AI Decisioning

AI Decisioning is learning-based marketing personalisation where AI agents decide which experience to deliver to each customer, optimising toward defined business goals within explicit guardrails.

Key points

• Hightouch defines AI Decisioning as a new class of marketing technology for 1:1 personalisation, replacing static segments and journeys with agents that continuously learn which experiences drive the best outcomes for each customer ^[src-023].
• The marketer's role shifts from drawing every journey branch to defining goals, decision dimensions, inputs, and guardrails ^[src-023].
• Decision dimensions can include subject line, content modules, channel, send time, day of week, communication frequency, and other experience variables ^[src-023].
• The technical foundation combines reinforcement learning, Multi Armed Bandits, and Contextual Bandits. Reinforcement learning connects actions to outcomes, bandits balance exploration and exploitation, and contextual bandits incorporate customer data into the choice ^[src-023].
• The key operating claim is that "right" content/channel/timing cannot be known in advance; it must be learned from customer engagement and non-engagement over time ^[src-023].
• Hightouch's second article expands the reinforcement-learning loop: an agent selects an action, interacts with the customer environment, receives a positive or negative response, then chooses the next action based on experience and desired outcomes ^[src-024].
• In marketing terms, actions can include message/no message, channel, timing, cross-sell offer, message type, and frequency; rewards can include opens, visits, cart additions, purchases, retention, or lifetime value ^[src-024].
• The third article identifies multi-armed bandits as the automated experimentation engine inside AI Decisioning: they shift resources to winning actions while still exploring new possibilities across marketing dimensions ^[src-025].
• The fourth article positions contextual bandits as the final 1:1 personalisation layer: they combine bandit exploration/exploitation with individual customer data so AI Decisioning can choose what works for this person, right now ^[src-026].
• Hightouch says its production version uses multiple contextual bandits optimised for different business outcomes and coordinates messages across campaigns ^[src-026].
• Braze uses the phrase AI Decisioning Studio for a broader optimisation architecture where Intelligent Selection and multi-armed bandit logic handle live allocation, while contextual bandits and predictive scoring support personalisation and next-best-action decisions ^[src-027].
• The Braze article also makes the reward loop explicit for campaign decisioning: customer interactions such as clicks or purchases update confidence across variations and allow the system to adapt quickly ^[src-027].
• Braze's later whitepaper frames AI decisioning as the answer to the personalisation bottleneck that remains after generative AI helps with data and content creation: deciding which action each individual customer should receive ^[src-152].
• In Braze's operating model, marketers define a success metric, decision dimensions, and available options; the AI decisioning agent then learns across dimensions such as offer, subject line, creative, channel, day, time, and frequency ^[src-152].
• Braze's Community Of Bandits architecture partitions the action space by dimension so collaborating contextual bandits can learn faster than a single monolithic model over every possible campaign combination ^[src-152].
• Braze reports a financial-services case study where its community-of-bandits approach increased conversion rate by 92% versus business as usual and produced $16M in annualised value; treat this as a vendor-reported claim unless independently corroborated ^[src-152].

Related entities

• Hightouch
• Braze

Related concepts

• Personalisation Gap
• Treatment Personalisation
• Contextual Bandits
• Multi Armed Bandits
• Exploration Exploitation Tradeoff
• Reinforcement Learning For Marketing
• Agentic Marketing
• Marketing Bandit Optimisation
• Customer Feature Matrix
• Intelligent Selection
• Community Of Bandits

Source references

• ^[src-023] Hightouch — "Under the hood of AI Decisioning, part one: Overcoming the personalization gap"
• ^[src-024] Hightouch — "Under the hood of AI Decisioning, part two: Reinforcement learning"
• ^[src-025] Hightouch — "Under the hood of AI Decisioning, part three: Multi-armed bandits"
• ^[src-026] Hightouch — "Under the hood of AI Decisioning, part four: Contextual bandits"
• ^[src-027] Team Braze — "What is a multi-armed bandit? Smarter experimentation for real-time marketing"
• ^[src-152] George Khachatryan, Nathaniel Rounds, Victor Kostyuk / Braze – "A community of bandits" (source page: "From multivariate testing to AI decisioning", 2025-09-30)