Kayla Siemon

Head of GTM Ops and Eng @officiallyiru

[Kayla Siemon @KaylaSiemon]

@brandon_galang Agreed. Until there is a date and something I can actually call, the Gemini 4 talk is just noise I have to filter out while deciding what we build on next.

[Kayla Siemon @KaylaSiemon]

Great post.

Point 2 is the one I'd want anyone who plans on using Jev for lead routing to consider. An output limited to your options can still be the wrong option, so the review step does not go away.

Hoping Jev opens new accounts again soon so I can start using it.

[brandon galang ▲ @brandon_galang]

the top 3 misconceptions I keep seeing about Jev:

1) Jev is NOT deterministic. it is typesafe and will only output pre-defined options you give it. within those options, it is still probabilistic.

2) Jev 'can't hallucinate,' but that doesn't mean it's never wrong. it will only ever output the options you give it, but it can choose the wrong option.

3) Jev is 'just a classifier.' this really undersells the value it provides. it's effectively a 'universal classifier' with frontier intelligence, so it can be quickly deployed across nearly an4y use case.

Jev has limitations but it is an incredible model in a space where there were previously very limited options.

Give Jev a try if you haven't. It's great.

[Kayla Siemon @KaylaSiemon]

@codyschneider Apify is one of the biggest unlocks in this stack.

Job listings are just one thing we run through it. With all of the Actors in the store, hiring intent is one of the thousands of signals you can start pulling.

[Kayla Siemon @KaylaSiemon]

The budget conversations I enjoy most at Iru start with builders saying what they need and what it saves the team.

Two of my GTM engineers did exactly that for an enterprise tier upgrade on a tool we use.

Clear ask, obvious time save, yes on the spot.

[Kayla Siemon @KaylaSiemon]

The maintenance burden line is the part that lands for me.

So much GTM ops time goes to patching routing rules and enrichment logic every time a field or an upstream tool changes, and I would happily hand that to an agent with the right tools.

[brandon galang ▲ @brandon_galang]

building in applied AI is a constant battle between fighting against and embracing the bitter lesson.

I estimate we may be only 1 generation away from models that are cheap and reliable enough where agents are the better defacto implementation over bespoke implementations.

I'm finding we are almost there with models like 5.6 luna and GLM 5.3 flash. I expect gpt 6 luna to solidly cross that threshold.

especially when you factor in the maintenance burden of deterministic logic, it will soon be cheaper, more performant, and more reliable to give your agents the tools to compose the workflow you need unless latency is a p0 concern.

[Kayla Siemon @KaylaSiemon]

Completely agree that you need a reliable content engine.

Too many good ideas stay buried in calls because turning them into posts is another task someone has to find time for.

This is the kind of work I want my skill files helping with.

[Fivos Aresti @fivosaresti]

This content engine has generated over 20,000,000+ impressions and got our team 180,000+ LinkedIn followers.

It turns call transcripts, raw ideas, and screenshots into LinkedIn drafts.

3 skills do the work, and they all sit inside Claude Code.

Here's the stack:

1 - Content Ideator

• Reads call transcripts, raw ideas, screenshots, or Notion pages
• Pulls the client's voice profile from our GitHub OS
• Queries our shared Pinecone DB for strong reference posts in the niche
• Returns hooks + body concepts in the Notion format the team already uses

2 - LinkedIn Drafter

• Takes the selected idea from the Ideator
• Applies the voice profile + our hook and body format libraries
• Stays Notion-first so the team can still edit before publish
• Speeds the draft, never replaces the writer

3 - Post QA

• Cross-references every claim against our claims database
• Locates supporting stats
• Grades the hook and body
• Flags items that need a human review before publish

The Ideator writes new posts back into Pinecone on every run.

And the engine gets sharper every week.

If you don’t have a reliable content engine in 2026, you’re behind.

[Kayla Siemon @KaylaSiemon]

Sad I missed this live! I really think CLI is the future of how we work with tools like Clay. The better the models get, the less patience I have for clicking through a UI to do something I could ask an agent to handle.

We were so excited when Clay’s CLI came out. Between that and the MCP, our whole team is enjoying being back in Clay.

[brendan j short @brendanjshort]

Last year, everyone building on the bleeding edge of GTM was using @clay.

Then, in Dec, most people started using Claude Code (I personally did as well).

For 6+ months, I watched many of the early Clay power users post about the workflows they built in Claude Code, that replaced Clay.

It was obvious to me that Clay would have to respond, with a product direction that supported these type of builders. It was NOT obvious how they could (eg: release an open API) given their business model.

BUT, this summer, Clay shipped. An API, a CLI, MCP, audiences, and more.

Enough that I'm not seeing several of the early Clay adopters who had moved over the code gen, start to use Clay again (often within their Claude Code/Codex/Cursor/Grok Bot setups).

I'm trying to explore exactly what is possible, what the people on the bleeding edge are doing, what's possible, and the nuance of it all.

