// Practice · Support operations

Enablement

How I kept a growing support team's knowledge findable, current and usable, by people and later by an AI agent.

Concept illustration of a runbook and automation workflow on a monitor; not a product screenshot.
My role
Customer Success Manager, then the first enablement lead, then operations owner, at the same B2B SaaS company
Team
Grew from about 10 to about 50 while I was there
Type
Practice
  • Confluence
  • Intercom
  • Jira
  • Slack
  • NotebookLM
  • LearnUpon
  • Atlassian Rovo

The approach

Answers were scattered across conversations, documents and people's memories, and every new hire started from scratch. My approach was the same in every role: put the answer where the next person will look, write down how it stays current, and make maintenance part of someone's job rather than a favour. Later, when an AI agent started answering from the same documents, that habit turned out to be the whole game: the agent is only as good as the page it reads.

The agent is only as good as the page it reads.

What I built

  • Internal knowledge base

    Scattered internal FAQs consolidated into one Confluence knowledge base, with operating procedures and a knowledge-base update built into every new CSM's onboarding

    Confluence

  • Customer help center

    Organized the customer-facing articles; later maintained with the Support and web teams

    Intercom

  • Integration-break runbook

    When a client integration broke: reproduce, document, escalate with the research done, and stay the customer's single contact through resolution. Deep client-side builds belonged to the technical services manager

    Jira, Confluence

  • Escalation paths

    One route for critical product blockers between CS, Product and Engineering, with a feedback loop to Development and UX, across a distributed team

    Jira

  • Enablement function

    The company's first dedicated CS enablement: onboarding curricula, release documentation, playbooks, coaching and continuous training

    Confluence, LMS

  • Learning management system

    Part of the migration from WorkRamp to LearnUpon, then ran the LMS behind internal onboarding and customer certification, in partnership with the VP of Customer Success

    LearnUpon

  • Technical runbooks and bug-reporting standards

    Written for Support in the enablement years: how to reproduce, what to capture, when and where to escalate

    Confluence, Jira

  • CS operations portal

    One living document for every operations project, with a Jira intake board, a Slack support channel and a weekly call

    Confluence, Jira, Slack

  • Tech-stack onboarding

    One-on-one onboarding to the CS tools for new hires, including people outside CS

    Intercom, Planhat, Jira

  • Knowledge notebooks

    NotebookLM knowledge bases for several teams, on a standard instructions template, with scenario packs and a methodology page

    NotebookLM

  • AI workshops

    Hands-on sessions on prompting, call intelligence with NotebookLM and the vendor's MCP server

    NotebookLM

  • Internal AI assistants

    Two assistants, one grounded in the customer platform and company docs, one in the CS tech stack. Built and used by me and a couple of teammates; not rolled out org-wide

    NotebookLM, Atlassian Rovo

  • The AI agent's knowledge corpus

    The help center, developer docs and internal FAQ, audited page by page before an AI agent was allowed to answer from them

    Intercom, Confluence

How it stays current

  • Procedures with add, update and retire rules, so a page has an owner and an exit.
  • A knowledge-base update as a learning task in every new junior CSM's onboarding. New people find the gaps; fixing one is how they learn the system.
  • A weekly operations call, a Slack channel for questions, and a Jira intake board, so requests and fixes have one door.
  • For the AI agent: every miss logged with a root cause (doc issue, agent issue, unclear), and the fix made on the source page, not in a prompt.

Where it paid off

  • The same audited corpus later grounded the company's customer-facing AI agent, so fixing an article fixed the agent too. That work is the Fin case study.
  • Account managers found answers faster with the knowledge notebooks, and used them for client dossiers during handoffs.
  • Moving the LMS reduced software spend.

I do not have time-to-ramp, deflection or completion figures for any of this, so there are none here.

Where this corpus went nextThe Fin case study