Compare · three categories
AI employee vs. chatbot vs. assistant
A chatbot answers the message in front of it. An assistant does what you ask while you stay in the thread. An AI employee holds a role: it keeps your facts between runs, works a queue on a schedule, obeys written limits on what it may do alone, and leaves money and outbound messages to a person.
Each of those parts is a record you can open. A role is a prompt with a validator, kept in code. Your facts are knowledge items, read into the prompt on every run. A timer wakes the work loop every 15 minutes. Each run lands in a journal with its inputs, its outcome and its error text.
At a glance
AI employee
Compare · six axes
Where they part ways
The chatbot and assistant answers describe how those two categories are usually built, with no vendor named. The AI employee answers describe gates that exist in this platform’s code, and each chip carries the literal string or number behind one.
| Axis | Chatbot | AI assistant | AI employee · Unistaff |
|---|---|---|---|
| Memory | The open conversation. Close the window and the context goes with it. | The current thread, plus whatever you paste back in or save by hand. | Knowledge items stored per brand and read into the prompt on every run, so two runs a month apart answer from the same facts. The last successful run's output becomes the next role's input.knowledge_items |
| Initiative | Waits for a visitor to type. | Waits for you to prompt it. | A timer wakes the work loop every 15 minutes. Each role carries its own cadence: 7 days for the market analyst, 24 hours for the morning brief, on arrival for a new lead.15 min tick |
| Queue | One message at a time, in the order people arrive. | One thread at a time, in the order you open them. | Work is rows in a job table with a status — queued, running, ok, failed. One job is drained per tick, a role that already has a pending job gets no duplicate, and a run past 8 minutes is closed as failed.queued · running · ok · failed |
| Limits of authority | Whatever the widget was wired to reach. | Whatever tools you granted for the session. | Approval rules, spend caps and the kill switch live in the database and are checked before every outbound write. An action the rulebook has never seen counts as needing approval.approval_required |
| Reporting | Transcripts, read after the fact. | The thread you were sitting in. | Runs, sync attempts, proposals and decisions are written with timestamps and outcomes, failures included. Each decision keeps who decided, when, and the note they left.journaled, failures included |
| Human control | You review the log later. | You are the one typing. | Drafts. The lead responder has no send path of its own, money and outbound messages stop at the approval gate, and a pending approval that runs out of time stays undecided.drafts only |
The gates named above — approvals, spend caps, the kill switch, the journal — are listed one by one on the product page. The cost side of the same choice is on AI employee vs. human hire.
Employee experience
The same parts, pointed at your own team
Internal questions have the shape of customer ones: a repeat question whose answer already exists in a document nobody rereads. The support role works the same way inside the company as it does on the front desk.
01 · Ask
An employee question is answered from the same material as a customer one
Your handbook, policy pages and past answers are stored as knowledge items — facts, policies, legal restrictions, audiences, cases — and read into the prompt on every run. The draft comes back with the passage it used attached, so whoever sends it can check the answer against the source in seconds.
02 · Onboard
Week one and month six get the same answer to the same question
A new hire asks the questions the last one asked: expense limits, time off, who signs what. Because the material is stored per brand and reread on each run, the answer stops drifting with whoever happens to be free. Marking a fact verified is a separate, deliberate action, so nothing gets a fresh timestamp just because a run touched it.
03 · Report
The daily brief is built from the system's own state
One role writes a brief every 24 hours out of the actual pulse of the platform, and its validator compares every number in the text against the inputs the role was handed. A brief carrying a figure that never came from an input fails as a run instead of landing in someone's inbox.
04 · Limit
Internal chat is scoped during a deployment
One platform connector ships today — Meta ads — plus an endpoint any form or quiz can post to, which is how an internal help form reaches a role right now. Telegram sits in the planned group on the integrations page. Slack, Teams and helpdesk tools are built against your account as part of a deployment.
The customer-facing version of this role, with its workflow and its first-month numbers, is on the AI customer support agent page.
Which one is the right call
A chatbot is enough when
- A single page needs a narrow FAQ answered on the spot.
- Volume is small enough that the follow-up work fits in someone's afternoon.
- The conversation starts and ends in the chat, with nothing to update afterwards.
- You want a two-week experiment before committing to a role.
An assistant is enough when
- A person is already at the keyboard and wants the draft faster.
- The work is one-off: a proposal, a summary, a rewrite.
- Judgement lives with the human on every single step.
- Nobody needs a record of what ran while the office was closed.
A role earns its keep when
- The job continues after the answer: a CRM update, a booking, an invoice chased.
- The same questions repeat with context specific to your business.
- Somebody has to look at the work overnight and have it ready by morning.
- Money, outbound messages and customer-sensitive actions need a human yes on record.
Pricing runs per digital employee per month — $149 Starter, $599 Scale, custom Enterprise — so the arithmetic in the ROI calculator stays flat while volume moves.
Frequently asked questions
A chatbot sits on a channel and answers whoever writes to it, using the material it was given. An assistant sits in a thread you drive and does what you ask of it there. Both hold context for the length of the conversation. Neither one carries a role, a work queue, or a rule about what it may do without you.
A role, kept in code as a prompt plus a validator that can reject the role's own output. Around it: your facts as knowledge items read on every run, a timer that wakes the loop every 15 minutes, a job queue with statuses, approval rules and spend caps checked before every outbound write, and a journal that records failures as well as successes.
Two ways. Your material is stored as knowledge items per brand — facts, policies, legal restrictions, audiences, cases — and read into the prompt on every run. And the output of one successful run becomes the input of the next role in the chain, which is how a decision made on Monday still shapes what gets drafted on Friday.
No. The lead responder writes the first reply and has no send path of its own. Launching a campaign, changing a budget or pausing something sits in the approval class, an action the rulebook has never seen counts as needing approval, and a pending approval that expires stays undecided. If the rulebook cannot be read, the outbound write is refused.
Only once you add the parts around it: a role with a validator, stored material it rereads, a schedule, a job queue, approval rules with spend caps, and a journal. That build is the work Unistaff ships, which is why a deployment starts by picking a role and briefing it in plain language.
Yes, with the same parts pointed inward. Your handbook and policies become knowledge items, an internal help form posts to the same endpoint a customer form uses, and the answer comes back as a draft with its source attached. Internal chat tools like Slack or Teams are scoped and built during a deployment.
Meta ads for reporting, with writes behind a separate token and the approval gates. An endpoint that any web form or quiz can post to, with an email to you when a lead arrives through your own site form. A daily capture of the public Meta Ad Library. Telegram is planned; everything else is a connector we scope with you before we promise it.
Starter is $149 per month, Scale is $599 per month, Enterprise is custom. The rate is per digital employee per month, with seats, messages and tokens left unmetered, so the bill holds still when a busy week doubles the volume.
Bring one role in and watch the journal for a week
Every run is timestamped with its inputs and its outcome, so the decision to keep it is made on a record rather than an impression.