Best shared inbox software for customer threads: a checklist, not a ranking

Despite the title, this page does not rank vendors, because no ranking survives a team’s own reply count and channel mix. It gives you the pass/fail criteria and a script to run on any tool. Here, a shared inbox is one queue where a team sees, assigns and answers the same customer threads in Slack and email. A help desk adds ticket fields, SLA reporting and automation, and the criteria below apply to both.

This page is published by Weft, so treat its rows as the publisher’s claims. Only Front and Weft are priced here, both checked on their own pricing pages in November 2026. We did not price or test other shared inbox or help desk products, so run the same script and ask the same written questions of any tool you are considering. Zendesk appears only as a source for email visibility rules. We have not run the test on any vendor.

Shortlist by how many people reply

Prices are annual list prices in USD as of November 2026. Take a team with 4 repliers and 26 readers. On the publisher’s Starter at $39 per seat per month, the bill is 4 × $39 = $156 per month, because its pricing page defines a seat as someone who replies to customers. Front’s page does not say how read-only users are billed, so ask Front in writing.

Front Starter needs a caveat. Its page says Starter supports a single channel type of Email, chat (Resolve) or SMS. Its Slack row carries the note “Yes, if single channel type”, and the page never names Slack Connect. A team running Slack and a support email address uses two channel types, so Starter does not fit unless Front confirms otherwise in writing.

• Up to 2 repliers. The publisher’s Free plan covers up to 2 people who reply and up to 3 channels, and the page calls it free forever. It is the cheapest place to run the test.
• Up to 10 repliers. Front Starter at $25 per seat per month billed annually, with up to 10 seats and 10 channels per license, for a single channel type only, per the caveat above.
• 11 to 50 repliers. Front Professional at $65 per seat per month billed annually, with up to 50 seats and 30 channels per license.
• More than 50 repliers. Front Enterprise at $105 per seat per month billed annually. The page lists no seat limit, which is not the same as promising none, so ask.
• Any size, on the publisher’s plans. Starter at $39 and Growth at $89 per seat per month, both billed annually, list no seat caps and cover Slack Connect and email.

At three repliers, the totals are $75 per month on Front Starter, $195 on Front Professional, $117 on the publisher’s Starter and $267 on its Growth. At twelve, Front Starter cannot hold the team under its 10-seat cap, Front Professional comes to $780 per month, the publisher’s Starter to $468 and its Growth to $1,068.

This is our inference, not a vendor statement: if your Slack Connect channel count exceeds a plan’s channels per license, ask whether you need a second license before you compare seat prices.

Assignment has to work without forwarding

A thread needs exactly one owner, and the owner changes inside the tool. If a handover means forwarding an email, pasting a Slack link into a DM or asking someone to take the thread, the tool fails.

Decision rule: if a reassignment leaves a trace (who held the thread and when it moved), keep the tool on the shortlist. If it leaves no trace, reject it. After any reassignment, check that the old owner’s queue no longer shows the thread.

The failure mode is two agents answering the same customer, because one view shows the thread as unowned and another as assigned. The fix is one owner field and a saved view of threads unowned for longer than a quarter of your first response target. On a one-hour target, that is 15 minutes.

Internal notes stay internal on both paths

Email is where rules differ. Zendesk’s help article says a comment is private when its author is an end user who is not the requester or a CC on the ticket, and that a CC who replies with Reply instead of Reply all gets a private comment. The #public Mail API command defaults to true.

This is our inference, not Zendesk’s wording: when the sender is not on the ticket, the tool’s visibility rule decides. A customer’s reply can land as a private comment your team has to find, and a note written in the wrong mode can reach the customer.

Decision rule: if the timeline does not label each message as internal or public, reject the tool. Also check whether a half-written internal note survives a switch to a public reply in the same composer.

Email intake and sender identity

The test needs a working email path, so set it up before minute zero. Ask the vendor which of these it supports: forwarding your support address to a tool-generated address, connecting the mailbox directly, or changing MX records. Then check three things.

• Reply identity. Replies must go out from your own support address, not a vendor domain, or customers will reply to an address you do not control.
• Existing threads. Forward an old customer email and check that the tool attaches the reply to the existing conversation rather than opening a duplicate.
• Rollback. Know how to switch the forwarding off. A tool that cannot be disconnected cleanly is a lock-in risk.

Decision rule: if a reply to the old thread opens a second conversation, or replies go out from a vendor domain, reject the tool. Send the test email as “Test E: does this thread attach?” so you can find it.

Slack Connect as a first-class channel

Slack’s Help Center says each Slack Connect channel can host up to 250 organizations, including your own. That cap counts organizations, not messages, so it does not tell you how a tool splits threads. The two-message step in the test below checks that directly.

Slack’s help article says an inviter chooses whether an invited person can “Can post & invite” or “Can only post”. This is our inference, not Slack’s wording: “Can post & invite” lets a contact bring more people in, which moves the channel toward the 250-organization cap. Decision rule: give customer contacts “Can only post” unless they need to bring in their own colleagues.

