Two bots want to run your GTM stack

Hey friends, a few weeks ago the internet decided Grok Bot was going to run everyone's outbound. Then OpenAI shipped dots this week, and the same crowd decided Grok Bot was already old news. Give it a month, though, and who's to say what makes dots the old news next?

In today’s Circuit:

🛠️ Operator Tip of the Week
🤖 The always-on agent race, and what it means for GTM
💡 Tool Spotlight
📡 Across The Wire
🧪 Worth Trying This Week

Before you move any real workflow onto a new agent, run it in parallel for a week against what you already have. Pick one repeatable job, outbound research or email copy, and let both do it side by side.

The demos always look cool, but what you care about is how much you end up reviewing and fixing. Because that time is the real cost, and it rarely shows up until a few days have passed (and you wasted a bunch of time, again).

OpenAI released dots this week, an always-on agent running on its own cloud computer and browser, powered by GPT-6 Astra. The pitch is one primary agent that coordinates background work and hands tasks off to Codex.

It plugs into Slack, carries a big library of app integrations, and adds approval controls that let you decide what it does on its own and what needs a human.

Dots needs a $100-a-month ChatGPT Pro plan. xAI got there first back in August with Grok Bot, which leans on named roles you teach by recording a workflow once and replaying it as a repeatable skill, and starts around $20 through Cursor.

Both chase the same job: an agent that runs the boring, repeatable parts of your day.

The two teams split on how an agent should be shaped. Grok Bot makes the roles visible, so you build a set of named teammates and watch each one work.

Dots hides that behind a single agent that farms tasks out in the background. For GTM that maps onto real choices. Visible roles are easier to trust when a bot is touching your CRM or sending on your behalf, and early reviewers found Grok Bot felt like a finished product while dots was slow and stumbly on launch day.

Dots counters with tighter governance, the approval rules and the Codex handoff, which starts to matter once an agent is allowed near anything that ships to a prospect.

The catch for most GTM teams is that your setup probably already lives somewhere. If you've wired research, enrichment and drafting into Codex or Claude Code, you can usually point one of these agents at that setup instead of rebuilding from scratch.

So the interesting question is how many people actually move their stack over versus just bolt a new agent onto what they already run. Our guess is most won't move. They'll point the new agent at one job, see if it holds up, and keep the rest where it is (otherwise you'd be rebuilding your whole setup every other day of the week and go crazy).

Then there's timing. The moment a standalone harness proves useful, the model makers have every reason to fold the same thing into the tools you already pay for.

Grok Bot rides inside Cursor. Dots rides inside ChatGPT. It's easy to picture the same coordinate-your-agents feature landing natively in whatever you use to code, at which point this month's must-have becomes obsolete again by the time you set it up.

So the safe move is staying portable. Keep your knowledge and your workflows somewhere you can move them, and treat each new harness as a layer you rent rather than a foundation.

OpenRouter is a single API that routes your requests across just about every model, OpenAI, Anthropic, xAI, Google and more, through one interface, so you can swap the model underneath without rewriting anything.

If the worry this week is getting locked to one provider, this is the portability layer for the model itself. Build your own agent on top of OpenRouter and GPT-6, Grok or whatever wins next month is a one-line change instead of a migration.

- OpenAI's dots launch, an always-on agent on GPT-6 Astra that coordinates background work and hands off to Codex. Read here.

- Hands-on head-to-head: one tester clocked dots taking about five minutes to book a restaurant on launch day while Grok Bot felt polished. Read here.

- A fuller feature-by-feature breakdown of the two, pricing, platforms and governance. Read here.

- Another side-by-side if you want a second read before you commit a workflow. Read here.

- Pick one repeatable GTM task and run a new agent against your current setup in parallel for a week.

- Write down what you had to review or redo after each run. That's your real cost. Time is money my friends.

- Keep your prompts, ICP and workflows somewhere portable so switching harnesses is a copy-paste.

- Before migrating anything that sends to a prospect, check how the tool isolates credentials.