Features
A governed operational platform — not a feature list.
Sophie is built around five layers — voice and actions, orchestration, the Insights Centre, governance and security, and the managed service that operates it. Below, each layer with a look at what it actually does.
Voice + actions
Conversations that feel human, governed end-to-end.
Sophie answers, captures, identifies and triages — in natural Australian English with the cadence of a real conversation. Every turn is transcribed, every classification is rule-based, and every escalation is traceable.
- Inbound and outbound call handling, 24/7
- Caller identification against your source system
- Sophie cadence engine — pacing, pauses, interruption handling
- Structured information capture and deterministic triage
- Governed escalation — immediate alerting on trigger conditions
- Call recording and real-time transcription
Orchestration + integration
Configured workflows. Connected systems.
Sophie follows configured conversation flows — not open-ended prompts. Workflows read from and (where agreed) write to your existing systems, so the call ends with the right action having actually happened.
- Governed workflow engine — rule-based, versioned
- Native integrations: AlayaCare, SharePoint, Dynamics 365, Microsoft Graph, Teams, Excel, Power Automate
- Custom REST integrations for line-of-business systems
- Real-time alerting to Microsoft Teams or email
- Scheduled reporting to Excel or SharePoint
- Read-only by default; write-back for agreed interaction types
Inside the Insights Centre
The governance surface, not just a dashboard.
Every conversation Sophie handles produces evidence. The Insights Centre is where that evidence becomes operational visibility, audit-ready records, and a controlled improvement loop.
Operations dashboard
See what Sophie did today, this week, this month.
Live volumes, escalation rate, average handle time, channel split, and Human Intelligence signals — at a glance. Drill-through to the underlying conversations is one click away.
- Real-time KPIs and demand patterns
- Channel and use-case breakdowns
- Activity feed of recent escalations and outcomes
- Human Intelligence Lite — sentiment and friction signals
Caller journey analytics
See how callers actually move through Sophie.
A Sankey-style flow shows the path from caller identity through triage to outcome — verified, transferred, resolved, escalated. Spot the bottleneck. Spot the drop-off. Spot the path you didn't expect.
- Per-stage breakdown with duration and exit reason
- Filter by use case, channel and environment
- Aggregate flow data behind every visualisation
- Direct links to representative conversations at each step
Conversation review
Audio, transcript and structured summary — side by side.
QA, compliance and clinical reviewers get a single workspace: play the call, read the transcript, see the decision trail, mark sensitivity, flag for follow-up. Every review action is recorded.
- Audio player with synchronised transcript
- Sensitivity levels, caller-consent flags, compliance markers
- Reviewer actions logged with name, timestamp and reason
- Linked to escalation, improvement and audit records
Governance & audit
Every change recorded. Every decision traceable.
A platform-wide audit trail captures who changed what, when, and why — across prompts, policies, modules, workflows, settings and users. Old and new values are diffed; environments are tagged.
- Filterable audit log across all object types
- Old/new value diff panels with field-level changes
- Environment chips (dev / test / prod) on every entry
- Aligned with Privacy Act 1988, APRA CPS 234, ASIC RG 271
Prompt management
Versioned prompts with a real lifecycle.
Draft → in approval → published → retired. Every prompt version is a snapshot — viewable, comparable, and restorable. Promotion across environments is explicit, and rollback is one click.
- Full version history with content viewer per version
- Approval workflow with reviewer name and timestamp
- Per-environment publish status (dev / test / prod)
- One-click rollback to any prior approved version
Continuous improvement
Evidence-based recommendations — applied as controlled changes.
Sophie surfaces patterns from the operational data — misrouted intents, slow stages, recurring escalation triggers — as proposed improvements. Each one carries the evidence behind it. Humans approve. Changes are versioned and reversible.
- Categorised by intent accuracy, escalation reduction, caller experience, compliance
- Priority and status tracked end-to-end
- Linked back to the conversations that triggered the suggestion
- Implementation goes through the same governed change process
Cost analytics
What Sophie costs — by use case, channel and environment.
A clear view of total spend, average cost per call, and the breakdown across the work Sophie is actually doing. Useful for budgeting, useful for prioritising, useful for the conversation with finance.
- Total spend, average per call, daily trend
- Breakdowns by use case, channel and environment
- Per-tenant view with environment-level cost isolation
Governance + security
Designed for regulated environments.
Security and governance are not a tab — they are the way the platform is built.
- Deterministic triage — no unsupervised AI judgements in classification
- Versioned, reversible configuration — Sophie does not self-modify in production
- RBAC and SSO support
- Encryption in transit and at rest
- Hosted in Azure East Australia — all data stays in Australia
- Multi-tenant SaaS with logical customer isolation
Managed service
We operate it. You use it.
Sophie is delivered as a managed platform. You don't run infrastructure, you don't tune prompts, you don't manage releases — we do.
- Platform operated by Taidotech — monitoring, maintenance, security updates, infrastructure
- Up to 5 hours per month of managed support included (configuration, issue resolution, monitoring)
- Governed optimisation reviews on a periodic cadence
- Conversation design and prompt management by Taidotech
- Onboarding, training, and go-live support