So, that brings me to today:

I am bringing @THArrowOfApollo (one of the most vocal agency owners who left Clay this year) onto my show to discuss the pros and cons of what Clay has shipped. We'll have @yash_tek on the call too, to make sure any technical questions are answered.

It'll be a fun conversation. Come listen live: https://t.co/cyuSQB12vd

[Kayla Siemon @KaylaSiemon]

Love this guide.

It’s very similar to the approach we use for our own GTM context brain. Give agents a real understanding of how the company operates, then let them work from that context. Saving this one.

[Fivos Aresti @fivosaresti]

‘Service-as-a-software’ is here...

We moved our entire company brain to GitHub and wired 25+ tools through MCPs.

Any one of our 20+ team members can now spin up a contextualized AI assistant in seconds.

The system has 5 layers:

1. Markdown company OS
↳ SOPs and campaign playbooks converted into .md files using research agents
↳ Most SOPs turned into agents that handle 70% of the task
↳ Output: 50+ actionable Claude skills

2. Context environment
↳ One Company OS GitHub repo propagated to every session via org-wide plugin
↳ Each client gets their own repo with Slack DMs, call transcripts, GDrive changes, and campaign data auto-synced through n8n
↳ Zero configuration needed per session

3. MCPs
↳ 25+ tools connected including InstantlyAI, HeyReach, Apollo, HubSpot, Slack, Notion, n8n, Supabase, Pinecone, Browserbase, Apify
↳ Not just research. Action through AI.
↳ We went from researching work to actually doing it

4. Self-improvement engines
↳ Pinecone database stores 1000s of LinkedIn posts and outbound campaigns with performance metrics
↳ Copywriting skills query this data to find winning formats to reuse
↳ Human corrections get fed back in so the system gets sharper over time

5. Operating principles
↳ Every repo has a safeguard file that prevents certain operations
↳ 100% AI outputs are not acceptable, everyone owns their work and every mistake
↳ Agent swarms split one task into 5-20 sub-agents when needed

Our goal is to become the most advanced AI-native services company for our niche (GTM).

[Kayla Siemon @KaylaSiemon]

Been building with GPT-6 Astra and the thing I keep noticing is how it handles gaps.

It fills routine ones on its own and only asks when the answer could change the outcome.

For ops automations that is the difference between a workflow that runs and one I babysit.

[Kayla Siemon @KaylaSiemon]

@dan__rosenthal Agree, and the cost compounds once you put AI on top. Enrichment and scoring only cover accounts the system knows about. When reps source outside the CRM, agents end up routing and prioritizing against a partial picture and just move the gaps faster

[Kayla Siemon @KaylaSiemon]

Astra is being described as the best model for software engineering to date, some saying they're seeing the early form of AGI...

The question I care about for GTMe is whether it changes what one ops person can build and maintain alone.

Can't wait to find out!

[Kayla Siemon @KaylaSiemon]

Not long ago GTM engineering often meant stitching together a 14-tool tech stack and calling it a workflow.

Now the work is building governed data pipelines, because the AI on top fails without clean inputs. Fired up about that.

GTM engineers, what changed most in your role?

[Kayla Siemon @KaylaSiemon]

@lightsilver323 Unified.

[Kayla Siemon @KaylaSiemon]

Okay, this is insanely cool.

@treg_ai took the part of Clay I care about most, picking providers, building the waterfall, and controlling cost, and handed it to the agent. Open source and pay per result.

Looking forward to trying this out in my own work.

[Jason Zhou @jasonzhou1993]

Introducing Claude for people-search

No more $600 subscriptions, juts $0.0089/lead

Claude is now the best people finder in the world

#1 on People Search Bench
Fully open source, 0% markup

Try it at https://t.co/s60Ye4EO3h

Git Repo below 👇

[Iru @officiallyiru]

With an on-site gym, showers, private courtyard, fully-stocked pantry, 30,000 sq ft office space and plenty of open roles, Iru is proud to be a key player in Miami's tech scene. Check out our open roles in the linked reply.

[Iru @officiallyiru]

Client rosters move every month, so license counts should too. Iru license counts flex up and down, with reduced minimums and no restrictions on which operating systems you cover. An MSP should not still be paying for seats that walked in March.

[Iru @officiallyiru]

"Kandji Passport strikes a nice balance of integrations and the native macOS experience." @bradleychambers at @9to5mac reports on the latest development by Kandji to sync single sign-on credentials to the Mac for a secure login. https://t.co/pIfMXeyHh3

[Iru @officiallyiru]

With Kandji's new enrollment customization feature, users can be authenticated via Single Sign-on as soon as they start up their new Mac computers. Here's how it works. https://t.co/jkIgkfQoFO

[Felicis @felicis]

We are proud to back CEO Adam Pettit and the @KandjiMDM team at this exciting time of growth. We are also happy customers. @nikipez unveils who else is "eating from the apple tree." 🖥 📱
https://t.co/O0Y1WWXiM1

[9to5Mac @9to5mac]

Apple's enterprise growth remains strong as Kandji announces $60 million Series B funding round https://t.co/PhADoalacr by @bradleychambers