64 workflow nodes across 8 categories, 4 cross-cutting engines and a five-tier model-routing foundation. Every node is draggable into the workflow builder; AI and non-AI nodes interoperate.
Roadmap — nodes marked Coming soon roll out in phases.
The nodes below compose into the applied capabilities your teams use day to day. From conversation and voice to scoring, prediction and risk & compliance — open any capability for the full picture.
Conversational AI across web, WhatsApp and app that answers from your data and acts.
Transcribes, analyses and summarises every call, in English and Indian languages.
Reads the emotion behind every call, chat and ticket and tracks how it trends.
Sends each lead, call and ticket to the right owner by skill, load and priority.
Scores every lead and deal by real intent, with the breakdown behind each score.
Builds living segments and personas that update as behaviour and value change.
Forecasts conversion, churn and revenue from your own history, not market averages.
Turns activity signals into intent and fires behaviour-based plays at the right moment.
Surfaces the sequences and correlations behind wins and losses, ranked by impact.
Catches the transaction, process or access that does not fit before it grows.
Suggests the next best action, product or offer, grounded in each customer record.
Drafts replies, summaries and content grounded in the record, in your tone, safe to send.
Ten non-AI primitives form the backbone of every workflow. They orchestrate flow and data movement, and provide the connective tissue that AI nodes plug into.
Start a pipeline on lead creation, stage change, or inbound message
Route flow based on CRM field evaluation
Update records, create tasks, assign owners
Pause between steps; time-based nurture
Bulk-process matching records
Fire outbound calls to external endpoints
Normalise data between steps and systems
Fan out into concurrent paths
Re-converge parallel branches
Close the pipeline and log outcome
These nodes read text and record data and emit scores, labels, and structured signals used to route, prioritise, and trigger downstream steps.
Route high-score leads to senior reps; trigger callback for 80+
Escalate negative sentiment; flag positive for testimonials
Route to correct department; trigger demo on purchase intent
Auto-populate CRM fields; create tasks from extracted deadlines
Tag records; trigger topic-specific workflows
Route to language-specific agent; select template language
Auto-archive spam; block from reaching agents
Set SLA timers; escalate P1 immediately
Trigger retention workflow; alert account manager
Identify disengaged customers; trigger re-engagement
Decision nodes replace brittle IF/ELSE chains with AI judgement that returns a decision plus its reasoning, enabling compound-signal branching.
Auto-execute top recommendation or present options
Intelligent round-robin by expertise and capacity
Auto-qualify; send nurture to unqualified
Compound-signal escalation, not single thresholds
Auto-approve within limits; queue exceptions
Dynamic branching replacing complex IF/ELSE
AI judgement for ambiguous scenarios
WhatsApp for urgent, email for formal, SMS for reminders
Generation nodes produce channel-ready content — messages, emails, proposals, quotes, and call scripts — from CRM context.
Auto-generate personalised follow-ups
Personalised outreach, follow-ups, responses
Auto-generate when deals reach a stage
Generate quotes on purchase intent
Prepare agents before outbound calls
Auto-create after meetings, demos, missed calls
Generate A/B test variants
Knowledge nodes answer questions from indexed content; agent nodes are tool-using AI agents that act within your permissions and approval rules.
Auto-suggest articles during ticket handling
Auto-respond to common questions
Upsell / cross-sell suggestions
Policy or contract question answering
Internal policy auto-answers
Autonomous or assisted sales conversations
Autonomous or assisted support resolution
End-to-end campaign management with A/B testing
Dynamic pricing during negotiation
Handle price objections in automated conversations
Suggest best-fit solutions
Prediction nodes run pre-trained ONNX models (some with LLM assistance) for fast forecasts inside the workflow on conversion, churn, revenue, activity, and campaign performance.
Prioritise follow-ups; allocate resources to high-probability leads
Trigger retention workflows; alert CSMs for at-risk accounts
Feed dashboard widgets; alert when forecast drops
Pre-schedule outreach; proactive staffing
Optimise campaign parameters before launch
Utility nodes provide memory, personalisation, scheduling, resilience, and channel-specific intelligence that other nodes draw on.
Maintain context across sessions
Enrich records; feed other agent nodes
Customise every touchpoint per recipient
Auto-generate follow-up tasks
Best-time scheduling for outreach and meetings
Intelligent retries for failed actions
A/B test and auto-select best variant per segment
Real-time voice intelligence
Intelligent WhatsApp conversation handling
Integration nodes connect the AI layer to internal data and external systems. They carry no AI cost but depend on external API availability and latency.
Fetch deals before churn prediction; fetch prior deals before proposal generation
Enrich leads with external data; push qualified leads to marketing tools
Sync contacts, deals, tickets, and AI scores with external systems
Trigger reports, campaign sends, bulk updates, dashboard widgets
Four engines operate across all nodes and surfaces, providing shared intelligence rather than being tied to a single workflow step.
Powers Next Best Action, product recommendations, and personalised suggestions across surfaces
Flags outliers in pipeline, activity, and account signals to trigger alerts
Drives revenue, activity, and campaign projections feeding dashboards and alerts
Records reasoning behind every AI decision to preserve auditability and user trust
Requests route to the cheapest capable tier and escalate only when needed — with retrieval and agent orchestration underneath.
Compose agents and workflows from the library — grounded in your governed data.
