Pineapple on pizza

No data available

Feature ownership
Members

Roadmap

Recently shipped

5 new MCP tools

Five tools landed in PostHog's MCP server this week:

Signals

  • Step off a report's suggested reviewers

Replay vision

  • Request session scans
  • Get a session scan request

Annotations

  • List recent production changes

Conversations

  • Start a support ticket by email

Goals

Q4 2026 Objectives

Objective 1: Fix the docs Marco Gancitano

A new user can tell what Support and the agent do, how Support works with the rest of PostHog, and what Business Knowledge is.

  • KR1: Docs explain Support and the Support Agent, including how Support works with other tools such as dashboards and workflows
  • KR2: Docs explain Business Knowledge

Objective 2: General availability for the Support Agent Aleks Veryayskiy Marco Gancitano

The agent is a great experience on real tickets, has clear pricing, and a customer can import their existing support history.

Objective 3: General availability for Business Knowledge Aleks Veryayskiy Marco Gancitano

Customers can store and sync product knowledge, pay for it, and bring an existing knowledge base with them.

Objective 4: Make Support email reliable Aleks Veryayskiy

Customers receive mail, a Mailgun failure does not stop the queue, and an agent can move a conversation from one channel to another.

  • KR1: Incoming and outgoing support email no longer depends on the current Mailgun path
  • KR2: A failed delivery is visible on the ticket and can be retried
  • KR3: An agent can switch the channel on an open ticket, so a conversation that starts in the widget can continue over email

Previous goals

Q3 2026 Objectives

Objective 1: PostHog fully migrates to our own support product

Handle all PostHog support through our own product, replacing external tools.

  • KR1: 100% of support tickets flow through our product (emails, Slack, API ingestion)
  • KR2: All missing features requested by support team are shipped (track blockers to zero)
  • KR3: All docs and internal procedures are updated to reflect the new workflow (Support mini crestSupport Team owns, we unblock)

Objective 2: Launch the Support product to public beta

Make the product available to all users who want to opt-in.

  • KR1: Product is available in feature preview menu and has a product page on posthog.com
  • KR2: At least 1000 customers have handled at least 10 tickets in the Support product
  • KR3: Marketing campaign launched to drive awareness

Objective 3: Build knowledge base product

Give customers a place to store product knowledge (code, articles, docs, text) that AI can search across PostHog.

  • KR1: Customers can add and manage knowledge sources (code, links, text, documents) in a searchable database
  • KR2: AI reply suggestions in the Support product use the knowledge base to find answers
  • KR3: Knowledge base is available via MCP
  • KR4: At least 100 beta customers have added knowledge sources and AI has used them in replies
Q2 2026 Objectives

Objective 1: Launch the Support product to general availability

  • Product is officially GA
  • 100% of PostHog support tickets flow through the Support product
  • Zendesk is stopped

Objective 2: AI bot investigates tickets

  • AI bot performs the initial investigation before a human responds
  • It gathers relevant context and identifies the likely issue
  • Its findings help the human respond faster and more accurately

Objective 3: Business knowledge improves itself

  • Resolved tickets trigger a pipeline that compares the AI investigation with the human response
  • The pipeline identifies missing knowledge and adds it to Business Knowledge as ready for review
  • Humans review and approve drafted knowledge instead of writing it from scratch

Objective 4: Make Support reliable

  • Support Amazon SES as an alternative to Mailgun for the email channel, removing the dependency on a single provider
  • Process incoming events through a dedicated Kafka/WarpStream queue, isolating Support traffic from other workloads

Objective 5: Set pricing for AI products

  • Define and validate pricing for the AI support bot
  • Define and validate pricing for Business Knowledge
Q1 2026 Objectives

Objective 1: Dogfood the support product at PostHog

Start handling PostHog support tickets in our own product alongside Zendesk.

  • KR1: Our support product shows all the same information as Zendesk tickets (messages, metadata, history)
  • KR2: Help sidebar is integrated to create tickets and show ticket history in our support product
  • KR3: Move from 5% to 50% of Help sidebar tickets going through our support product (the rest still go to Zendesk)

Objective 2: Customer alpha for chat widget

Get real customer feedback on human-to-human chat.

  • KR1: Launch the chat widget alpha to 5% of PostHog customers (from the early access list)
  • KR2: At least 10 customers actively use the chat widget

Objective 3: Make AI reply suggestions useful and measurable

Help humans reply faster and better with AI. Improve reply quality via in-product knowledge.

  • KR1: AI reply suggestions are used every day
  • KR2: Content management for AI context is added and used
  • KR3: AI suggestions have thumbs up / thumbs down feedback and are tracked

Handbook

Who we're building for

Support, Customer Success, and Product Engineers at product-led B2B or B2C tech companies, where:

  • The support team is small and can escalate to engineers
  • Most tickets are product questions or bugs, and the fastest way to an answer is seeing what the user did and having good docs to point to

Conversations is a good fit if they:

  • Already use PostHog (or should) and want session replay, product analytics, and other product context right next to each ticket
  • Want AI to investigate and answer tickets, with a human stepping in when it's unsure
  • Need a knowledge layer to power their support answers or their own app
  • Want to track support quality and speed
  • Want to build workflows around their support process
  • Want value right away, without a long setup

Who else at their company might use it:

  • Product Managers to spot where users get stuck or the product falls short
  • Marketing to see common questions and points of confusion

Who we're NOT building for (right now)

Large support orgs and contact centers that need:

  • Big teams of dedicated agents with shift scheduling, SLAs, and complex routing
  • Phone and voice support
  • Support across many languages

Teams that want AI to fully replace their support team and need:

  • An agent that handles everything on its own, with no human in the loop
  • AI that takes actions in other tools, like issuing refunds or changing orders
  • Heavy testing and simulations before anything goes live

Later-stage competitors are a better fit for these teams today. If they try us and churn, that's fine. We haven't built what they need (yet).