Guides

Tags

Tags describe what your conversations are about, in your own words. You define a short, deliberate list of them once, and after that both your team and the AI can apply them — the AI is only ever allowed to use tags you created, so the record stays consistent instead of filling up with hundreds of one-off labels.

Tags are available on the Growth plan and above.

What tags are for#

Two jobs, mainly.

The first is finding things: filter the inbox to every conversation about refunds, or every one where the customer is still waiting on you.

The second is building a consistent record of what your customers contact you about. That only works if the same idea always gets the same label, which is why HelpStack asks you to define the list up front rather than letting anyone invent a tag as they go. A tag list that grows on its own quickly turns into refund, refunds, Refund request and money back, and then nothing can be counted.

Reporting on tags is not built yet. Today tags are for filtering the inbox and for the record they leave behind. There is no chart or breakdown of tag volumes anywhere in HelpStack, and the Analytics page does not read them. Applying tags consistently now is what will make that reporting worth having when it arrives.

Two kinds: topic and state#

Every tag is one of two kinds, and the difference matters because they behave differently.

Topic tags accumulate. They are the historical record of what a conversation was about. A conversation that was ever about a refund keeps its refund tag forever, even after the refund is done. Topic tags are never removed automatically.

State tags describe the conversation right now, not its history. Waiting on customer is true until it isn't. State tags are recomputed every time the AI re-reads the conversation, and dropped the moment they stop being true.

A tag's kind is fixed when you create it and cannot be changed afterwards. That is deliberate: flipping a tag from topic to state would retroactively change the meaning of every conversation it was ever applied to.

Categories#

Categories group related tags — "Sentiment", "Delivery", "Product area". They are your own organizing idea and are optional; tags you create without one appear under Ungrouped.

A category can be marked Exclusive, which means at most one of its tags applies to a conversation at a time. Picking a new one replaces the old one. This is what you want for a category like Sentiment, where a conversation should read as either Angry or Happy, never both.

Exclusive only makes sense for state tags, so a category set to exclusive must have State as its default kind, and only state tags can be put in it. Topic tags accumulate by definition and are never displaced, so there would be nothing for exclusivity to enforce.

When your organization first gets tags, HelpStack seeds two categories to get you started, Sentiment and Status, with a few tags in each. They are ordinary tags with no special powers — rename them, archive them, or ignore them entirely.

A note on sentiment. Sentiment tags are a soft signal, useful for spotting patterns across many conversations. Don't wire them to anything automatic. An LLM's read of tone is uneven across languages and cultures — terseness reads as anger in English and as perfectly ordinary in German or Slovene — so it is a poor basis for deciding priority or who gets escalated.

Go to Settings → Tags. Owners and admins can edit the list; everyone can apply tags.

Add Tag — give it a name, pick its kind and colour, optionally put it in a category and add a description.
Add Category — name it, choose its default kind, and decide whether it's exclusive.
Edit — rename a tag or change its colour and category at any time. The kind stays fixed.

Keep the list short. A taxonomy people can hold in their head gets used consistently; one with eighty entries gets used at random.

The description field is worth filling in. It's what tells the AI when a tag applies, so a vague name with a clear description works better than a clever name with none.

Applying and removing tags#

Open any conversation and use Add tag in the header. Tags applied to a conversation show as coloured chips there and on the conversation's row in the list.

Removing a tag is permanent as far as the AI is concerned. When a person removes a tag, HelpStack records that as a decision, not an accident, and the AI will never re-apply that tag to that conversation. This is the mechanism that lets you correct the AI and have the correction stick. If you remove a tag by mistake, you can simply add it again by hand.

How the AI tags conversations#

Shortly after a customer message arrives, the AI reads the conversation and applies tags from your list. It waits for a short quiet period first, so a customer who types "hi" and then spends thirty seconds writing the real question gets tagged on the question, not the greeting.

What it will and won't do:

It only applies tags you have created. It cannot invent new ones.
It adds topic tags and leaves existing ones alone.
It recomputes state tags each pass, adding and removing them as they become true or stop being true.
It never re-applies a tag a person removed.
Inside an exclusive category it can replace the current tag, including one a person applied — otherwise a conversation marked Happy once would stay Happy through a customer's every later complaint.

Tagging runs in the background and never delays or affects the reply the customer gets. If it fails, the conversation is simply left untagged.

Faded chips

A topic tag chip fades when the AI has re-read the conversation several times and not seen that topic come up again. Hover it to see how long it's been. This is a display hint only — the tag is still applied, still counted, and still matched by filters. It's there to tell you a conversation has moved on to something else.

Filtering the inbox by tag#

The conversations list has a Tags filter. Pick one or more tags, and choose how they combine:

Any — show conversations carrying at least one of the selected tags.
All — show only conversations carrying every selected tag.

Use Clear tag filter to go back to everything.

Archiving and restoring#

Tags are archived, never deleted. Deleting a tag would take its history with it, so HelpStack does not offer it.

Archive a tag to take it out of circulation: it stops being offered on new conversations and the AI stops applying it, while every conversation that already carries it keeps it.

Turn on Show archived tags to see them, and Restore to bring one back. An archived tag keeps its name reserved, so if you try to create a tag whose name is already taken by an archived one, restore that one instead of making a duplicate.

Archiving a category doesn't touch its tags — they fall back to Ungrouped.

Conversations & inbox — filtering, searching, and working through the list.
Analytics — response times, AI performance and channel breakdowns. It does not report on tags yet.
Routing & handoff — getting conversations to the right person.

Edit this page on GitHub