DCS-PIS · PREDICTIVE INTELLIGENCE SYSTEM
CurrentDCS-PIS
Predictive Intelligence for modern businesses.
DCS-PIS connects a business’s data into one graph, reasons over it, predicts what comes next, recommends what to do, and — inside governance a person sets — does it. It is an intelligence system, not another analytics dashboard.
The flagship product of Datachondria Systems, an AI-native B2B technology company.
- BUSINESS DATA
- BUSINESS CONTEXT
- INTELLIGENCE
- PREDICTION
- DECISION
- AI ACTION
- OUTCOME
- LEARNING
Good morning
Command Center
- Customers
- 862
- Predicted to leave
- 41
- Value at risk
- ₹4.9L
- Awaiting approval
- 1
Key signals
What changed
Churn
30 days
Evening-class members
Now
4.1%
Predicted
9.6%
Key drivers
- Visits per week
- Late payments
- Class bookings
Decisions
What to do
Recommended · ranked 1st
L3Re-engagement message to the evening group
- Reaches
- 83 members
- Channel
- Cost
- ₹83
- Held back
- 1 in 10
Measured by retention at 30 days, treated against the held-back group.
Agents
What PIS is doing
Five questions
Five questions every business asks. One system that answers them in order.
01Business Intelligence
What happened?
Signals and alerts across revenue, pipeline, churn risk and operations; a daily brief of what changed and why it matters; metrics with one definition each, and “how is this computed” on every chart.
DDCS · PIS/ SignalsEvening-class attendance · 4 weeks
High−31%
against this group’s own baseline
How is this computed
attendance = check_ins ÷ booked_sessions scope evening classes · rolling 4 weeks
One definition — the brief, the report and every agent read the same number.
Late payments rising
Same group, same four weeks
23 packages expire this month
Renewal still recoverable
02Causal discipline
Why did it happen?
Drivers and patterns behind the change. DCS-PIS labels a relationship correlational until an experiment concludes — a causal claim is earned, never assumed.
DDCS · PIS/ Signals · DriversWhat moved with the drop
Correlational · no experiment yet- Fewer visits per week
- Late payments
- Fewer class bookings
- Instructor change, Tue & Thu
- Associationweak → strong
A driver is called a cause only after an experiment with a held-out group concludes.
Propose an experiment03Predictive Intelligence
What is likely to happen?
Churn, lifetime value, lead conversion, deal win and revenue forecast, each with calibrated confidence. When the data underneath is incomplete or stale, the prediction is withheld and the reason is shown.
DDCS · PIS/ Predictions · ChurnChurn · next 30 days
CalibratedNow
4.1%
Predicted
9.6%
± 1.8 pts at 80% confidence
- PPriya K.High0.71
- AArun M.High0.64
- NNeha S.Watch0.38
04Decision Intelligence
What should we do?
Recommendations ranked by expected value — what, for whom, when, through which channel, at what cost, with the downside stated. Each carries an executable action, or the reason it cannot execute.
DDCS · PIS/ DecisionsRanked by expected value
3 recommendations
Ranked 1st
L3Re-engagement message to the evening group
- Who
- 83 members
- When
- This week
- Channel
- Cost
- ₹83
action attachedsend_message · measured against 1 in 10 held back2ndRenewal reminder for expiring packages
Action attached · email
3rdMove Tuesday's class to 7 pm
Cannot execute — no write access to the booking system
05AI Agents inside governance
Can AI execute it?
Sixteen catalogued agents, each with a role, allowed tools, data scope and budget. Read-only and low-risk actions run and are logged. Customer-facing and financial actions queue for a person to approve, modify or reject. Six actions are available to no agent, ever.
DDCS · PIS/ ApprovalsNeeds your approvalRetention agent
Drafted 2 minutes ago
Send the re-engagement message to 83 evening-class members
- Expected impact
- Retention at 30 days, against the held-back group
- Confidence
- 0.82 · calibrated
- Evidence
- Churn model · the four drivers
- Why approval
- Customer-facing message — level 3
ApproveModifyRejectL0
Read-only
L1
Low-risk, automated
L2
Configurable approval
L3
Explicit approval
L4
Never autonomous
Never in any agent’s tools: refunds, contracts, compensation, terminations.
Evening-class attendance · 4 weeks
High−31%
against this group’s own baseline
How is this computed
attendance = check_ins ÷ booked_sessions scope evening classes · rolling 4 weeks
One definition — the brief, the report and every agent read the same number.
Late payments rising
Same group, same four weeks
23 packages expire this month
Renewal still recoverable
HOW DCS-PIS WORKS
From business data to business action.
- 01
OBSERVE
Understand the business
- 02
ANALYZE
Find patterns and drivers
- 03
PREDICT
Anticipate what comes next
- 04
DECIDE
Recommend actions
- 05
ACT
Execute through workflows and agents
- 06
LEARN
Measure outcomes
and back to the data
Then and back to the data.
