Customer Support and Voice AI
AI receptionist for small business: calls, bookings and handoffs
What an AI receptionist does on a real call, the rules that apply, a sample greeting and a pass/fail checklist to launch one safely.
Support Operations
AI Support Copilot
Agents get a drafted reply and the right article the moment a ticket opens, and stay in control of what actually sends.
Your agents write the same kinds of replies all day, and each one starts with a blank box and a search for the right article. This copilot sits inside your helpdesk, drafts a suggested reply for every ticket an agent opens, and surfaces the most relevant article. It never sends anything itself. Your agents stay in charge, just faster and more consistent.
- A drafted reply waiting on every ticket an agent opens
- The most relevant help article surfaced alongside
- Agents edit and send every reply themselves
$4,900
20–30 working days
Optional maintenance $249/mo
What it is
AI Support Copilot sits inside your helpdesk beside your human agents, drafting a suggested reply and surfacing the most relevant help center article for every ticket they open. It never sends a message on its own; your agent reviews, edits and sends.
Why businesses buy it
Good agents are expensive to hire and slow to train, and much of their day goes on typing answers that already exist somewhere. New starters take weeks to find their feet. After launch, every agent starts each ticket with a grounded draft and the right source, so replies are quicker and more consistent while a person still decides what goes out. The Scope Card fixes the channel, systems and workflows before you order, so you know exactly what the copilot drafts and what it never does.
Problems this solves
New agents take ages to write replies they are unsure about
The copilot drafts a suggested reply for every ticket an agent opens, inside your helpdesk. Newer agents have a starting point grounded in your approved content.
Agents hunt through the help center for the right article
The most relevant help center article is surfaced beside each ticket. Agents can check the source quickly, confirm the answer and link the article in their reply.
You do not want AI replying to customers unsupervised
The copilot never sends a message on its own. Your agent reviews, edits and sends every reply, so a person stays responsible for each answer.
A good fit if
Your agents spend real time each ticket searching for the right article or writing a reply from scratch, even for common questions.
You want faster replies without removing the human agent from the conversation entirely.
New agents take a while to ramp up, and a suggested reply grounded in your approved knowledge would shorten that curve.
You are comfortable with AI assisting the agent, as long as nothing reaches the customer without a person reviewing it first.
Not the right fit if
You want the agent to resolve tickets on its own, not just assist a human one. See AI Warranty Support or AI Refund Workflow for that fully autonomous model.
Your main problem is finding and tagging tickets correctly, not drafting replies to them. See AI Ticket Classification.
You want to review how well past conversations were handled rather than assist the next one. See AI Support QA.
You need copilots working across more than six workflow types or an unusually complex approval chain. Talk to us through our Custom AI enquiry form.
How it works
- A ticket opens in your helpdesk and your agent starts reading it exactly as they do today.
- The copilot reads the same ticket and drafts a suggested reply, grounded only in the FAQs and documents you approved.
- It surfaces the most relevant help center articles alongside the draft, so the agent can check the source in one click.
- The agent edits the draft as needed, or writes their own, and sends it; the copilot never sends anything by itself.
- For actions the agent takes, such as updating a record, the copilot can prepare the steps but the agent confirms them under your approval rules.
- Every suggestion, whether used, edited or ignored, is logged, so you can see where the copilot is actually saving time.
What you provide within 72 hours
Your portal lists these the moment you order. Delivery starts the working day after everything is in and your access is checked.
- Admin or API access to the helpdesk your agents already use, with sandbox or test records if available.
- Your FAQs, policies and playbooks, which become the only source the copilot drafts from.
- Ten real ticket conversations so drafted replies match how your team actually writes.
- The name of the person who approves any change to what the copilot is allowed to draft.
- Admin access and API keys for every connected system, shared by adding our team as a user or creating a key you control, never by email.
- Your business information (company basics, hours and brand voice), shared during intake.
- One decision-maker who can confirm scope questions during the build.
- An API key for your own AI model account (for example OpenAI or Google Gemini), created for this build with a spending limit.
The 72-hour rule. Send everything on your checklist, including access, within 72 hours of ordering so your build starts on time.
Timeline
Start
Intake reviewed, agent workflows and playbooks confirmed.
Weeks 1–2
Drafting and article-surfacing logic built around your knowledge base.
Week 3
Helpdesk and CRM connected, visible inside the agent's own view.
Weeks 3–4
Fifty ticket scenarios run, with results sent to you.
Week 4
Your review, plus two rounds of requested changes.
Go-live
Live beside your agents, recorded handoff session, support begins.
| Option | Delivery time | Price |
|---|---|---|
| Standard | 20–30 working days | $4,900 |
| Express | 15–25 working days | $5,750 |
| Rush | 10–20 working days | $6,600 |
What you receive at handover
- The support copilot deployed inside your helpdesk for your agents
- Workflow, prompt and guardrail configuration for up to six workflows with your approval rules
- Connection to up to two systems and up to 100 documents as source content
- Test report covering 50 documented scenarios
- Admin guide, recorded 60-minute handoff session and 30 days of post-launch support
Expected benefits and how to measure them
- Agents can reply faster with a draft to start from.
