I gave ChatGPT Dots and Grok Bot the same job: build an agent that finds remote roles that fit my resume, tailors the application, finds a contact email, and drafts the outreach. In my test, Dots worked for about 45 minutes, came back with three jobs and couldn’t find the contact emails. Grok Bot set up an orchestrator agent with an Airtable database in a couple of minutes and started returning matches.

That’s one test, by one person, on one day. These tools change every few weeks, and your results will depend on your prompt and your plan. Below I explain what each tool is, how I set up the test, how to build a similar job application agent, and the mistakes that will get you into trouble.

Prefer to watch? See the full walkthrough on YouTube. I build the whole thing on screen.

What ChatGPT Dots and Grok Bot are

ChatGPT Dots

Dots is OpenAI’s always-on agent inside ChatGPT, powered by GPT-6 Astra. It works on tasks in the background and reports back while you do other things. As of late October 2026 it’s available on Pro plans (from $100 a month, with $200 and $500 tiers), Business Premium and in beta for Enterprise. It isn’t offered on Pro in the EEA, Switzerland or the UK. Your first dot is included. Chatting with a dot doesn’t count against your usage, but the work it does in ChatGPT Work or Codex does.

Grok Bot

Grok Bot is an agent that builds and runs its own sub-agents, tools and workflows. In my test it used Treg, a tool marketplace, to call data providers such as LeadMagic for job and contact lookups.

One correction to the video: I mention Grok Bot is free to start, with $20 Pro and $60 Pro Plus plans. That pricing changed on October 12, 2026. New customers now get Grok Bot through The X Subscription Premium ($30 a month), Super ($100) or Ultra ($200). The Free and Plus ($8) tiers don’t include it. Existing subscribers on the older plans were kept on them, so check the current pricing page before you sign up.

The test: one job application agent, two tools

I wanted a realistic task with several moving parts, not a single prompt. The agent had to:

• Read my resume and understand what roles fit.
• Search for open remote roles that match.
• Store results somewhere I can review and filter.
• Find a contact or application email for each role.
• Draft a tailored application and outreach email for my approval.

The last step is the one that matters most. An agent that sends job applications without your review can embarrass you quickly.

How the Dots run went

Dots took the task and worked through it in the background. After about 45 minutes it came back with three jobs. It couldn’t find contact emails for them. For a single research pass that’s useful. For a repeatable pipeline that runs every day, it’s slow, and you still have to do the contact research yourself.

How to build the agent in Grok Bot, step by step

Here’s the flow I followed. The same structure works in most agent tools that can create sub-agents and connect to a database.

1. Write a detailed brief. Spell out the goal, the inputs, the output fields and the approval rule. A vague prompt produces a vague agent. There’s an example below.
2. Let it set up an orchestrator. Grok Bot created a main agent that coordinates the work and set up an Airtable base to hold results.
3. Give it your resume. Upload a master resume so it can build a profile of your target roles and skills.
4. Answer its clarifying questions. Expect questions about job titles, remote or on-site, regions and salary expectations. Specific answers mean fewer bad matches.
5. Run a small test search. Start with a handful of results. Check that each job link opens and matches what you asked for.
6. Review the Airtable records. Each match should have a job title, company, job link, company link, location, remote status, employment type and salary where listed.
7. Approve before anything goes out. Review each draft and send it yourself, or approve it one at a time.

I show the Airtable fields and the clarifying questions on screen in the video, so you can copy the structure.

If you get stuck connecting the database or the data tools, AI Income Lab has templates for automation builds like this one and a place to ask questions.

Example brief you can adapt

This is a starting point, not the exact prompt from the video. Edit the filters to fit your situation.

You are my job search assistant.
Goal: find remote roles that fit my background and prepare
applications for my approval.
1. Read my master resume (attached) and build a short profile:
   target titles, core skills, seniority, locations I accept.
2. Search for open roles posted in the last 14 days that match
   the profile. Skip roles that need relocation or sponsorship
   I don't have.
3. Save each match in Airtable with: job title, company,
   job URL, company URL, location, remote status, employment
   type, salary (if listed), date found.
4. For each match, find a published application email or
   hiring contact. Mark it "unverified" if you can't confirm it.
5. Draft a tailored resume summary and a short outreach email.
6. Do NOT send anything. Wait for my approval on each one.

Notice the last line. Keep a human approval step in the brief from day one, even when the agent seems reliable.

Keeping a job application agent safe

• Verify every email address. Contact lookups can return guessed or outdated addresses. Use only addresses published by the company or confirmed by a verification tool.
• Respect site rules. Many job boards limit automated scraping. Prefer official APIs, feeds or data providers that license the data.
• Protect your resume data. Your resume includes personal details. Check where the tool stores files and who can see the database.
• Watch the costs. Data lookups are usually billed per call. Set a budget and check the spend after the first run.
• Keep volume low. Ten well-matched, tailored applications beat a hundred generic ones.

Want a second set of eyes on your setup? Join the AI Income Lab community and share your build before you scale it.

Common mistakes to avoid

• Letting the agent send applications on its own. One wrong email to a hiring manager costs more than the time you saved.
• Skipping the test run. Run five results first and check them by hand before you run fifty.
• Using a one-line prompt. Without filters for location, seniority and sponsorship, you’ll get lots of irrelevant matches.
• Trusting salary and remote labels blindly. Listings are often vague. Confirm on the employer’s own careers page.
• Judging a tool from one run. Both tools update often. Re-test with your own task before you commit to a paid plan.

Which one should you use?

If you already pay for ChatGPT Pro and need occasional research, try Dots first. If you want a repeatable pipeline that writes to a database and uses data providers, Grok Bot handled that better in my test.

FAQ

Can an AI agent apply to jobs for me?

It can find roles, draft tailored applications and prepare outreach. I’d still review and send each one yourself, or approve them one by one.

Is ChatGPT Dots available on the Plus plan?

No. As of late October 2026, Dots is on Pro (from $100 a month), Business Premium and Enterprise beta. It isn’t available on Pro in the EEA, Switzerland or the UK.

How much does Grok Bot cost?

For new customers since October 12, 2026, it comes with The X Subscription Premium ($30 a month), Super ($100) or Ultra ($200). Data lookups through tools like Treg may cost extra.

Why couldn’t Dots find contact emails?

I can’t say for sure why. Grok Bot had access to dedicated contact data providers like LeadMagic through Treg, which likely made the difference.

Do I need to code to build this?

No. Both tools work from a written brief. It helps to understand basic database fields and to test each step before you rely on it.

Conclusion

For this task, Grok Bot got a working job application pipeline running much faster than Dots in my test. Dots is still useful for background research if you’re already on ChatGPT Pro. Whichever you choose, write a detailed brief, start with a small test, verify every contact, and keep yourself in the approval loop.

Your next step: write your own brief using the example above, run it for five results, and check each one by hand before you scale.

If you’d rather not figure this out alone, join AI Income Lab for the courses, templates and help that go with these builds.