How it works
What exactly is it doing?
Most support software is a filing cabinet: it records the problem and waits for a human to solve it. Boltsolve is the opposite. The filing is a side effect; the work is the point. Below is the whole path a message takes, from the channel it arrived on to the reply your customer reads.
Step one · the sources
Where the work can come from
Each source is a rule you configure, not an integration you have to live with: an address, a form, a feed of alerts. Every source carries its own product, its own default urgency and its own approval policy, and every arrival is stored in full.
Support inboxes
LiveForward any address you already publish — support@, help@, billing@ — and every reply threads back onto the same ticket.
Shared team mailboxes
LiveAdd as many addresses as you have products or brands. Each one carries its own default urgency and its own project.
Web and in-app forms
LivePost a form or a contact page straight into the desk, with the fields your customers already fill in.
Your own team
LiveStaff raise a ticket by hand for anything spotted in a call, a meeting or a hallway conversation.
Error and uptime alerts
Rolling outRoute machine alerts into the same queue so a failing checkout opens a ticket before a customer writes in.
Chat and social messages
Rolling outLive chat transcripts and direct messages land as tickets, with the conversation attached.
Phone and voicemail
Rolling outMissed calls and voicemail become written tickets, transcribed and summarised on arrival.
Reviews and app stores
Rolling outPublic complaints get picked up as work rather than sitting unread in someone else's dashboard.
Step two · the handling
What the AI does with it, in order
- 01
It reads everything that arrives
Every message from every connected source is captured in full — sender, thread, attachments, and the whole history — and stored against one ticket per conversation, so nothing lives only in one person's inbox.
- 02
It works out what the message actually is
The AI classifies each arrival as a bug, a request or a question, sets urgency from the language used, and attaches it to the right product. An outage reads as urgent even when nobody typed the word.
- 03
It investigates before anyone is paged
For a defect, the AI opens the affected codebase, reproduces the fault, finds the cause and writes the fix as a reviewable change with tests — the same work a developer would do, started within seconds of the email landing.
- 04
Nothing ships until a human says yes
Every fix stops in an approval queue with the diagnosis, the proposed change and what it touches. You approve, you edit, or you decline. Routine requests can be pre-approved; anything sensitive never is.
- 05
It answers the customer and closes the loop
Once approved, the change is released and the reply goes back on the original thread in your own words. The ticket resolves itself, with the full trail of who decided what.
- 06
It counts the time and shows the portfolio
Every ticket carries the real time spent from first touch to resolution, rolled up per product and per client — so support becomes a number you can bill, staff and forecast against.
Step three · the control
What it will never do on its own
Approval-first by default
The AI proposes; a person decides. Auto-resolution is opt-in, per source and per urgency, and can be switched off in one click.
Scoped to what you allow
Work can only ever touch the systems on your own allowlist. Anything pointed elsewhere is refused and recorded instead of quietly run.
Isolated per customer
Each workspace's data is separated at the database level, not by a filter in the application — a query that forgets whose data it wants returns nothing at all.
Every action is on the record
Arrival, classification, investigation, approval, release and reply are all timestamped and attributed, so an audit is a page you open rather than a project you run.
Questions
- What exactly does BoltSolve do?
- It watches every channel your customers reach you on, turns each arrival into one ticket, works out what it is and how urgent it is, then investigates it and prepares the actual fix. A person approves, and it ships the change and answers the customer on the original thread.
- Will it change my software without asking?
- No. Approval is the default state: work stops in a queue with the diagnosis and the proposed change until someone accepts it. You can pre-approve narrow categories of routine work if you want speed, and nothing outside your allowlisted systems is ever touched.
- Which sources can it monitor?
- Email addresses you already publish, additional team mailboxes, web and in-app forms, and tickets your own staff raise are supported today. Error alerts, chat and social messages, phone and voicemail, and public reviews are rolling out, and each one arrives as a ticket in the same queue.
- Do my customers need an account?
- No. They keep emailing the address they always used. Accounts are only for the people working the queue and, on Scale, for your clients viewing their own workspace.
- How is my data kept separate from other customers?
- Separation is enforced in the database itself rather than by application filters, so a query that fails to say whose data it wants returns nothing instead of somebody else's tickets.
- How does billing work?
- Subscriptions are monthly or annual, per workspace, and you can change plan or cancel at any time. Annual is two months cheaper. Book a call if you need invoicing or a purchase order instead.