Can the Tab AI assistant run your job search?

What did Tab just launch, and can it run your job search?

Tab is a personal AI assistant you text on iMessage or WhatsApp, and as of 7 October 2026 it's officially out of stealth at a $300 million pre-seed valuation — but the product is still gated behind a waitlist, so for most job seekers it's a plan rather than a tool. Brennan Erbz, Stafford Schlitt and Ammar Amdani started building it earlier this year, with backing from SV Angel, Valar Ventures and American Spirit. The pitch is deliberately unglamorous: an agent should live where you already spend your day, so there's no app to install and no dashboard to learn. You text it like a contact, and it goes and does the thing.

Tab's capability list reads like a personal errand runner. It books flights, places phone calls to businesses, drafts emails, chases wire transfers and processes payments. The distinction the founders keep drawing is execution over suggestion — it does the task instead of explaining how you might. Under the hood it runs its own computing environment and carries context between conversations, so it remembers what you told it last Tuesday. It accepts text, photos, PDFs and video as input. On privacy, the company says Tab operates with its own phone number and wallet so your personal data isn't used to train AI, that financial information stays confidential, and that significant actions need your explicit approval before they execute.

None of that was built for job hunting, and Tab hasn't claimed otherwise. But strip the thinking out of a job search and look at what remains: forty applications in a spreadsheet nobody updates, the follow-up you meant to send on day eight, a recruiter email that needs an answer before lunch, a phone call to reception to work out who the hiring manager actually reports to. That's errand work, and errand work is precisely where this generation of assistants earns its keep. The catch is simple, and it governs everything below — an agent acting on your behalf sends whatever document you handed it, flaws included, at machine speed.

What can a personal AI assistant actually do for your job hunt?

A personal AI assistant can run the administrative layer of a job search end to end: tracking what you applied for and when, drafting follow-ups and sending them on a schedule, answering recruiter emails, phoning a company to confirm an interview slot, and holding your work history in memory so you stop retyping the same three bullets. What it can't do is judge. It doesn't know that the role paying slightly more is a sideways move, that your "stakeholder management" bullet says nothing concrete, or that the hiring manager you're about to cold-email left in August. Delegate the logistics. Keep the strategy, the targeting and the final read of anything with your name on it.

In practice the useful instructions are narrow and repetitive. "Here's the job ad text — add it to my tracker with today's date and remind me to follow up in eight working days." "Draft a four-sentence note to this recruiter referencing the two things in the ad I actually have." "Pull the three roles I applied for last week and tell me which ones have gone quiet." That's unremarkable work, and unremarkable work is what eats a job seeker's evening. An assistant that reliably clears it is worth more than one that promises to write your whole application, because the writing is the part a human still needs to own.

Here's the contrarian bit: the biggest risk of handing your search to an agent isn't that it does too little, it's that it makes you sound synthetic at volume. Recruiters now read dozens of letters a week with the same cadence, the same triads of adjectives, the same bloodless opening line. One agent-drafted follow-up is fine; thirty of them leave a trail. Before you let any assistant send outreach in your name, it's worth checking whether your own CV already reads as AI-written — the tells in a CV and in an email are the same tells, and they're easier to fix than to live with.

Why fix your CV before any AI agent touches it?

Because agents send files — they don't repair them. A two-column PDF, a template with icons instead of section headings, contact details stranded in a header, text trapped in a sidebar or a scanned image: applicant tracking systems routinely lose that content during extraction, and an agent that applies to sixty roles instead of six simply multiplies the same broken document. The screening layer you're trying to clear doesn't see your layout. It sees whatever text came out of the file, in whatever order, mapped into fields. If your employment dates land in the wrong column or your skills section never extracts at all, speed works against you.

This is the part worth doing by hand before you automate anything. A free CV analysis from HRLens takes a PDF, Word file or even a photo of your CV — Hebrew OCR included — and returns an overall score out of 100 rated Weak, Decent, Strong or Excellent, plus five category scores covering experience, tech stack, impact and ownership, clarity and structure, and ATS compatibility. You also get your genuine strengths and a visual layout analysis showing how the page reads. That's enough to tell you whether the document an agent is about to fire off is actually parseable, or whether you've been applying into a void for weeks.

The second thing to fix is substance, and no assistant will do it for you. "Responsible for improving onboarding" is a sentence an agent can send a thousand times without ever becoming persuasive. "Cut onboarding from 14 days to 5 and raised 30-day activation from 41% to 58%" is the version a human remembers. If you're rebuilding rather than patching, the chat-based HRLens CV builder lets you paste an old CV or just describe what you did, refine it by conversation, and export to PDF or Word across six templates — Hebrew CVs are written in Hebrew and exported right to left.

How does Tab compare to OpenAI dots and Meta Muse?

For a job seeker today, Meta Muse is the one you can actually use, OpenAI dots is the most capable on paper, and Tab is the easiest to adopt once you get in. Muse launched on 8 September 2026 with free, $20 and $100 monthly tiers, runs in its own app, on the web and through WhatsApp, and is US-only for adults 18 and over — though Meta requires a payment card on file even for the free tier, which is metered rather than unlimited. Dots arrived at OpenAI's DevDay on 29 September 2026, powered by GPT-6 Astra, and shipped to ChatGPT Pro and Business Premium subscribers.

Dots is the one built for delegation at scale. OpenAI says the agents connect to more than 4,000 apps, including Slack and Teams, run on their own cloud computer, work toward goals around the clock and learn from feedback over time. There's a read-only mode that blocks browser and computer access while you're away. Tab takes the opposite route — no new surface at all, just a thread in the messaging app already on your phone, with its own phone number and wallet rather than access to yours. That's a genuinely lower-friction trade, and the reason the valuation raised eyebrows for a product still behind a waitlist.

My honest read: for a job search, the gap between these three matters less than the marketing suggests. All of them can hold a tracker, draft a follow-up and remember your history. None of them knows whether your CV survives extraction, and none will tell you that your bullets have no numbers in them. If you want a deeper breakdown of the two most mature options, this comparison of which agent should run your job search goes through price, availability and privacy in detail. Pick on availability and cost, not on feature lists you won't use.

How do you set up an AI assistant for your job search?

Set it up in this order, and the whole thing takes an afternoon. First, score your current CV and fix what the analysis flags — layout, missing numbers, an unparseable template. Second, decide on one target role and write a two-line description of it, because an agent with vague instructions produces vague output. Third, give the assistant a clean brief: your current CV text, the three roles you're targeting, your notice period, salary floor and locations. Fourth, define the recurring jobs — log applications, nudge at eight working days, summarise the week on Friday. Only then let it send anything on your behalf.

Two practical notes that save real time. Paste job description text rather than links: LinkedIn and Indeed block automated fetching, so a URL often returns nothing useful while the pasted text works immediately. And when you paste a job description into an HRLens analysis, you get back the skills your CV is missing against that specific ad, plus ATS-friendly rewritten versions in six templates that mirror the ad's own keywords — which is the material your agent should be working from. If you want the prompt-level detail, there's a walkthrough of setting up Muse for a job search with twelve starting prompts you can adapt to any of these assistants.

Finally, be deliberate about what you upload. These agents run persistent cloud environments and remember context across sessions, which is exactly what makes them useful and exactly what makes a full CV a meaningful disclosure — your address, your ID number, your referees' phone numbers. Strip what the agent doesn't need before you hand the file over; a redacted CV still gets you useful feedback and useful drafting. The questions worth asking first are covered in this piece on what uploading your CV to an AI tool actually exposes. Start every agent in its most restricted mode, keep approval required for anything it sends, and widen permissions only once you've watched it work.

Frequently asked questions