workfOS: A Temp Agency for AI Workers

The idea in one sentence

workfOS is like a temp agency for AI workers. You hire dedicated specialists to handle your workflows, so you don’t have to juggle generic chat tools on your own.

If you’ve ever hired a contractor through an agency, you already understand the model: you describe the job, the agency matches talent, handles onboarding, and the worker shows up ready to work. workfOS does the same thing with AI.

Why this isn’t “just ChatGPT with a login screen”

Most AI products send every question to a single large model and hope for the best. That works for casual questions. It breaks down when you need a worker who:

• Watches your inbox every hour
• Answers questions using your documents — not the internet
• Screens résumés against your open roles
• Eventually, draft an email in your voice.

Each of these jobs needs its own approach. workfOS is designed with that in mind: it matches the right type of intelligence to each task and only uses costly AI when it truly adds value.

We call this a routing engine, not a wrapper. The platform decides how each job gets done, not just which chatbot responds.

A platform that grows with you, not just a fixed menu of four bots

workfOS is designed as a platform for hiring AI workers, not as a product with strict limits. The workers available now make up the launch roster, which includes general-purpose specialists that many businesses need from the start. Over time, the roster will expand in two ways:

• New worker types for common jobs (inbox triage, document Q&A, hiring, communications, and beyond)
• Industry vertical bundles: pre-configured teams of workers tailored to how a specific business operates

You hire the workers who fit your needs. The platform manages accounts, billing, scheduling, and data isolation, whether you use one worker or a small team.

The launch roster: specialists available today

These are the workers who currently live on the platform. Each is a specialist with a clear job description. You hire the ones you need; they share the same platform for accounts, billing, and settings.

Research Assistant

Job: Answer questions about your documents — research papers, reports, URLs, notes.

Why it matters: Knowledge workers often get overwhelmed by reading. This worker understands your documents and cites their sources, so you can trust the source of each answer.

How it works (in simple terms): You upload your documents. When you ask a question, the worker first finds the relevant passages and answers based on them, not on guesswork.

Job Scout

Job: Monitor your inbox and surface job opportunities worth your time.

Why it matters: Job searching often means sorting through LinkedIn alerts, Indeed digests, recruiter spam, and the occasional real lead. Most of it is just noise.

How it works (in simple terms): Connect your email. Job Scout checks your inbox on a schedule, sorts messages into categories (strong match, worth a look, or skip), notifies you right away for great fits, and sends a summary twice a day for the rest.

Recruiter

Job: Help hiring teams find strong candidates — including people whose résumés don’t keyword-game the system.

Why it matters: The best candidate is not always the one who optimized their résumé for your ATS. This worker reviews incoming applications, matches them to your open roles, and highlights hidden gems in a review queue and a twice-daily summary.

How it works (in simple terms): Connect your hiring mailbox, set your open roles, and let the worker score incoming résumés based on those roles.

Communications Manager (early access)

Job: Eventually draft emails and messages in your personal writing style — with your approval before anything is sent.

Why it matters: Generic AI writing often sounds bland. A worker trained on your writing samples can handle routine messages without making you sound robotic.

What’s available now: The system collects your writing samples and checks if you have enough quality material to train a personalized model. Full drafting will be available in a future release.

What verticals look like

These are not products you can buy off the shelf today. Instead, they show how the platform can assemble workers for a specific industry. The infrastructure stays the same, but the worker mix changes.

Property Management

The business: A property manager overseeing 40 units. About 200 emails a week. One inbox.

The problem: Maintenance requests, lease questions, and vendor follow-ups all arrive in the same inbox. Every morning starts with sorting through these messages to figure out what’s urgent, what can wait, and what needs to be checked in the lease or policy documents.

The worker team:

• An inbox worker monitors the main mailbox and highlights what really needs attention, such as maintenance emergencies, routine questions, or just noise.
• A Research Assistant, equipped with your policy documents, lease templates, and building rules, answers tenant and staff questions using your materials and provides citations, instead of giving generic internet advice.