- New team members can find their feet sooner.
- Customers still hear from a person on every ticket.
- Average handle time
- Compare average handle time per ticket in your helpdesk for four weeks before and after launch.
- Draft acceptance
- Sample tickets weekly and note whether agents sent the draft as is, edited it or discarded it.
- Ramp-up time for new agents
- Compare how long new starters take to reach your target ticket volume, before and after launch.
Before you start: Record average handle time by agent for one month, and how long recent new starters took to reach full volume.
These are things to track, not promised results. Outcomes depend on your data, your team and your market.
Running costs
The build price is one-time. Running AI Support Copilot costs roughly $50 to $250 a month in AI model usage and hosting, on top of your existing helpdesk or CRM plan. More agents drafting more replies sits toward the top of that range.
AI usage during the build and afterwards runs through your own accounts, so the provider relationship and the bill stay visible.
Ask about running costsCustom requirements
Need more than six workflow types, a third connected system, or copilot support for a second team? Add a note at checkout for a written Change Quote, with work starting only once you accept it. Larger or bespoke projects go through the Custom AI request form instead.
Request a custom versionKeep it running
Monthly maintenance covers monitoring, fixes and change hours for work such as updated playbooks or a new suggested-reply category. It starts after the post-launch support period, bills monthly and can be cancelled any month.
$249/mo (the minimum fee, includes 2 change hours)
Optional. Starts after the post-launch support period. Cancel any month.
How maintenance worksHow this compares
| Do it yourself with a SaaS tool | Hire a custom agency | Upmark Labs | |
|---|---|---|---|
| Price | A helpdesk AI-assist add-on bills per agent seat every month | A custom agency quotes after discovery calls | Ours is one fixed price, shown on this page before you buy |
| Scope | A generic assist feature drafts from whatever it can find, not an approved source | An agency agrees scope during the project | Ours is written on the Scope Card before purchase |
| Delivery | An assist add-on can switch on in days but drafts from unreviewed content | An agency build usually takes several weeks, set per project | Ours has a fixed delivery time shown before you buy, with faster options |
| Control | Some assist tools can send low-confidence replies automatically | An agency build depends on what was agreed | Ours never sends a message; your agent always reviews and sends |
Questions about this product
Can AI Support Copilot send a reply without an agent reviewing it?
No. It drafts a suggested reply and surfaces relevant articles, but a human agent always reviews, edits if needed, and sends it themselves. This is what separates a copilot from an autonomous agent: every message a customer receives has had a person's eyes on it first.
What does AI Support Copilot cost to run each month?
Roughly $50 to $250 a month, covering AI model usage and hosting. That is on top of your existing helpdesk or CRM plan, which you continue paying directly. Cost scales with how many agents are actively using drafted suggestions, not a fixed per-seat fee from us.
How long does AI Support Copilot take to launch?
Delivery runs on a fixed schedule from a verified intake, and the exact date is shown in the delivery panel before you buy, with express and rush options if you need it sooner. Every option includes the same fifty test scenarios and a recorded handoff session before your agents start using it live.
What access does AI Support Copilot need?
Admin or API access to the helpdesk your agents use, and usually your CRM as the second connected system, so drafts can reference full customer history. Sandbox or test records work for the build. Credentials are shared by adding our team as a user or creating a key you control, never by email, revocable at any time.
Do customers know a reply was drafted with AI help?
That is your call. Because a human agent reviews and sends every message, many teams choose not to disclose the drafting step, treating it the same as a canned-response library. If you prefer to disclose it, we can add that wording to your reply templates during intake.
Can I update what AI Support Copilot is allowed to draft?
Yes, through monthly maintenance. Its change hours cover work such as updated playbooks or a new suggested-reply category, along with smaller updates to the drafting rules. Your FAQs, policies and playbooks stay the only source the copilot drafts from, so updating them updates its replies. Maintenance starts after the post-launch support period and can be cancelled any month.
How is AI Support Copilot different from a fully autonomous support agent?
A copilot assists a human agent inside every ticket; it never acts or replies alone. If you want the agent handled end to end without a person drafting or sending anything, see AI Warranty Support or AI Refund Workflow, which are built to act autonomously within your rules.
Guides on this topic
All guides Give every agent a first draft, not a blank reply box.
Order AI Support Copilot and every agent starts each ticket with a grounded draft and the right article, while still deciding what is sent.
Works well with
Support Operations
AI Support Copilot for Help Scout
Drafts replies and surfaces relevant articles for human agents inside the helpdesk, connected natively to Help Scout so it reads and updates…
Support Operations
AI Support Copilot for Gorgias
Drafts replies and surfaces relevant articles for human agents inside the helpdesk, connected natively to Gorgias so it reads and updates tickets,…
Support Operations
AI Support Copilot for Intercom
Drafts replies and surfaces relevant articles for human agents inside the helpdesk, connected natively to Intercom so it reads and updates conversations,…