Root AI
It reads the same live database your sales, inventory, schedule, and books run on, so it can answer with the real number and then do the next step: reorder the stock, draft the schedule, send the win-back.
The difference
A chat window bolted onto four tools that sync overnight can only repeat last night's export. Root AI queries the same live database, so it reads today's sales, today's stock count, and today's open invoices, then acts on them: adjusts a reorder, drafts a schedule, queues a payment retry.
Ask the business anything
Ask in plain language and get an answer drawn from billing, staff, scheduling, and revenue together, in one context, instead of opening one siloed report at a time.
Because it reads the live operational database, the answer is always now, not whenever the last export ran.
Then it acts
An answer is the start. Within the limits you set, Root does the work, retries the failed card, drafts the leaner roster, queues the reminder, and brings you the result or the one approval left to tap.
Covers were down 14% on Tuesday, driven by a 22% drop at the bar after 8pm. Two bartenders were scheduled where three is typical. I drafted a leaner Wed–Thu roster and flagged it for approval.
When a payment fails, Root retries it on an optimized schedule rather than letting it lapse, and tells you only if it still can't collect.
Root finds the best available person for an open shift, by role, hours, and availability, then reaches out to confirm them.
When a regular's pattern starts to fade, Root flags it and drafts a personal message for one-tap approval, before they're gone for good.
A dip in revenue or a spike in voids is brought to your attention before you would have thought to look for it.
Built right, not bolted on
The AI isn't a separate app calling Root over an API. It runs inside the same system, behind one model layer, a per-business spend cap, and an off switch you hold.
Every model call goes through a single vendor-agnostic layer, so the best model can be used for each job and swapped without a rewrite.
Usage is metered against a per-business spend budget, so the cost of intelligence is known and bounded, never a surprise.
When the AI lacks a tool to do something, that gap is written down, so the product keeps improving toward what people actually ask of it.
One toggle in Settings turns Root AI off for the whole business. When it's off, no model call is made and no data leaves the database.
Private by default
Root AI reads your live records to do its work. It never trains a model on your data, it obeys the same per-business isolation and permissions as the rest of Root, and it can only see and do what the signed-in user can.
It reads your live data to do its work, but your data is never used to train models, yours or anyone else's.
Conversations are kept on a fixed schedule and then deleted. Nothing lingers indefinitely.
It respects the same per-business isolation and permissions as everything else. It can only see and do what the signed-in user can.