Decisions ride on before-and-after
Seasonality and audience mix get mistaken for impact.
Platform Foundation Experimentation
AI agents turn goals into experiments, run them as adaptive tests that learn while they run, and keep a fixed control so the result is honest. If it can be measured, it can be improved.
Traffic share · day 7
Why change
Most marketing decisions are still made on correlation. The teams that do test properly run a handful a quarter, because every test is hand-built.
Seasonality and audience mix get mistaken for impact.
Ideation, setup and analysis are all manual.
Overlapping exposures and broken assignment contaminate the result.
The learning lives in a slide and the rollout never happens.
What it does
Language models plan, explain and generate. Statistical learning chooses and allocates. A control group keeps everyone honest.
01 Adaptive learning
A fixed A/B/n test gives every variant an equal share until it ends. An adaptive policy shifts share as evidence builds, keeps exploring in case it is wrong, and records the exact probability behind every choice.
Traffic share by creative
Fixed A/B/n test
25% each for 14 days, then pick a winner.Adaptive agents
Shift share as evidence builds. Keep exploring.Illustrative allocation, not a prediction. Actual shares depend on evidence, uncertainty and guardrails.
02 The right action for each context
A simple test might crown one winner. A contextual policy learns that the discount wins with leisure travellers, lounge access with business travellers and extra baggage with families, and serves each accordingly.
| Customer context | 10% discount | Extra baggage | Lounge access |
|---|---|---|---|
| Leisure travellers | 55% | 30% | 15% |
| Business travellers | 10% | 25% | 65% |
| Family travellers | 20% | 65% | 15% |
Illustrative learned preferences for a fictional airline, not conversion rates.
03 Trustworthy measurement
Every policy runs against a randomised holdout. Results come with intervals, data-quality checks and a clear recommendation: ship, iterate or stop. For spend decisions, experiments calibrate attribution and mix modelling.
Agents at work
Planner, creative, optimisation, measurement, learning and governance agents each own a step. They are logical roles, not six chatbots: the choosing is done by statistics, the checking by policy, and the final call on anything material by a person.
Assignment, allocation within exploration limits, quality checks, stopping a variant that breaches a guardrail.
Launching a new test, changing the reward metric, full rollout of a winner and budget moves above threshold.
Tests without a control, customers outside consent, and any variant that breaks brand, legal or frequency rules.
What's inside
Across web, app, messaging, recommendations, loyalty and media, on one set of metrics.
Hypotheses, rationale and prioritisation in one backlog.
Primary, secondary and guardrail metrics; power and duration.
Client-side and server-side, with feature flags.
Adaptive allocation with logged selection probabilities.
For sequential decisions, used where the complexity is earned.
Consistent bucketing, targeting and mutual exclusion groups.
Sample-ratio mismatch, event checks and collision detection.
Frequentist or Bayesian, sequential testing, variance reduction.
Short-run controls and long-run holdouts for novelty and retention.
Geo tests, response curves and budget scenarios.
Gradual release, approvals and one-click rollback.
Every result, searchable, linked to its assets and data.
How it runs
Each step is a named, logged stage. Nothing autonomous is switched on until its data, permissions, monitoring and rollback have been checked.
How you'll measure it
We agree the numerator, denominator, cohort and success gate with you before anything goes live, and compare against a baseline or a holdout, never against a hunch.
Questions
Not quite. A bandit minimises the cost of learning by sending less traffic to losers while the test runs. It does not, by itself, prove causal lift, which is why every adaptive policy here also runs against a randomised holdout.
No. Language models plan experiments, generate variants and explain results. Allocation and policy updates are made by statistical learning algorithms on logged reward data.
When the decision is sequential and the reward is long-term, such as a multi-step lifecycle journey. Most programmes should start with bandits, move to contextual bandits and add full RL selectively.
Yes, with realistic expectations. Power planning tells you up front what size of effect you can detect, and adaptive methods make better use of limited traffic than fixed splits.
Composable by design
Experimentation is the proof layer. Each capability it connects to gets a fair test and a growing evidence library.
Compare decision policies with stable allocation and measured lift.
ExploreCreative and budget tests designed properly, with incrementality.
ExploreFind out whether a reward changed behaviour or just cost margin.
ExploreRuns on your web, app, messaging and ad channels through SDKs and APIs. Reads outcomes from your analytics, warehouse and order systems.
Small language models trained on marketing data are deployed inside your cloud or data centre, so customer data and the models reasoning over it stay under your control.
Getting started
Pick a question with real money attached. Agree the reward and the guardrails, run it properly once, and let the result settle the argument.