Book a demo

See the decisions, risks, and dependencies your project management tool misses.

Orbit reasons across the work and context in a Space. It surfaces what is missing, connects related evidence, and helps the team decide what needs attention next.

Book a demo See how Quely works
Threads Orbit Chat
what’s the update on QUEL-418 this week? 11:40 AM
Orbit
Analyzing work item context
Retrieving information processing

Here is the current status of QUEL-418, “Rate-limit the ingestion API”:

  1. Rate limiter merged

    • Token-bucket shipped behind ratelimit_v2; PR #2291 merged with CI green.

  2. Load tested at scale

    • Held at 5k req/s with p99 under 180ms.

  3. Rollout is the last step

    • Staged rollout owned by Dwayne, targeted Thursday.
    • Redis quota store (QUEL-402) is already Done.
Actions 3 BETA
Add tool
Ask Anything
Connects approved sources
Jira Jira Azure DevOps Azure DevOps GitHub GitHub monday.com monday.com Google Meet Google Meet Microsoft Teams Microsoft Teams Google Docs Google Docs Confluence Confluence
Bring it in

Bring related work and context into one Space.

Choose the documents, conversations, meetings, tickets, and updates Orbit should reference. Orbit reasons across those sources and surfaces the context your team can discuss and act on.

Type Info 3Connect
Orbit

Where can I find the work and context for our move off the legacy billing system?

Choose the sources I should reference, then tune how I reason over them.

What should I focus on?

RisksMomentumDecisionsBlockers

How far back should I look?

30 days90 daysAll time

Anything else I should know?

Describe what I should pay attention to
Understanding accuracy 57%
Add to space
Before work starts

Ask better questions before work starts.

Orbit reviews the task and its connected context to surface missing information and open questions before the team commits.

QLY-214 / Threads
QLY-214 From Jira

Improving chatbot handoff accuracy

Orbit Orbit raised 4 questions before kickoff
? Have we accounted for rate-limiting on the handoff endpoint under burst load? Open question
? Who owns the fallback flow when the agent transfer fails mid-session? Open question
? Is the EU data residency path in scope, or deferred to a follow-up? Open question
? How do we measure “accuracy” here: resolution rate or context retained? Open question
Answer now Send threads
QUEL-418 / Orbit Chat
what are the risks and blockers on this task? 05:12 PM
Orbit

Here are the open risks and blockers on this unit right now:

1. Blocked on the intent model

Handoff currently relies on explicit keywords like “agent” or “human support”, and misses softer / frustrated language. Retraining is blocked until:

  • The labelled escalation dataset is due from Support. The owner is out until Thursday.
  • Staging access to the NLU service is approved (request open 3 days).

2. Risks worth flagging

  • Scope creep: the thread now includes sentiment routing, which wasn’t in the original spec.
  • False positives: a looser trigger risks escalating conversations the bot could resolve.
  • Dependency: the rollout flag ratelimit_v2 must ship before this can go live.

Two of these need a decision from you before the next sprint:

Ask Anything
As work progresses

Ask the questions your ticket cannot answer.

Ask Orbit about a task, decision, blocker, update, requirement, or next step inside your Space. It reasons across the connected conversations, documents, assets, and updates to show the context that needs attention.

Discussion → ticket

Convert discussions into ticket descriptions.

Orbit can prepare a clearer ticket description from the connected discussion. Your team can review the result before it enters the source tool.

Threads Orbit Chat
ONOnyi Nwankwo3 min ago

We’re seeing too many chatbot conversations stay with the bot even after the customer clearly asks for a human. The biggest issue is when they say “can I speak to someone?”…

0Resolve0
DKDwayne K.2 min ago

Agreed. The handoff logic only catches explicit phrases such as “agent.” When it does fire, tell the customer what is happening instead of dropping them into a queue.

