01
Capture transcript and participants
LiveThe meeting becomes structured context instead of a note buried in someone’s inbox.
Workspace
Meeting orchestration
Turn Zoom or Teams meetings into structured follow-ups, CRM actions, product intake, asset requests, and cross-functional coordination with visible agent routing and approval control.
Signals ingesting
Agents reasoning
Actions queued
Motion View
How the meeting workflow moves through the organization
This is the cross-team operating map for a customer conversation: one intake, multiple specialist actions, and a controlled path to execution.
02
Extract business and product signals
LiveRevenue, roadmap, advocacy, risk, and design/build requests are detected from one source.
03
Route work to specialist agents
QueuedFollow-ups, CRM updates, feature demand, design tasks, and engineering work are drafted in parallel.
04
Execute or pause for approvals
WatchingExternal actions and high-risk changes route to humans, while safe actions continue automatically.
Architecture View
Architecture map for meeting-driven orchestration
This is the live platform view: connectors feed context, the run core classifies the meeting, specialists generate work, and enterprise systems receive governed outputs.
01Sources
Meeting connectors
Connect Zoom or Teams to start live transcript intake.
Customer meetings
External conversations route into follow-up, CRM, beta, and advocacy flows.
Internal meetings
Internal calls stay inside launch, build, asset, and coordination lanes.
02Run Core
Classifier
Determines whether the flow is internal, external, or blended before any side effects happen.
Policy engine
Uses approval risk and tenant posture to decide which work pauses and which work executes.
Orchestrator
Chooses specialist agents and fans work into revenue, product, creative, and engineering lanes.
03Specialists
Revenue + follow-up
Handles CRM updates, recap emails, beta enrollment, and account next steps.
Product + engineering
Converts repeated customer evidence into PFRs, build worksets, and architecture planning.
Creative + file generation
Builds decks, calculators, briefs, videos, and reusable files with lineage.
04Enterprise Systems
Approvals
High-risk or external-facing work pauses for human review with full action previews.
Auto-queued work
Low-risk internal tasks move automatically into files and coordination lanes.
Files workspace
Artifacts land in one S3-style workspace with lineage back to meetings and runs.
Workflow Playground
Paste real meeting text and let Synth build the workflow
This is the simplest demo path before full connector auth: give Synth the meeting context, then inspect the branch, the agents involved, and the actions it generated.
Ask Synth
Why this branch?
Synth chose the customer branch because it detected external buying signals, requested follow-up resources, and account-specific asks.
What gets approved?
Customer-facing and risky actions pause for approval first. Internal low-risk work can be auto-queued.
Who is involved?
Jamie Chen, Alex Morgan, Mira Patel
Agent Motion Canvas
Watch the workflow move across classifiers, specialists, and approvals
This is the operator view for a single workflow run. It should make routing obvious without making the user decode a wall of text.
Input
Zoom
Q2 expansion and rollout review
Classifier
Customer
Customer expansion and beta routing
Branch
Revenue + product branch
3 actions routed from one transcript.
Outputs
2 approvals, 1 auto-queued
Review the generated actions, inspect recipients and destinations, then approve or let policy execute.
Active branch
Customer
Signals found
3
Actions created
3
Meeting Command Center
Internal and customer calls branch into different workflows automatically
Customer calls trigger revenue, follow-up, beta, and advocacy work. Internal calls shift toward launch planning, asset creation, product intake, and build execution.
CustomerZoom
Acme Manufacturing
Q2 expansion and rollout review
April 13, 2026
Owner: Jamie Rivers
Detected signals
Beta interestPositive ROI proof pointFeature gap: group-sync roles
We would expand this to three more teams if launch reporting had role-mapped access controls.
Synth will do next
1Update Salesforce opportunity
2Draft beta follow-up email
3Nominate reference candidate
4Open or update grouped feature request
Action Preview Center
Webhook-shaped ingest now produces real action previews
This is the review surface between transcript classification and approvals or autonomy. It shows what Synth inferred and exactly what it wants to do next.
Awaiting approval
2
Auto-queued
2
Drafted
0
Awaiting ApprovalEmail
Draft customer beta follow-up
External customer call included explicit beta interest and a request for ROI materials.
