Support where members already worked
Members could ask by mentioning the assistant in Slack, sending a direct message or using a slash command. They did not need to leave Slack and search another system.
Customer story Fishburners · Sydney
Member support, moved into Slack.
We built Fishburners a Slack-native member support and onboarding assistant grounded in its approved Notion content and attached PDFs. Members could get consistent answers without leaving the place where they already communicated.
The problem
Useful knowledge was outside the workflow.
Onboarding and facility information existed across Notion pages, supporting documents and PDFs. Members communicated in Slack, so answering a simple question meant leaving their normal workflow to locate and interpret the right material—or waiting for staff.
The repeated questions covered practical issues such as room costs, after-hours access and facility information. That created avoidable interruptions for staff and slower answers for members.
What we implemented
A support assistant with clear boundaries.
The goal was not to let an AI answer everything. It was to make approved information easy to access and make uncertainty visible.
One approved knowledge boundary
The assistant searched Fishburners’ approved onboarding material, including nested Notion content and attached PDF documents such as room and rate-card information.
A safe answer policy
When the approved material was insufficient, the assistant was instructed to say so and direct the member to staff rather than invent an answer.
Context kept inside each conversation
Conversation history remained separate for each direct message or Slack thread, so follow-up questions made sense without mixing different members’ conversations.
Production-ready layer
Built for real Slack traffic.
- ✓ Workspace-specific Slack connection and access controls
- ✓ Slack request-signature validation
- ✓ Asynchronous event processing and streamed replies
- ✓ Retry deduplication, caching and operational timing logs
- ✓ Human handoff when the approved sources did not contain an answer
Illustrative value model
A$18,720 in modelled annual staff-time value.
There is no production usage telemetry or verified deflection rate available, so these figures are a transparent scenario—not a claim about Fishburners’ measured results.
The calculation
75 × 52 × 8/60 × 60% × A$60 = A$18,720/year
A$360
Modelled staff-time value per week
A$1,560
Modelled staff-time value per month
A$18,720
Modelled staff-time value per year
Base-case assumptions
- Repetitive questions
- 75 per week
- Staff handling time
- 8 minutes each
- Successfully deflected
- 60%
- Loaded staff cost
- A$60 per hour
This model estimates six staff hours returned each week and values that time at A$60 per hour. Member time saved, faster responses and reduced staff interruptions are excluded.
Your first workflow
Start with one repeated task and a clear business case.
Describe the workflow. We will review the fit, likely first scope and whether automation is likely to create meaningful value for your team.
Assess my workflow →