The outcome: Your mornings no longer begin with sorting messages. Urgent maintenance is flagged, lease questions are answered from your documents, and you only need to review a short list of items that require a human touch, not 200 unread emails.

Customer Email & Support (e.g., banking, insurance, regulated services)

The business: A support team handling hundreds of inbound customer emails — account questions, complaints, product inquiries — across a shared mailbox. Responses must follow company policy, align with the customer’s situation, and recognize when a human needs to step in.

The problem: Every message appears equally urgent in the unread folder. Agents end up retyping the same answers, and the tone can be inconsistent—sometimes too cold for complaints or too casual for policy questions. Complex cases can get overlooked until it’s too late.

What the worker team would handle:

The outcome: Routine inquiries receive a solid first draft in seconds. Complaints get an empathetic, policy-compliant response or are flagged for escalation when they need a senior agent. Your team can focus on judgment calls instead of retyping the same paragraphs.

That’s the idea behind the platform: assemble the right workers for your industry, set them up with your data and integrations, and let them do their jobs. Whether it’s property management, customer support, legal intake, or clinic scheduling, the pattern remains the same, but the mix of workers changes.

How you actually use it

The experience is deliberately simple:

1. Hire: Choose a worker from the roster
2. Configure: Connect your email, upload documents, and add your API keys if needed
3. Work: Your worker runs on a schedule, around the clock

You manage everything from a web dashboard: start and stop workers, see recent activity, and review what each worker found or answered.

What you bring vs. what we handle

Trust, privacy, and cost: the things buyers actually care about

Your data stays yours.

Each customer gets isolated storage. Your documents, résumés, and worker settings stay in your own space and are not mixed with anyone else’s.

Your keys, your bill

For production use, you connect your own Anthropic or OpenAI account keys. workfOS routes workers through your keys, so you are not dependent on the platform subsidizing unlimited AI usage, and you keep direct control over your spending with your provider. This policy will change as the platform matures and may provide workers with a salary.

Cost-conscious by design

Workers do not use expensive AI for every step. Job Scout uses quick rules to sort mail first, and the language model only ranks the shortlist. The recruiter tries simple matching before using AI for unclear cases. Research Assistant removes irrelevant document sections before answering.

Every AI call is logged, so you can see exactly what your workers cost to run. This is important when you pay a monthly “salary” for each one.

Human approval for high-stakes actions

For any action that affects the outside world, especially sending email on your behalf, the system assumes you will approve it before anything is sent. Autonomous outbound communication is not the default.

The “salary” model

You pay a recurring fee per worker, similar to a salary, based on complexity and usage. This matches familiar SaaS pricing while keeping the staffing metaphor accurate:

• A document Q&A worker has a different “seniority level” compared to a fine-tuned communications worker.
• Usage has real costs; transparency matters.

Billing and plan management live in the same dashboard as your workers.

Where is this going

Today’s launch roster includes Research Assistant, Job Scout, and Recruiter, with Communications Manager in early access. The platform roadmap extends in two directions:

• More worker types — inbox triage, drafting, domain-specific workflows, and integrations we haven’t built yet
• Industry vertical bundles — pre-configured teams like the property management and customer-support examples above, tuned to how that business actually runs
• Custom fine-tuned models trained on your writing and domain data
• Ensemble answers — asking multiple models and synthesizing the best response
• Deeper integrations — Gmail draft creation, approval queues, and more

The main idea stays the same: a growing agency of specialized workers, smart routing, and management for people who do not want to handle infrastructure themselves.

The bottom line

workfOS is not trying to be the smartest chatbot. Instead, it aims to be the best employer of AI workers — a platform where you hire the right specialists for your work, match intelligence to each job, and grow your team as new worker types and industry bundles become available.

If your team is overwhelmed by inbox noise, buried in documents, spending hours on initial résumé screening, or drafting the same customer replies again and again, that is the problem this platform was built to solve. It starts with the launch roster and expands into the verticals where these patterns repeat.

workfOS: A temp agency for AI workers.