Destination
jamie@acme.com, alex@acme.com
•Recap requested role-mapped reporting gap
•Attach ROI model and beta enrollment note
•Ask for sandbox contact and rollout owner
Awaiting ApprovalCRM
Update Salesforce opportunity
The call signaled expansion intent, broader rollout scope, and a named blocker tied to close confidence.
Destination
Salesforce / Opportunity ACME-443
•Increase opportunity amount to reflect three additional teams
•Set next step for beta-readiness follow-up
•Mark permissions gap as adoption blocker
Auto-QueuedArtifact
Queue launch bill of materials refresh
Internal launch planning call confirmed missing assets and did not require external system writes.
Destination
Artifact workspace /customer-assets/launches/role-based-launch-reporting
•Create demo video brief
•Refresh overview deck narrative
•Update ROI sheet assumptions for launch day
Auto-QueuedEngineering
Open engineering workset from VOC cluster
Internal build-planning call aligned transcript evidence and telemetry around the same adoption blocker.
Destination
Build Studio / PFR-241 workset
•Generate architecture diagram
•Split issue set by permissions service, reporting layer, and audit trail
•Create design brief for rollout visibility states
Workflow Graph
What happens after a customer meeting
This is the transcript-first operating flow: ingest the meeting, understand the business signal, route to the right agents, draft the work, and execute through approvals or policy-driven autonomy.
01
Stage
Meeting captured
Zoom or Teams recording, transcript, attendees, and chat context are ingested.
02
Stage
Signals extracted
The system detects asks, blockers, buying intent, commitments, competitors, and launch interest.
03
Stage
Agent routing
An orchestrator chooses the right specialist agents based on the meeting outcome.
04
Stage
Action plan drafted
Follow-up emails, CRM updates, tickets, assets, and internal reach-outs are prepared with rationale.
05
Stage
Approval or autonomy
Users approve actions individually, by policy tier, or allow safe actions to execute automatically.
06
Stage
Execution and tracking
Tasks are posted, opportunities updated, tickets opened, assets requested, and outcomes tracked.
Post-Meeting Actions
What the system can draft or execute
Customer follow-up email
Why: The customer requested pricing collateral and an ROI model for the next call.
Output
Draft email with recipients, requested resources, and meeting recap.
Opportunity update
Why: The meeting showed expansion intent, timeline movement, and an identified economic buyer.
Output
CRM stage update, next steps, close-date confidence, and action tasks.
Product feature request
Why: The customer said rollout would expand if the platform supported feature X.
Output
PFR with transcript evidence, account count, and revenue influence.
Internal reach-outs
Why: The account needs input from solutions engineering, security, and another project owner.
Output
Draft Slack or email asks to internal stakeholders with context and requested help.
Launch asset request
Why: The next customer call needs a tailored overview deck and deep dive material.
Output
Asset request with owner, due date, source materials, and priority.
Ticket or GitHub issue
Why: Validated demand is strong enough to move into product or engineering intake.
Output
Jira/GitHub issue with business rationale and structured requirements.
Agent Routing
Which agents get called and why
Orchestrator Agent
Chooses workflow branches, selects specialists, and enforces budget and policy.
Model strategy: Use a small reasoning model first; escalate only when branching is ambiguous.
Writer Agent
Drafts emails, recaps, customer resources, and executive summaries.
Model strategy: Prefer a lower-cost writing model for standard recaps and templates.
Revenue Agent
Handles opportunity updates, task creation, buying-signal scoring, and follow-up sequencing.
Model strategy: Use structured extraction plus lightweight reasoning for CRM changes.
Product Agent
Creates PFRs, clusters demand, and drafts product briefs or tickets.
Model strategy: Use mid-tier synthesis only when multiple customer signals must be merged.
Creative Agent
Generates decks, calculators, docs, and routes video work to the right toolchain.
Model strategy: Use task-specific generation tools and avoid large-model use for simple formatting.
Autonomy Modes
How users stay in control
Review every action
Best for new tenants or highly regulated teams. Every side effect pauses for approval.
Auto-run safe tasks
Internal notes, summaries, and asset requests execute automatically while external writes require review.
Policy-driven autonomy
Trusted workflows can execute automatically based on tenant rules, risk tiers, and role permissions.