Rules Plateau. A Learning Engine Outgrows Them
Most recommendation engines in India still run on hand-written rules: if the cart crosses ₹5,000, show free shipping; if the visitor is from Mumbai, show this banner.
Rules are easy to start and impossible to scale, a hundred of them, conflicting, no one sure which still helps. Meanwhile your best customers behave in ways no rule anticipated.
ONE CX replaces static rules with an intelligent decision layer. A decision engine learns instead of guesses: it reads each customer's signals and history and decides what to recommend, what to offer, what to do next, sharper with every interaction, because the data, not a quarterly review, tells it what works.
We Build on Your Stack, Whatever It Runs On Build on Your Stack
We build the decision engine on the platforms you already run, or as custom models on your stack, no rip-and-replace.
rStorefront, a hyperlocal retail platform
Built the recommendation and decisioning layer across 5,300+ stores, eleven AI modules learning from every order to surface the right
SUD LIFE INSURANCE
For SUD Life's agent recruitment, we built the decisioning approach behind the personalisation, statistically scoring, via Statsig, which content and offer fits each prospect by their signals, then letting the best-performing decision scale automatically as the evidence built.
We Build on Your Stack, Whatever It Runs On Build on Your Stack
We build the decision engine on the platforms you already run, or as custom models on your stack, no rip-and-replace.
Recommendation Engine
Show each customer the right product, content or offer. More discovered, more bought, even on day one
Decision Engine
Sequence what reaches each customer and when, offers and nudges timed across app, web and WhatsApp, never clashing
"Walk into almost any enterprise and you'll find the same thing, hundreds of rules built over years, but very little confidence that they're still delivering value.
The fear isn't creating new rules; it's removing old ones. So we don't ask teams to take a leap of faith. We run AI alongside what's already in place, measure every outcome, and let the numbers decide. Over time, the best-performing decisions replace the old rules naturally."
- Chandra Prakash, CX Architect - ONE CX
Decision Stack We Build
Decision Stack We Build
We build the decision engine on the platforms you already run, or as custom models on your stack, no rip-and-replace.
Our Expertise
Custom ML Models
Our pick when the decision is your edge, propensity, recommendation and ranking models built on your data, owned by you.
AWS Personalise
Google Cloud (Vertex AI)
CleverTap
MoEngage
Baseline Architecture
Signals In
Identity, behaviour and history, from the foundation Data Engineering builds, feed the engine.
Candidate Generation
Scoring & Ranking
Business Rules
The Decision
Learn & Retrain
Why India's CMOs and CPOs Trust ONE CX™ to Run the Engine
01
We Move You From Rules to ML Without the Rip-Out
We don't replace your rules overnight, we run them alongside models, prove the lift, then shift the weight to ML as the numbers earn it. Migration without the risk.
We Move You From Rules to ML Without the Rip-Out
We don't replace your rules overnight, we run them alongside models, prove the lift, then shift the weight to ML as the numbers earn it. Migration without the risk.
02
Recommendations That Fit Indian Catalogues
We tune models for sparse data, cold-start products and shared accounts, the conditions that break off-the-shelf recommenders trained on Western behaviour.
Recommendations That Fit Indian Catalogues
We tune models for sparse data, cold-start products and shared accounts, the conditions that break off-the-shelf recommenders trained on Western behaviour.
03
Decisions Tied to Margin, Not Just Clicks
The engine optimises for the outcome you choose, margin, LTV, retention, not just the click, so the smartest recommendation is also the most profitable.
Decisions Tied to Margin, Not Just Clicks
The engine optimises for the outcome you choose, margin, LTV, retention, not just the click, so the smartest recommendation is also the most profitable.
04
Auditable, Tuneable, and Yours
Every decision is logged and explainable, the models documented and handed over, running on your stack. Your team owns the logic and can tune it. No black box.
Auditable, Tuneable, and Yours
Every decision is logged and explainable, the models documented and handed over, running on your stack. Your team owns the logic and can tune it. No black box.
What Indian Growth Leaders Ask ONE CX About Decision Engines
What does ONE CX actually do in a decision engine engagement?
We build the engine that decides what each customer sees, recommendations, offers and next-best actions, rules where they fit, ML where it pays. It scores every option by likelihood to convert, applies your guardrails, and hands the decision to your experience to serve. Learning from each outcome.
Why move from rules to ML at all?
Rules are quick to start and impossible to scale, a hundred conflict, and none adapt to a customer who behaves unexpectedly. ML learns the patterns rules can't encode and improves every interaction. We don't rip rules out; we run them alongside models and shift weight to ML as it earns the lift.
How is this different from experience personalisation?
Two halves of one system. The decision engine decides what should reach each customer, the product, content or offer most likely to convert. Experience personalisation is the system that serves that decision: the DAM, signal routing and activation that put the chosen content on screen. This page is the engine that makes the call; personalisation is the machine that delivers it.
Why won't an off-the-shelf recommender just work in India?
Off-the-shelf recommenders are trained on dense, Western, one-account-one-person data. India breaks all three: sparse catalogues, cold-start products, shared accounts where one login is a household. We tune models for those conditions, else the engine recommends to a person who isn't there.
Can the engine optimise for margin, not just clicks?
We optimise for the outcome you choose, not just clicks. Tell the engine to weight margin, lifetime value or retention, and business rules cap it, inventory, compliance, margin floors. The smartest recommendation is filtered to the one that's also right for the business to make.
How does the DPDP Act affect a decision engine?
A decision engine acts on personal data, so every decision must run within verifiable, purpose-linked consent, under an Act with penalties up to ₹250 crore. We build consent into the decision: only consented data feeds the models, with audit trails on what fired and why. Defensible by design.
How long until the engine beats our current rules?
We run the engine alongside your current rules from week one, measuring lift head-to-head. Most see ML decisions beating rules on priority journeys within the first quarter, then widen the gap as models retrain on live outcomes. You shift weight to ML on evidence, not faith.
How do you prove a decision actually works?
Every decision is tested live, which model, rule or recommendation fires, measured per cohort against a control. This is decision-level experimentation: distinct from UX experimentation, which proves which design wins. One proves the screen; this proves the logic behind what's on it
Stop Guessing What to Show. Let the Engine Decide. Product, content and offer recommendations decided by data, tied to revenue, on your stack