Dixa AI conversation tagging: what it tags, what it costs, limits

Kurnia Kharisma
Written by

Kurnia Kharisma

Katelin Teen
Reviewed by

Katelin Teen

Last edited November 22, 2026

Expert Verified
Hand-drawn illustration of a support agent with a headset at a laptop, a customer holding a phone, and three chat bubbles with a heart, a robot and a typing indicator, beside a large Dixa logo on a yellow circle
TL;DR

Dixa tags conversations in four ways: Co-Pilot suggested tags (EUR 39 per agent per month), a beta Automation AI agent, Flow Builder rules and Mim's automatic tags on its own chats. I would trust the rules first, since you wrote them, and audit any AI tag on a weekly sample before routing on it. Dixa doesn't publish accuracy, language coverage or the plan for the Automation agent.

eesel has no Dixa connector, so I can't offer tagging inside Dixa. On Zendesk, Freshdesk, Gorgias, Front or Help Scout, the AI helpdesk teammate drafts replies and tags tickets by topic. With the eesel CLI, an AI coding agent can upload approved tag definitions and policies, then ask test questions before anything goes live. Helpdesk admin access is still needed to connect it.

How does AI conversation tagging work in Dixa?

Dixa puts a tag on a conversation from one of four sources, and the source decides who is accountable for the tag. A rule you built tags what you told it to. An AI suggestion waits for an agent. An AI agent tags in the background.

Hand-drawn map: a new conversation feeds four sources of tags, Flow rules, the Automation AI agent in beta, Co-Pilot suggested tags that an agent confirms, and Mim auto-tagging its own chats, all landing on the conversation
Four sources of tags in Dixa, and who stands behind each one.

Dixa's own pages describe each one:

  1. Flow Builder rules. Dixa's Flow Builder page says you can automate "tagging, routing, escalations, follow-ups" without code, and its FAQ describes tagging conversations by topic or urgency. Rules act on tags, SLA status and events.
  2. The Automation AI agent. The same page says it analyses conversations as they come in, extracts details, categorises contact reasons and adds tags, so agents start with context.
  3. Co-Pilot suggested tags. Dixa's Co-Pilot page says AI recommends relevant tags from the conversation so agents tag consistently without the manual work.
  4. Mim's own tags. Dixa's Q1 2026 roadmap says every conversation Mim analyses gets an "AI promise resolved" or "AI promise pending review" tag.
Dixa's Flow Builder page hero with a flow of Incoming Conversation, Intelligent Routing, Queue and Offers, and the line that you can automate tagging, routing, escalations and follow-ups, as taken from Dixa's Flow Builder page
Dixa's Flow Builder page puts tagging first in its list of automations.

If you want the neighbouring pieces, my posts on Dixa AI draft replies and Dixa AI conversation summaries cover what else Co-Pilot does.

Which Dixa features tag conversations, and are they live?

Status is where I found the most gaps, so here is the table I would take into a demo.

FeatureWhat Dixa says it doesStatus in Dixa's own pagesSource
Suggested tags (Co-Pilot)AI-suggested tags based on similar conversationsBeta on the Q1 2026 roadmap; listed in Co-Pilot's feature listQ1 roadmap
Automation agentEnriches conversations with summaries, tags, attributes and routingBeta on the Q2 2026 roadmap; not named in Q3Q2 roadmap
Suggested conversation attributesCo-Pilot recommends custom attribute values for wrap-upBeta on the Q2 2026 roadmapQ2 roadmap
Auto-tagging of Mim conversationsTags Mim's own chatsGeneral availability, Q3 2026 roadmapQ3 roadmap
Intent detection in automationsIntent node works in automations for automatic tags and attributesEarly access, Q3 2026 roadmapQ3 roadmap
Contact reason categorisationAdvanced Insights groups contact reasons across channelsDescribed in present tenseAdvanced Insights

My read: Mim's tags are the only AI tags Dixa marks as generally available. The rest ranges from beta to early access, or has no stated status, and the marketing pages describe them in present tense. Ask which ones are on in your account.

What do Dixa's customers say about tagging?

Dixa's product page carries one tagging testimonial. Benedikt Peine of Celebrate Company says that during peak times up to four agents a day spent their full time tagging and assigning tickets, and that they no longer need any.

"Dixa automatically tags all conversations, regardless of channel."

