New month, new face: The engineer teaching our internal AI systems to think

Product Updates Published: September 16, 2026
New month, new face: The engineer teaching our internal AI systems to think
Why it Matters
In reading this post, marketers will gain a clear picture of what happens when AI takes over the repetitive parts of internal tooling. This post explains how Oktopost's own internal GTM systems run on a stack of AI agents, and why the strategic judgment behind that stack still has to come from a person.
Key Points
  • Work that used to take a full day of manual effort now runs in minutes, across hundreds of items at a time.
  • The system pulls from a real stack, Salesforce, n8n, Claude Code, ZoomInfo, Explorium, Outreach.io, and HeyReach, each with its own job.
  • Nothing that comes out of the system looks templated, personalization is built into the automation, not bolted on after.
  • AI runs the system, but every judgment call about how it's built and where it should go next still belongs to Itamar.

The role

Itamar is Oktopost’s AI Systems Engineer, reporting to Operations leadership. The role owns the AI-driven operating model behind Oktopost’s internally built GTM tools, including the strategy, standards, and execution framework for building and maintaining AI-led systems with generative software engineering tools.

Day to day, that means designing and owning the applications that power internal GTM operations, integrating third-party tools, APIs, and internal systems so they connect cleanly. It also includes coordinating multi-agent workflows built with tools like Claude Code, with an eye on reliability, monitoring, and clean handoffs at deployment. Finally, the role covers maintaining the prompt libraries, safety guardrails, and security and data privacy standards that keep those systems trustworthy, and staying close to internal stakeholders so the tools keep improving based on real feedback.

The day to day

Itamar’s day to day isn’t one job, it’s several running at once.

Part of it is making sure the agents and the underlying system are actually working, no bottlenecks, no silent failures. Another part of it is watching the market, because new tools show up constantly and most of them aren’t worth the team’s time, but some are. And the final part of it, probably the hardest part, is finding new ways to solve problems that don’t have an obvious playbook yet. Everyone’s methods have gotten so condensed that standing out takes real creativity, not just another integration.

How the system is built

The system itself is built from a handful of tools stitched together with purpose. Salesforce holds the data. n8n handles the workflows. Claude Code writes and maintains the automation. ZoomInfo and Explorium surface signals. Outreach.io and HeyReach handle delivery. None of these tools do the whole job alone. The value is in how they hand off to each other.

What actually changed

Here’s what changed. Work that used to eat a full day of manual effort now takes minutes. The moment a signal fires, the system picks it up, filters it, drafts what’s needed, and creates the task. Nobody is copying information from tool to tool manually anymore. That’s not an exaggeration of scale, it’s the actual before and after.

Where judgment still lives

None of this replaces human judgment and decisioning. The system is complex enough that every part of it needs its own thinking, its own method, its own round of improvement.

In Itamar’s own words

“AI helps me build faster. It doesn’t decide what ‘faster’ should even mean for a given part of the process. That’s still on me.

I didn’t get here alone. Yaniv Setty, VP of Operations at Oktopost, brought the operations knowledge, the sense of how this should actually serve the business, not just look impressive in a demo. Julius Schwartz, AI Systems Engineer, brought the technical depth to make it real. Learning from both of them, in their own areas, made me better at this than I would have gotten on my own.

What I build touches more than one part of the business. Cleaner processes, better-handled data, and smarter automation don’t stay in one lane. Build it right in one place and the rest of the operation benefits too.

What’s changed most isn’t how fast I work. It’s what I spend my time on. Less doing, more deciding how things should work. More time asking what could be better, not just keeping up with what’s already running.

If I had to boil this down to three lines: AI isn’t here to change us. It’s here to amplify what we can already do. Nobody gets replaced, we just get handed better tools and the room to use them. Think Inspector Gadget. Smart tools don’t make the person, they make the person more of who they already were.”

Itamar, AI Systems Engineer

In summary

More New Month, New Face stories are on the way. Stay tuned!

Check out more from this series: Less than 2 weeks in: I gave an AI the keys to Oktopost, and 1 person, 3 marketing roles: how I run product, customer, and creative marketing on one platform.

For more insights, contact us to request a demo.

Frequently Asked Questions

What does an AI systems engineer do at Oktopost?

The role owns the AI-driven operating model behind Oktopost's internally built GTM tools, including the strategy, standards, and execution framework for building and maintaining AI-led systems. Day to day, that means designing internal applications, integrating third-party tools and APIs, coordinating multi-agent workflows, and maintaining the prompt libraries, safety guardrails, and data privacy standards that keep those systems trustworthy.

What tools power Oktopost's internal AI-driven GTM systems?

The stack includes Salesforce for data, n8n for workflows, Claude Code for writing and maintaining automation, ZoomInfo and Explorium for surfacing signals, and Outreach.io and HeyReach for delivery. Each tool has a specific job, and the value comes from how they hand off to each other rather than any single tool doing the whole job alone.

How has AI automation changed Oktopost's internal workflows?

Work that used to take a full day of manual effort now runs in minutes, across hundreds of items at a time. When a signal fires, the system picks it up, filters it, drafts what's needed, and creates the task without anyone copying information from tool to tool by hand.

Get ready!

The latest B2B marketing magic is about to land in your inbox
Join 30K+ pros already on the inside

Engaging social media content and interaction, illustrating B2B social media marketing tips and insights.