0Resolve0
Actions 2 BETA
Update descriptionUpdate your description with missing context
AskUncover blind spots and answer questions
Auto-send to threads
Add tool
Ask Anything
TOPIC

Improving chatbot handoff accuracy

💡 Topic·ONOnyi Nwan…·Jun 15, 2026
Description

overview

We’re seeing too many chatbot conversations stay with the bot even after customers clearly ask for a human. The handoff logic relies on explicit keywords (e.g. “agent”, “human support”) and misses softer language like “this isn’t helping” or “I need someone,” causing delayed handoffs, repeated questions, and customers arriving already annoyed.

next steps

  1. Analyse recent conversations
    • Pull the last 30 days of chats where handoff should have fired but didn’t.
    • Label them “correct resolution”, “missed handoff”, or “unnecessary handoff”.
  2. Improve handoff intent detection
    • Add a broader intent layer beyond simple keyword matching.
    • Combine intent with confidence scores and conversation context.

requirements

  • Detect handoff intent from indirect phrasing, not just “agent”.
  • Combine detected intent, bot confidence, and repeated attempts.
  • Show a clear message to the customer when a handoff is triggered.
Ready for review · description draft prepared QLY-214
Lenses

Surface decisions, risks, dependencies, and open questions.

Orbit reads through the conversations inside a Space and organizes them into decisions, actions, risks, opportunities, open questions, and next steps, giving your team a clear view of what was decided, what still needs attention, and what should happen next.

Lens Map / Estimation | Agile Coaches Report 13 items · 7 signals
Report design approved
Landing page is canonical
Launch scope locked
@all As stated last…
The report’s design is…
Ojelola to fold in edits
Teammates to comment
Implement Onyi’s feedback
@aditi@quely.io this i…
Differentiate Orbit/Quely
Market Orbit/Quely as…
quely.io/agile-coaches…
www.canva.com
The Canva editor error…

Estimate, don’t guess

Stop going round in circles.

Orbit reads the work, asks the questions your team would normally raise during planning, and recommends a story point once the task is clearer. You can use the estimate as-is, edit it, or override it.

8 Estimations

Adjust if needed
1235 8132134?

Recommended seniority & time

Junior~10h1.3h / point
Mid-level~5h0.6h / point
Senior~3h0.4h / point

Suggested for review: senior developer, based on task complexity.

Wait for team Submit estimation
An 8-point suggestion is ready for team review

Break into subtasks

~5h total

Orbit suggests a breakdown into 2 subtasks for the team to review.

Select all Review selected (2)
Feature implementation 5 sp #1
Overview

Implement the process and system integration for improving chatbot handoff accuracy based on the provided diagrams.

Success criteria

Accurate handoffs with no errors across the specified nodes and edges.

Testing and validation 3 sp #2

Refine this estimate

6 questions

Use the answers to review the estimate with the team.

What are the current issues with the chatbot handoff that need to be addressed?
What methods or techniques are we considering to improve handoff accuracy?
What are the specific deliverables for this task?
What are the completion criteria for this task?
Are there dependencies that could affect the timeline or execution?
What is the estimated time required to complete this task?
Answer now Send to threads
Why teams run on Orbit

Less guesswork. More momentum.

Orbit works across the entire project lifecycle so your team spends its time building, not searching for context.

Catch scope creep earlier

Surface missing requirements, hidden risks, and unclear assumptions at kickoff, before commitment is made.

Unify related context

Orbit connects active tasks to the bugs, customer requests, specs, docs, and conversations that shape the work.

Keep project state current

As chats, updates, decisions, and documents change, Orbit keeps the latest context available without manual recap work.

Get status without a meeting

Ask what changed, what is blocked, what was decided, or what needs attention next. Orbit reviews the context connected in the Space.

Keep decisions connected to their context

Review an update, milestone, or scope change with the conversation, document, or related work connected in the Space.

Async ⇄ live

Bring meeting context into the work it changes.

Keep the relevant meeting record and notes connected to the related work, discussions, and decisions in the Space.

Start with async conversation

Keep the conversation in one shared space, connected to the task from the beginning.

Schedule a meeting from the space

When the work needs a live conversation, book it from the same space or add a meeting link.

Orbit joins and records

Orbit captures the meeting, transcribes it, and turns the conversation into usable context.

Context comes back to the task

Notes, decisions, and follow-ups are saved against the work item, so they do not disappear after the call.

Orbit
REC Handoff sync · QLY-214 12:04
Live transcript
PriyaIf we persist context, where does the payload live between transfers?
MarcoSession store, 30-min TTL. We’ll need a fallback owner on-call.
AditiAgreed. The EU path is out of scope for this work. It needs a separate ticket.
OrbitCaptured 2 decisions and 3 follow-ups, attaching to QLY-214…
Notes 2 decisions 3 follow-ups
back on QLY-214
Orbit Ready when you are

Build with clarity. Execute with momentum.

Orbit turns conversations into decisions, follow-ups, and next steps your team can act on.

Book a demo See how Quely works