A meaningful change is captured
BrandOS records a material change such as a price update, budget move, campaign restructure, creative change or promotion.
Decision Memory
BrandOS is NOT another LLM wrapper.
BrandOS is not a chatbot sitting on top of Shopify and Meta. It builds a structured memory of the important decisions made across a brand — what changed, who or what changed it, why it changed when that is known, and what happened afterwards.
The simplest way to think about BrandOS
Analytics tells you what is happening. BrandOS remembers the decisions that changed what happened next.
Why this matters
Brands keep relearning the same lessons.
E-commerce brands make dozens of important changes every month. Prices move. Budgets are raised. Creatives are switched. Promotions launch.
Campaigns are restructured. Some decisions are made by people, some by platforms, and increasingly some by AI agents and LLM-powered systems.
The problem is that the learning behind those decisions gets fragmented across people, agencies, tools and platforms. The reason might sit in Slack, the action in Meta or Shopify, the AI recommendation in another system, and the result in a dashboard weeks later. By the time a similar decision comes around, the brand is often piecing the story back together, or relearning the same lesson again.
BrandOS keeps that decision history with the brand, regardless of whether the decision came from a person, a platform or AI.
From change to memory
What actually happens when something changes
BrandOS does more than pull data into a prompt. It follows a decision through a sequence that can be checked later.
The context at that moment is frozen
The system keeps the relevant commercial and marketing evidence as it looked when the decision happened, before the result was known.
Who or what made the change is recorded
Where the platform can verify it, BrandOS records whether the change came from the brand, an agency, platform automation or an AI system.
The reason is kept separate from the facts
If the person or agent supplied a reason, BrandOS stores it. If no reason was recorded, BrandOS can suggest a likely reason, but it is clearly labelled as an inference rather than a fact.
BrandOS waits for the right review window
Different decisions need different amounts of time. BrandOS does not immediately declare a result just because a metric moved the next day.
The outcome is checked against real data
When enough time has passed, BrandOS checks what happened and whether another event, such as a promotion, stock issue or restructure, made the result difficult to interpret.
The decision becomes part of the brand's memory
A settled case can then be brought back when a similar decision appears in the future.
Where AI fits
AI helps explain the memory. BrandOS turns what actually happened into knowledge the brand can trust and reuse.
BrandOS is not just an LLM connected to your data.
An LLM can analyse the information it is given. BrandOS creates a decision record as things happen — what changed, why, who changed it, what was expected, and what happened next.
It then keeps that history and brings it back when a similar decision comes around.
AI helps explain and retrieve the memory. BrandOS creates the memory itself.
AI helps explain the memory. BrandOS creates the memory by turning every important decision into reusable knowledge, so the next decision starts with what the brand has already learned.
A different foundation
An LLM can analyse a result. BrandOS knows what the brand was trying to do before the result existed.
An LLM can analyse the data your business already has. BrandOS creates a new kind of data around the decisions themselves, what was changed, who changed it, what was known at the time, what was expected, and what happened afterwards.
That decision history does not exist in Shopify, Meta or a generic AI model. BrandOS builds it prospectively, then turns each settled decision into evidence the brand can use the next time a similar decision has to be made.
An LLM connected to your data
BrandOS
Can analyse the history it is given.
Builds a persistent record of decisions as they happen.
Usually looks backwards from the result.
Freezes relevant evidence before the outcome is known.
Can generate a plausible explanation.
Separates recorded rationale, verified facts and inferred rationale.
Does not inherently know who changed a platform setting.
Preserves actor and execution provenance where it can be verified.
Can tell you a metric moved after a change.
Waits for the review window and checks whether the outcome is usable.
May answer from whatever context is currently available.
Keeps a structured, reusable Decision Memory for the brand.
Can suggest what might happen next.
Can later compare a new decision with what actually happened in similar past cases.
A simple example
A brand changes a product from £39 to £44.
- A generic AI assistant can look at the store later and tell you that price went up and sales changed.
- BrandOS records the change when it happens, keeps the commercial context from that point in time, records who or what made the change, and keeps the stated reason if one was supplied.
- It then waits for the appropriate review window rather than judging the decision immediately.
- When the window closes, BrandOS checks what happened and whether anything else made the result unreliable.
- The next time a similar price decision appears, that previous case can be retrieved as evidence rather than rediscovered from scratch.
That is the difference between analysing data and building memory from decisions.
What BrandOS remembers
The material decisions that can genuinely alter performance.
BrandOS is focused on material decisions, the changes that can genuinely alter how the business performs.
Commerce
- Prices
- Promotions & discounts
- Inventory decisions
Paid media
- Budgets
- Pause / revive
- Campaign structure
- Creative changes
Who made the change
- Brand or team
- Agency
- Platform automation
- AI agent
Evidence quality
When BrandOS does not know
A useful memory system also needs to know when not to turn a messy result into a lesson.
If a promotion overlaps a price test, inventory disappears halfway through a review window, tracking breaks, or a campaign is restructured before the outcome can be read properly, BrandOS can mark the result as insufficient or confounded instead of forcing a neat answer.
That matters because a confident explanation is not useful if the evidence underneath it is weak.
Commerce is automating
Why this becomes more important as commerce automates
As e-commerce becomes more automated, more decisions are made across more systems, and the brand risks losing the complete story of what changed, why it changed and what happened next.
Meta can optimise a campaign. An AI agent can move budget. A merchandising system can change a price or promotion. Each decision may make sense in isolation, while no single system understands how those actions interact across the wider business.
BrandOS becomes the decision memory layer across that automated stack.
It records important decisions, preserves the reasoning and evidence behind them, and connects each action to its eventual outcome, whether the decision was made by a founder, an agency, a platform or an AI agent.
As automation increases, BrandOS ensures the intelligence created by every decision stays with the brand.
The idea in one sentence
The idea in one sentence
BrandOS turns real decisions and their outcomes into memory the brand can use the next time a similar decision has to be made.