The agency that stoppedlosing what it knew.
A VA agency with 150+ assistants was losing institutional knowledge every time someone left. We built an autonomous continuity agent that captures meetings, monitors Slack, and writes SOPs in real time, so the knowledge stays even when the people change.
Knowledge walkedout the door.
The VA industry has notoriously high turnover. At Optinizers, around 27 assistants left each year, moving to full-time roles, pursuing other opportunities, or burning out from the demanding work.
The real cost was not recruitment. It was the institutional knowledge that walked out the door with each departure: years of client context, preferences, and relationship nuance, gone with the person who held it.
70% of client-specific processes existed only in assistants' heads. One person leaving turned a routine into a crisis.
Years of communication history, preferences, and relationship nuance disappeared. New assistants started from zero.
6 to 8 weeks to get a new assistant fully productive, during which client satisfaction dropped and errors rose.
Without documented learnings, new assistants made the same mistakes their predecessors had already solved.
An agent thatcaptures and recalls.
An autonomous knowledge-capture system in five layers, built on LangGraph and LangChain. It listens where the work happens, turns what it hears into structured SOPs, connects it in a knowledge graph, and answers natural-language questions with citations.
Integrates with Zoom and Google Meet to record, transcribe, and analyse every client call, with speaker identification linking statements to people.
A Slack bot passively reads client channels, extracting decisions, commitments, and process changes, and ignoring casual conversation.
Converts extracted knowledge into structured, versioned SOPs, detecting when a process is being described and documenting it step by step.
Connects every captured detail in a semantic graph, so a question synthesises insight across meetings, Slack, and documented SOPs.
New assistants ask natural-language questions and get contextual answers with citations, returning exact quotes with timestamps.
Productive onday one.
New assistants now query the knowledge base from their first day instead of starting from zero.
Thirteen weeksto handoff.
Discovery and integration
Mapped existing workflows, integrated with Zoom, Slack, and Notion, and stood up the infrastructure.
Pilot deployment
Deployed to 3 client accounts and tuned extraction accuracy and SOP-generation quality.
Full rollout
Extended to all 40+ clients and trained assistants on the retrieval interface.
Advanced features
Implemented knowledge-graph connections, semantic search, and cross-client insight synthesis.
Training and adoption
Ran comprehensive training sessions and built custom dashboards for team leads and managers.
Optimisation and handoff
Refined against real usage, added custom prompts for specific client contexts, and completed knowledge transfer.
What it runs on.
We went from losing weeks of productivity every time a VA churned to having new team members productive on day one. The agent does not just document, it understands context.
When a new assistant asks how a client prefers their reports, the system does not just return a doc. It synthesises insight from six months of meetings and Slack into a coherent answer. It is like the institutional memory never left.