Slack’s shared-channels help article does not say whether people from other organizations are billed or take a paid seat, and the pricing page does not address it either. People from other organizations normally belong to their own workspaces, so we would not expect them on your workspace bill, but that is our inference. Confirm it with Slack in writing, since the cited help article and pricing page don’t address it.

• Sender identity. A reply must post under the agent’s own name and avatar. A bot-labelled reply means reject the tool.
• Thread splitting. Two unrelated top-level messages from the customer should become two threads, with later replies staying in the right one.
• Emoji reactions. A reaction must not mark a thread answered.
• Reporting by channel. Ask the vendor to show first response time filtered to Slack Connect only and to email only. If the report blends them, reject it, because a blended average hides a slow Slack queue behind a calm email one.

Pick the first-response target before you test

Start from the customer contract. If it promises a response within 4 business hours, set your internal target at 2 hours, so a handover or a missed notification does not breach it. With no contract, take your last 30 days of first pickup times, set the target at the 80th percentile and tighten it each quarter.

When AI drafts or replies, write the rule into the SLA first.

• An AI draft that a person reviews, edits and sends stops the clock at the person’s send time.
• An AI reply posted with no person in the loop stops the clock only if your SLA says AI replies count. If they count, report human and AI first responses as separate lines.
• A holding reply counts only if it names an owner and a next step. Example wording: “Thanks, Priya is on it. We’ll send the refund status by 3 pm UTC.” A message that says only “Got it” does not count.

Decision rule: if the tool cannot show whether the first reply came from a person, an AI draft or an AI reply, reject it for this use. Ask: “Can a person see and approve an AI draft before it posts into a Slack Connect channel?” This is our inference: a no means the tool cannot gate AI output, so you would need a rule outside the tool to keep AI from posting unreviewed. The pricing pages checked here state no AI unit or rate, so get both in writing and ask whether AI is in the seat price or billed per resolution.

The 10-minute test with two agents and one test customer

Create one Slack Connect channel shared with a test account in a second workspace. Give the two agents separate Slack logins, and give the test customer one email address the inbox monitors. Copy the table below and fill it in per tool.

• 0:00 to 1:00. The customer posts “Test A: can you see this?” and, about 30 seconds later, “Test B: second request”. Pass: two threads, each labelled with the customer’s name. Fail: merged, or one missing.
• 1:00 to 2:30. Agent 1 assigns Test A to Agent 2 inside the tool. Pass: Agent 2 sees it in their own queue, and no DM or forward was sent.
• 2:30 to 4:00. Agent 2 posts the internal note “Internal: check refund history before answering.” Pass: the inbox labels it internal, and the customer sees nothing in Slack or email.
• 4:00 to 5:30. Agent 2 sends the public reply “Thanks, we’re checking this now.” Pass: it appears in the Test A thread under Agent 2’s own name and avatar.
• 5:30 to 6:30. The customer answers in the Test A thread. Pass: the answer lands in the same thread.
• 6:30 to 7:30. The customer emails the test address from an address not on the ticket. Pass: the inbox records whether it landed as public or private, and the rule behind that result is visible.
• 7:30 to 8:30. Agent 1 adds an emoji reaction to Test B. Pass: Test B still shows as unanswered and the clock still runs.
• 8:30 to 9:30. The customer attaches a file to Test B. Pass: the file appears in the inbox, and a reply can be sent in that thread.
• 9:30 to 10:00. Search the inbox for a word from the internal note. Pass: Test A is found, labelled internal, with no customer-facing copy.

Scoring: a failure in assignment, note visibility, thread merging or per-channel reporting rejects the tool. A slow sync or a missing file goes in the notes column and breaks a tie only between tools that pass everything else. The publisher’s trial runs 14 days with Growth features on and needs no card to start, so run a second test on a live channel before you sign.

What to ask in writing, and how to choose

• SSO and SCIM. Pass: SAML SSO and SCIM are included on the plan you are buying, with a date. If either is missing or planned, require a dated commitment in the contract, or keep the tool for a pilot team with no regulated customer data.
• Audit log. Pass: assignments, status changes and internal notes are exportable with user and timestamp, kept at least 12 months. If not, export weekly through the API and store the files yourself.
• History and export. Pass: all threads exportable with timestamps, and a stated retention period for deleted messages.
• Channel scope. Pass: the plan covers public, private and Slack Connect channels.

Worked example: 3 repliers, mostly Slack Connect customers plus a support address, both tools pass every row of the table. This team uses two channel types, Slack and email, and Front’s page limits Starter to a single channel type. If Front confirms in writing that Starter covers both Slack Connect and email despite that limit, Front Starter at $75 per month beats the publisher’s Starter at $117 per month. Otherwise Starter is out for this two-channel team, you compare Front Professional at $195 per month against $117 per month, and the publisher’s Starter wins on price. If only one tool passes, it wins at any price, and if neither does, keep your current tool and fix the failing rule first, such as the owner field or the AI-draft rule.