Benedikt Peine, Celebrate Company, on Dixa's product page

It is a vendor-published quote, so I read it as a direction rather than a benchmark. It describes tagging to categories so conversations land in the right queue, which is the best use of a tag: a decision downstream. I found no G2, Capterra or Reddit review that discusses Dixa tagging, so I have no independent user voice on accuracy. My notes on reviewing AI conversations show how to build one yourself from a sample.

Which plan or add-on do you need?

Seats first, then Co-Pilot for suggested tags, and the plan decides what the rules can reach.

ItemPriceNotes
Team seatEUR 49 per agent per month, billed annuallyMim included, billed per conversation
Professional seatEUR 139 per agent per month, billed annuallyStandard AI Intent Detection
Enterprise seatCustomAdvanced AI Intent Detection, Advanced Insights included
Co-PilotEUR 39 per agent per monthPage lists suggested tags; Dixa's pricing page says suggested tags are included
Advanced InsightsEUR 29 per agent per month, included in EnterpriseContact reasons, sentiment, recurring issues
Auto QAEUR 0.10 per reviewAdd-on on every plan
MimEUR 0.40 per conversationOn every plan

All prices come from Dixa's pricing page. A 10-agent team on Team with Co-Pilot pays (EUR 49 + EUR 39) x 10 = EUR 880 a month. On Professional it is (EUR 139 + EUR 39) x 10 = EUR 1,780, and adding Advanced Insights makes it EUR 2,070. Dixa lists no separate price for the Automation agent or Flow Builder tagging, and it doesn't say whether Mim's auto-tags are part of the EUR 0.40. Older Dixa pricing coverage used different plan names, so read the live page, and my Dixa review covers the rest of the platform.

What does Dixa not document about tagging?

Six things, and each decides how far you can lean on the tags.

  • Accuracy. No figure is published for suggested or automatic tags.
  • The tag list. Dixa doesn't say whether the AI picks from tags you created or invents new ones. Dixa's API reference says tagging a conversation needs an active tag, while the bulk endpoint creates a tag if the name doesn't exist. That tells me the data model allows both, not what the AI does.
  • Language coverage. Mim replies in the customer's language, but nothing I found covers languages for tag suggestions.
  • Timing. Whether the AI tag lands at open, at handoff or at close isn't stated, and that decides whether a tag can route a live chat.
  • Backfill. Whether history gets tagged when you switch a feature on is unclear.
  • Plan for the Automation agent. Beta on Dixa's roadmap, with no add-on or plan named.

Dixa's help center at docs.dixa.help returned a server error on every attempt I made, so I relied on Dixa's marketing, roadmap and API pages. If a help article exists that settles these, it beats my read.

How do I set up Dixa tags the AI can use?

Fewer tags beat more tags. Every extra label is another decision the AI can get wrong and another bar in your report that is too small to read. A routine that has worked for me on any helpdesk:

  1. List 12 to 20 contact reasons your team already uses, such as "where is my order", "refund", "address change" and "product question".
  2. Give each a one-line definition and one example, and keep them in a shared note.
  3. Switch on AI suggestions for one queue, so agents confirm them and you can see the misses.
  4. Route on a tag only after a sample shows it is right most of the time. My ticket classification guide has a way to score that.
  5. Re-run the audit weekly for the first month, then monthly.
Hand-drawn weekly loop: sample 20 tagged chats, compare to what you would tag, merge or split tags, fix the rule or help article, with a centre label of tag audit, 30 minutes
The 30-minute tag audit I would run weekly.

For a store, ticket tagging for ecommerce gives a starter list, and the ecommerce contact rate post helps you decide which reasons deserve their own tag. A ticket-tagging glossary entry defines the basics.

How do other helpdesks handle AI tagging?

Mostly the same split between rules and AI suggestions, with different gates. Help Scout documents Topics as a condition rather than an action, and HubSpot gates its category field to Enterprise.

Gorgias, Zoho Desk and Re:amaze each have their own version, covered in my posts on Gorgias tagging and Zoho Desk tagging. Zendesk ticket classification is the most mature of the lot.

Dixa's angle is that tagging is part of an agentic stack: tags feed routing, Mim hands over with context, and Advanced Insights reports on contact reasons. If tagging is your only reason to buy, my comparison of Reamaze vs Dixa and the best AI for Dixa round it out.

Where does eesel fit if you use Dixa?

I'll be direct: eesel has no Dixa connector, so it can't read or write tags inside Dixa. Dixa does have a public API with tag and untag endpoints, but I won't promise a bridge that eesel doesn't ship. What I can say is what eesel does on the helpdesks it supports: Zendesk, Freshdesk and Gorgias, plus Front, Help Scout and HubSpot.

