budgets. That combination is where implementation fails.

How fast is agentic AI growing and why are sales teams paying attention to it?

The agentic AI market is at $7.6B in 2025. Projected to hit $199B by 2034. That's a 44% CAGR, not software growth rates, adoption-wave growth rates.

Sales teams are paying attention because the numbers are hard to argue with. Teams using AI are twice as likely to hit quota. Win rates up 30%. Lead volume up 50%, cost per lead down 60%. Those are the headline stats, and they're directionally real even if you cut them in half.

But the underlying reason is simpler. Sales is a volume and timing game. Right message, right person, right moment. AI doesn't change the game: it runs it at a scale no human team can match.

The "agentic" part is what's new. It's not AI helping you write a better email. It's AI deciding who to contact, when, with what message, processing the reply, updating the CRM without a human touching each step. That's a different order of magnitude.

We run a version of this for TGA's own sales. We watch for specific buying signals: a company re-posting the same sales role twice, or cancelling an AI SDR tool subscription. When a signal fires, the pipeline enriches the contact, generates personalized copy, and pushes it into our email platform. My involvement is the judgment layer: which signals matter, what the copy strategy should be, and how to optimize the architecture. The repetitive steps run automatically.

That's what's drawing sales teams in. Not the AI hype, the operating leverage.

What types of AI agents are sales teams using today?

A few categories that are actually in production, not just demos:

1. Prospecting agents. Scanning databases and job boards for ICP-matching leads, scoring them, prioritizing outreach. Apollo, ZoomInfo Copilot. 61% of sales teams have some version of this.

2. Outreach agents. Personalized email sequences at scale. Regie.ai, Lavender, a dozen others. ~52% adoption. The challenge is that most of them sound like AI, which kills reply rates. The teams winning here are the ones solving personalization at the signal level, not just at the template level.

3. Research agents. Deep account intel before outreach (tech stack, funding, headcount, news etc). Clay has become the popular tool most B2B teams are using.

4. SDR agents. End-to-end early funnel: prospect, outreach, qualify, hand off to an account exec. Salesloft Rhythm, Exceed.ai. 67% of enterprise sales orgs have deployed some version. This is the most contested category with lots of noise about what "full SDR replacement" actually means in practice.

5. Scheduling agents. Calendar coordination and booking. Reclaim.ai, Clara. Already table stakes. x.ai processed 50 million meetings by 2025.

6. CRM agents. Zero-touch data entry, auto-logging, forecasting. Salesforce Agentforce. 71% of CRM users are running something here.

The pattern that matters is that these are chaining together. Apollo prospects → research agent enriches → outreach agent writes and sends → scheduling agent books the meeting → CRM agent logs it. That full chain is what we build for clients rather than individual point tools.

Which parts of the sales process are AI agents expected to fully take over in the near future?

The honest answer: anything that scales poorly with headcount.

• Prospecting and enrichment — mostly done already. Gartner puts 85% autonomy at end of 2026. A well-trained agent finds better leads, faster, with cleaner data than an SDR manually pulling from Apollo. No contest.

• Initial outreach — not fully automated yet because the AI fingerprint is still visible. The teams that crack genuine personalization at scale — not mail-merge-level personalization, but signal-informed copy that reads like it was written by someone who actually looked at the account — will get there by 2027.

• Scheduling — essentially over. x.ai processed 50 million meetings. Nobody needs a human coordinating calendars.

• CRM hygiene — done. Zero-touch logging should be table stakes in 2026. Manual data entry is just a tax on rep time.

What AI won't fully take over: the conversation where both sides are figuring out if the problem is real and whether they trust each other enough to work together. Negotiation, complex deal architecture, high-stakes proposals. The closer you get to money moving, the more human it stays.

My rough frame for this: anything that happens before a real conversation starts is automatable. The conversation itself isn't — not because AI can't handle the words, but because buyers won't accept it for decisions that matter. (educated guess, could be off — but I haven't seen a client close a $50K deal without a human in the room yet.)

The upshot: AI agents replace the machine work in the sales motion. The judgment work, the trust work, the relationship work — that's where human time gets concentrated. That's actually a better deal for sales people, assuming they make the shift.