Current through early August 2026. Updated as Intercom ships; past editions are linked at the bottom.
Intercom shipped around 30 updates over the last six weeks. Your inbox settings haven't changed since whoever configured them left the company.
New routing settings need a feedback loop or you're guessing. Supportman posts every Intercom rating into Slack as it lands and scores the conversations nobody rates, so you can see CSAT by skill and by channel two weeks after you change anything.
Most of those releases don't ask anything of you. Four of them change how conversations get assigned to your team, and if you skip them your queue keeps distributing work the same way it did in June.
Here's what to turn on, and the numbers I'd start from.
Skill-based routing is useless until you define the skills
Shipped July 13. Admins define skills, and conversations only get assigned to teammates who have the matching one.
Every team that turns this on makes the same mistake in week one: they define eleven skills. Billing, Refunds, Subscriptions, Invoices, Dunning, Tax. Now a refund question matches exactly one person, and your queue stalls the second that person goes to lunch.
Here's the rule that keeps it sane. A skill is something that requires a login or a permission other reps don't have. Not a topic.
| Skill | Who has it | What it actually gates |
|---|---|---|
| Billing | 3 reps + team lead | Stripe dashboard access, refunds over $50 |
| Technical | 4 reps | Log access, API debugging, error traces |
| Default queue | Everyone | Everything else |
Three skills. If a fourth doesn't map to a system permission, it's a tag, not a skill.
Then the coverage rule: three people minimum per skill. Two people means one holiday plus one sick day takes the skill offline, and conversations sit unassigned while everyone stares at a queue they can't touch.
Intercom added channel-based specialisation on August 3. Different feature, and worth keeping separate in your head. Skills route on what the conversation is about. Channel specialisation routes on where it came from. Turning both on at once, on the same team, is how you end up with conversations nobody is eligible to take.
Split capacity by channel, then protect the wrap-up
Two releases that belong together. On July 16 Intercom added separate assignment limits for email and Messenger. On July 17 it extended wrap-up time to chat and email, which previously only existed for phone.
The broken default is one capacity number for everything. A rep set to 5 gets five live chats or five emails, treated as identical load. They're not close. A live chat needs a reply inside 40 seconds. An email can sit for three hours and nobody notices.
Starting numbers:
| Channel | Concurrent limit | Wrap-up |
|---|---|---|
| Messenger (live chat) | 3 | 90 seconds |
| 12 | 3 minutes | |
| Phone | 1 | 5 minutes |
Adjust from there by seniority, not upward across the board. Tier 1 handling password resets can run 4 concurrent chats. Your senior reps should sit at 2, because the conversations that reach them are the ones that need actual thinking.
On wrap-up, measure before you set it. Export last month's conversations and look at the gap between the final reply and the close. If your reps genuinely use four minutes to finish an email, setting 90 seconds doesn't make them faster. It just moves the admin work to 6:30pm, unpaid and invisible.
Per-teammate office hours makes your response time honest
Also July 17, and it's the release with the biggest gap between how boring it sounds and how much it changes your reporting.
Before this, one office-hours schedule covered the whole team. If you have a rep in Manila and a team schedule set to 9–6 London, their responses looked bad and their productive hours registered as dead air.
Turn on per-teammate hours and your median first response time will drop without anyone working faster. That isn't gaming the metric. The old number was wrong, and you were probably coaching someone on it.
One warning: fix your SLAs the same day. SLAs adapt to office hours, so changing the schedule moves the SLA clock underneath every open conversation. Check every SLA rule before you touch this, and don't touch it on a Friday afternoon.
Wait for Webhook is the one Fin release that changes what you own
July 24. Fin can now pause a Procedure while an external system does its work, then resume when that system calls back. An identity check, a payment, a bank-linking flow.
Previously Fin had to hand off anything involving a wait. Now it holds the conversation through the gap. That moves a real category of work from your queue to Fin's.
The setting nobody configures is the timeout, and the default is a 7-day ceiling. Seven days is not an escalation policy. A customer whose identity check died silently doesn't open a second conversation. They churn.
Set the timeout to roughly twice the p95 of the external system. A bank-linking flow that normally returns in four minutes should escalate at 15 minutes, not next Thursday.
Then route that escalation to a named skill rather than the general queue. An escalation from a stalled payment Procedure needs the person with Stripe access, which is exactly the routing you defined at the top of this post.
What I'd leave alone
Not everything on a changelog is a decision.
- The modern Help Center (July 23), custom hold music (July 15), macro CSV export (July 10) and side conversation translation (July 8) are all fine. None of them has an ops decision attached.
- Telegram as a native channel (July 28). Only if customers are actually asking. A new channel with no coverage plan is just a slower email queue with worse expectations.
- Merge preview (August 6) shows which attributes survive before you merge duplicate contacts. Pure improvement, requires nothing from you.
- Bulk actions across every search result (July 28) deserves one minute of thought. Anyone who could previously bulk-close 40 conversations a page at a time can now bulk-close 4,000 in a single click. Decide who gets that before you announce it in Slack.
The part Intercom didn't ship
Every routing release above solves the same problem: getting a conversation to the right person. Not one of them tells you whether that person handled it well.
Skill routing sends the billing question to your billing rep. If that rep is having a rough week, all skill routing does is guarantee every billing customer gets the rough week. Concentrating work on specialists concentrates their bad days too.
So when you change these settings, set the check at the same time. Segment CSAT by skill and by channel, weekly, and don't look at the overall number. If Billing sits at 71% while Technical runs 89%, that's a specific routing and coaching target, not a team-wide training problem. There's more on reading those numbers in Intercom CSAT benchmarks.
Give it two weeks. If you can't tell whether the new routing improved anything by then, you didn't change your settings — you shuffled them.
Previous editions
This page carries the current recommendations. Each edition below is the complete dated record for its window, frozen at publication.
- July 2026. All 25 releases, dated and grouped: skill-based routing, per-channel capacity, wrap-up time, Wait for Webhook, Telegram, and the Fin for Sales connectors.
What's new in Intercom FAQ
Should I use skill-based routing instead of teams?
No. Keep teams as your organisational structure and add skills on top for permission-gated work. Skills replace the informal "ask Sarah, she knows billing" habit, not your team setup.
How many skills should I define?
Two or three to start, each with at least three people holding it. If a proposed skill doesn't correspond to a tool login or an approval permission somebody else lacks, use a tag instead.
What's a reasonable wrap-up time for chat and email?
Start at 90 seconds for chat and 3 minutes for email, then check it against your actual data: the gap between a rep's last reply and the conversation close. Set the value your team already needs rather than the one you wish they needed.
What timeout should I set on a Wait for Webhook step?
About twice the p95 response time of the external system, and always well below the 7-day maximum. Route the timeout escalation to the skill that can actually resolve it, not the general queue.
Will per-teammate office hours change my historical reports?
It changes how responsiveness is measured going forward, so expect a step change in median first response time rather than a gradual trend. Note the date you enabled it so nobody reads the shift as a performance improvement.