Use cases
What it looks like in a working week.
- 01
Businesses drowning in disconnected data
The CRM, the books, the ad accounts and the support inbox describe the same customers differently.
One graph, one definition per metric, one place to ask.
Customers · Metrics - 02
Teams spending hours turning data into decisions
A dip in pipeline used to mean a week of exports.
Recommendations arrive ranked, costed, and with the action attached.
Sales · Actions - 03
Companies wanting predictive visibility
Which accounts are likely to churn, what each is worth, and what to do about it.
Churn, value, conversion and revenue, ahead of time, with confidence stated.
Customers · Insights - 04
Organisations exploring AI agents
Agents sound useful and feel risky.
Sixteen catalogued agents, each inside a permission grant you set.
Agents · Governance - 05
Businesses operating across multiple systems
Every function runs its own tool, and none of them agree.
Connectors for the CRM, commerce, finance, communication and advertising tools already in use.
Connections · Data
Capabilities
Capabilities
Ten capabilities, one system. Each exists in the product today.
Business IntelligenceContinuous monitoring of the business, not a static dashboard.
- Signals and alerts on revenue, pipeline, churn risk and operations
- A daily brief of what changed and why it matters
- Every metric has one definition, inspectable
Screen · Command Center
Predictive IntelligenceChurn, lifetime value, conversion, deal win and revenue forecast — with calibrated confidence.
- Each prediction carries its confidence
- Predictions are suppressed, with the reason shown, when data health is below threshold
- Models are retrained on a schedule and monitored for drift
Screen · Predictions
Business AnalyticsAnalytics as a view over one business graph, not a separate store.
- Revenue, sales, marketing, customer and operational KPIs
- Custom metrics without engineering
- Reports that can be scheduled and shared
Screen · Metrics
Decision IntelligenceRecommendations ranked by expected value, each with an action attached.
- What, who, when, channel, spend, expected value and downside
- A recommendation that cannot execute says why
- Causal language only when an experiment supports it
Screen · Actions
AI CopilotAsk the business a question; get data, analysis and a recommended action.
- Diagnostic, predictive, causal, prescriptive and operational intents
- Answers cite the data they used
- Actions route through governance
Screen · Copilot
Business ContextOne typed model of customers, deals, campaigns, invoices, people and events.
- Every module reads and writes the same entities
- Events are catalogued; an event not in the catalogue cannot be emitted
- Data Health reports completeness and freshness
Screen · Customers
AutomationWorkflows built from triggers, conditions and governed actions.
- Draft campaigns, follow-ups, alerts and tasks
- Simulate a workflow before enabling it
- Every run logged
Screen · Automations
AI AgentsSixteen catalogued agents, each with a role, tools, data scope, budget and escalation rule.
- Executive brief, sales follow-up, CRM hygiene, retention, finance monitoring, data quality and more
- An agent that behaves unusually is throttled to suggest-only
- Some actions are never available to any agent
Screen · Agents
Business MemoryThe organisation remembers: policies, goals, decisions, experiments, customer history.
- Tenant-isolated
- Decision memory is append-only: what was decided, by whom, why, what happened
- Injected into Copilot and agent context
Screen · Memory
IntegrationsConnects to the systems the business already runs.
- CRM, commerce and payments, finance and HR, communication, advertising
- Generic REST and webhook connectors for the rest
- Secrets encrypted at rest
Screen · Connections
Governance
Autonomy is granted, not assumed.
Every action an agent proposes is classified before it runs.
- Level 0
Read-only
Pull a report, summarise an account.
No approval
- Level 1
Low-risk automated
Log an activity, update a score, tag a record.
Runs, and is logged
- Level 2
Configurable
An internal notification, a draft campaign.
Your organisation decides
- Above this line, a person approvesLevel 3
Financial or customer-facing
A discount, an outbound email.
Explicit approval, every time
- Level 4
Highly sensitive
Never automated, with or without approval.
Never automated
No agent can reject a candidate, terminate an employee, change compensation, promote, issue a refund, or alter a contract. These are absent from every agent’s tools, not merely restricted.
Every run is logged. Every organisation’s data is isolated at the database. An agent behaving unusually — even within its permissions — is throttled to suggestions until a person releases it.
Integrations
Connects to what you already run.
CRM
- HubSpot
- Salesforce
- Zoho CRM
Commerce & payments
- Shopify
- Razorpay
Finance & HR
- Zoho Books
- RazorpayX
- Keka
Communication
- Slack
- Exotel
- Email (SMTP)
Advertising
- Meta Ads
- Google Ads
- LinkedIn Ads
Generic
- REST / webhook
- ERP (REST)
- Inventory (REST)
See DCS-PIS on your own business questions.
A working session with the people building it. Bring the systems you run and the questions you cannot currently answer.