STIL Classics runs eesel on Gorgias and Shopify. In one month the agent prepared about 320 tickets, left about 430 internal notes, and tagged about 80 tickets by topic so the team could see what was coming in, with a person sending every reply. Tags can also steer the agent. At Luxedo, the agent skips tickets tagged commercial, along with anything from a school, gym or municipality, and prepares a draft only for home customers. That is the useful loop: a tag decides what the AI does next.

The STIL Classics customer story on eesel.ai, headline Every Gorgias ticket arrives with the order already checked and a reply already written, with stats for 320 tickets prepared and 430 draft replies, as taken from eesel's customer stories
The STIL Classics story on eesel.ai.

Before it touches live tickets, I would run any tagging agent against past conversations and keep a human in the loop. The handoff best practices apply on any stack, and my guide to agent permissions covers what to allow first.

What would I ask Dixa before buying?

Six questions, in this order:

  1. Show me suggested tags on three of our real conversations. Which are right, and which would an agent reject?
  2. Do the AI tags come from our tag list, or can they create new ones?
  3. Is the Automation agent generally available today, and which plan or add-on includes it?
  4. When is a tag set, at open or at close, and can a rule route a live chat on it?
  5. Are Mim's auto-tags part of the EUR 0.40 per conversation?
  6. Can I see the same tags in Advanced Insights, and are they the same as contact reasons?

If the answers are specific, Dixa's tagging is a real time saver. If they are vague, treat the tags as a draft and keep your own rules for anything that routes money or risk.

Try eesel for ticket tagging on your helpdesk

If you run Gorgias, Zendesk or Freshdesk, eesel's AI helpdesk teammate works like a new hire who reads your help center, drafts the reply and tags the ticket by topic, with your team approving what goes out. Try eesel free, from $299 a month for 500 credits, or book a demo to see it on your stack.

Frequently Asked Questions

Does Dixa have AI conversation tagging?
Yes, in four places. Dixa's pricing page says Co-Pilot includes suggested tags, its Flow Builder page lists auto-tagging and an Automation AI agent that adds tags, and Dixa's roadmap lists auto-tagging of Mim conversations as generally available. Intent-based tagging is covered in my notes on Dixa AI conversation routing.
How much does Dixa AI auto-tagging cost?
Dixa doesn't price tagging on its own. Suggested tags sit inside Co-Pilot at EUR 39 per agent per month, on top of seats from EUR 49 (Team) or EUR 139 (Professional) billed annually. The Automation agent has no listed price. My Dixa pricing breakdown has the full plan math.
Does Dixa tag conversations automatically without agent input?
Dixa says it does, through its Automation AI agent and Flow Builder rules. The Automation agent was in beta on the Q2 2026 roadmap, and Co-Pilot's suggested tags were in beta on the Q1 2026 roadmap. Whether your account has either on needs a demo to confirm. See my guide to automating ticket tagging with AI for the general method.
Can Dixa tags trigger routing or automations?
Yes. Dixa's roadmap lists automations based on custom attributes and tags, and its Flow Builder page describes rules that act on tags, SLA status and events. Intent detection is a separate routing feature, Standard on Professional and Advanced on Enterprise. My intent detection routing guide explains how the two differ.
How accurate is Dixa's AI tagging?
Dixa publishes no accuracy figure, tag list size or language coverage for it. Treat the tags as a first pass and audit a sample every week. My notes on reducing false positives in ticket tagging show how to run that audit.
What is the difference between Dixa tags and contact reasons?
Tags are labels on a conversation that rules and reports can use. Contact reasons are what Advanced Insights, a EUR 29 per agent add-on, categorises automatically across channels for reporting. Dixa doesn't say whether the two share a taxonomy, so ask. Related reading: Dixa AI sentiment analysis.
Is there a better AI for tagging Dixa conversations?
Within Dixa, I would start with what the platform ships and audit it. If tagging is the main reason you are shopping, compare the best AI for Dixa and best Dixa alternatives, and weigh the helpdesk switch honestly.
Can eesel tag Dixa conversations?
Not natively. eesel has no Dixa connector, so I won't promise tagging inside Dixa. On helpdesks it does integrate with, such as Gorgias and Zendesk, the AI helpdesk teammate drafts replies and can tag tickets by topic, as STIL Classics does.

Share this article