Every conversation lands in
one place, routes itself,
and shows up in our reporting.

Problem

our processes, guides, and training videos live across Slack, Loom, Google, Tango, and more. Staff struggle to find what to send a customer, new hires struggle to learn, and an AI agent can't use any of it.

Solution

make the Zendesk help center our one library.

• Customer-facing articles the AI agent can also use.
• Internal-only sections for staff procedures, hidden from customers.
• The AI agent can also pull in documents we keep in Confluence, SharePoint, or Google Drive if we don't want to move them.

Benefit

one place to look, one place to maintain, and the foundation every AI item below depends on.

Problem

our agents type the same answers to the same questions, over and over.

Solution

a shared library of approved replies.

• One click drops in the approved answer.
• The same click can update the ticket too (owner, category, attachments).
• Shared with everyone, or only with the team that needs it.

Benefit

consistent answers, faster replies, and one place to update wording when something changes.

Problem

customers can't see where their request stands, and there's nowhere for them or other operators to talk shop or give us product feedback.

Solution

a self-service portal and a community.

• Portal: customers track their own requests, reply, reopen, and rate the outcome. Regional managers can see their stores' tickets.
• Community: customers and other operators answer each other, share best practices, and give feedback in one place we can see.
• We turn useful threads into help center articles, which is how they reach the AI agent.

Benefit

fewer "any update?" emails, customers helping customers, industry trends we can spot early, and product feedback collected where we can act on it.

Problem

today's chat turns whatever the customer types into a ticket, so support opens tickets with no context. Meanwhile the redesigned Alleaves site plans to build its own knowledge base and AI agent, duplicating what we already have.

Solution

put our Zendesk AI agent on the site, the app, and email.

• It answers from the help center first.
• It asks follow-up questions and gathers the details support needs before handing off.
• When it can't answer, it hands the conversation to a Support team with the context attached.
• Works on the web app, the mobile app, and incoming email.

Benefit

fewer empty tickets, faster answers for common questions, one AI agent to govern, and no separate build for the site team.

Problem

the website offers customers only a generic email address.

Solution

department forms.

• Separate forms for Support, CX, and Product Escalations.
• Each form asks for what agents need up front.
• Submissions go straight to that department's queue.

Benefit

complete requests land with the right team, not in a shared inbox.

Problem

outside Support, most teams handle customer email from their own inboxes. Nobody else can see the thread, nobody can pick it up if that person is out, and if someone leaves, their open conversations leave with them.

Solution

every department gets a Zendesk address and a shared queue.

• Department addresses such as cx@ and escalations@ come into Zendesk, and replies go out from the department address, not a personal one.
• Each department gets its own queue where the whole team can see open work.
• Staff can forward anything that still lands in a personal inbox into Zendesk, and the customer stays the requester.

Benefit

any team member can pick up where a colleague left off, and customer history doesn't walk out the door.

Problem

customers text managers directly, so one person sees the issue and nobody else can help.

Solution

a dedicated support text number.

• Inbound texts become tickets with full history.
• Texts route and autoreply like every other channel.
• Managers point customers to the new number; Zendesk doesn't document forwarding texts from other numbers.

Benefit

a team works the request instead of one phone, and managers get their time back.

Problem

busy days produce missed calls and thin voicemails.

Solution

connect 3CX now; consider Zendesk phones later.

• Now: connect 3CX to Zendesk so callers are identified and calls are logged.
• Option: move support lines to Zendesk voice. An AI agent answers, gathers the caller's details, and tries to resolve from our knowledge base. It transfers to an available person with the transcript attached; if nobody is free, the call still becomes a ticket with a summary. This adds usage charges.
• Later, if we choose: hand top-tier customers to a dedicated senior team that works them first, on calls and every other channel. Our published support terms would need updating first, since they currently promise first-come-first-served.

Benefit

every call logged and the caller identified today; a path to round-the-clock phone coverage without adding headcount if we want it later.

Problem

tags are inconsistent, so we can't filter or report on them.

Solution

one tag standard.

• A fixed, documented tag list instead of free-form tagging.
• Customer-level tags apply to every ticket from that customer automatically.

Benefit

tags become something we can trust for filtering and routing, and they feed the pattern detection in item 14 so recurring issues show up early.

Problem

someone has to go through the queue manually, work out what each ticket is about, and spread the work evenly across the team.

Solution

let Zendesk assign the work, with AI reading each ticket first.

• Zendesk assigns each ticket by who's available and how much they already have.
• AI triage reads each new ticket and classifies it by topic, customer sentiment, and product on arrival; classification is already included in our plan.
• With the Copilot add-on, triage can act on what it finds: send the ticket to the right team; autoreply with self-service links; ask for missing details before an agent opens it; flag upset customers for faster handling.

Benefit

no more manual queue balancing; work is spread evenly and reaches the right team without anyone reading it first.

Problem

company, contact, and store records are inconsistent and re-keyed.

Solution

one record for everything.

• One record per company and per contact; new users attach to their company automatically.
• One record per store, linked to its people and tickets.
• monday.com connected.

Benefit

when a customer emails in, their details are already there and stay there; no re-keying.

Problem

today we only have Open, Pending, and Blocked. None of them say whether we're waiting on the customer, on Product, or on a third party, so tickets sit and nobody knows why.

Solution

descriptive statuses, with automations that act on them.

• We create statuses that say what's actually happening, such as "Waiting on customer," "Waiting on Product," "Waiting on partner".
• Customers can see a plain-language version of the status too.
• Waiting on the customer: automatic reminders, then we close the ticket if they stay silent.
• Waiting on us: the assigned agent gets a nudge when a ticket has sat too long.

Benefit

everyone can see why a ticket is waiting, and follow-ups on both sides happen without anyone chasing.

Problem

escalations disappear inside Product, and customers reporting the same defect get different answers.

Solution

problem tickets and a Jira link.

• When several customers hit the same defect, we open one problem ticket for the defect and link each customer's ticket to it. Solving the problem ticket solves every linked ticket and sends every customer the same update.
• Engineering keeps working in Jira; their progress flows back to the ticket automatically.
• We measure how long tickets sit with Product.

Benefit

one fix, one consistent message to everyone affected, and a visible Support-to-Product handoff.

Problem

no reliable view of support performance, recurring issues, or backlog health.

Solution

• Dashboards built on standardized fields.
• A free AI app that surfaces recurring issues and their root causes.
• A named owner for the system and a named owner for knowledge.

Benefit

recurring problems surface before they grow, and performance is visible without manual analysis.