Your app asks. EkamFlow answers.

Batch scores answer yesterday's question. A decision point answers the one in front of you: a customer is on the cancel page or has left a cart, your application asks what to do for them, and it gets an action back chosen from that customer's predictions and context.

What you can do

A moment, and the actions you allow

You name the moment — abandoned cart, cancel page, inbound call — the actions you are willing to take there, and what each is worth. EkamFlow never invents an action you did not declare.

Chosen by what it is worth

Each allowed action is scored from this customer's own predicted chance of taking it, what it would be worth, and what it costs. The best-scoring action wins. Where your rules exclude every action, the point answers with the fallback you set rather than inventing one.

Your rules are applied, not suggested

Conditions, exclusions, a cooldown per action and a contact cap per customer are enforced on EkamFlow's side before anything is chosen, and the answer says which actions were left out and why.

Your systems call it

An API key scoped to decisions and nothing else. Your server asks; your own application decides what to do with the answer. Nothing is sent to your customer by EkamFlow.

A holdout that keeps you honest

A share you choose is held back from the decision, so you can see whether deciding this way actually beat not doing it, rather than assuming it did.

It can learn from what it tried

Optionally, a small share of calls picks at random among the actions still eligible, so later models have fair comparisons to learn from rather than only the choices you already favoured.

Idle costs nothing much

Context is refreshed only for decision points that have actually been called recently. A moment you stop asking about stops costing anything beyond its storage.

Setup and requirements

Steps

  1. An admin approves storing the context a decision point reads.
  2. Define the moment, the actions allowed there, and what each is worth.
  3. Create an API key scoped to decisions and call it from your server.
  4. Read the results, including how the held-back share compared.

What you need

  • A profile for the customers the moment is about
  • An admin's approval to store decision context, which can be withdrawn
  • A server of your own to make the call and act on the answer

An admin's approval to store the context a decision point reads is required — withdrawing it clears what those points kept.

Build an Agentic Worker for your team.

Tell us which workflow you want to hand off.

Book a demo