Tag: AI Agents

I often find myself grappling with the essential need to make sure AI-generated work aligns with our brand’s identity and campaign history. This is where a structured ‘client brain’ comes in handy, providing context that grounds AI in brand guidelines and technical nuances.

Each SEO agency I know deals with what I like to call a ‘context tax.’ It’s the unspoken burden when strategists and analysts have to remember intricate account details like brand voice, past keyword decisions, CMS limitations, and what angles the founder disliked.

The challenge with integrating AI into complex SEO tasks is giving it enough context to be genuinely useful. One approach I’ve been exploring is the concept of a ‘client brain,’ a system that retains account-specific knowledge, allowing AI to act with the intelligence of someone who’s been involved from the start.

Context is the Problem

Understanding context is crucial for any successful worker or AI. As a senior SEO lead, I onboard new team members by sharing vital details about client preferences, past strategies, and technical constraints. Without this, even AI struggles to deliver effectively.

In SEO, we’re increasingly focused on data integration—bringing together metrics from various sources to have a comprehensive view. However, AI isn’t just about data. It’s about using that data in the context of what we’ve learned from the client so far.

I’ve realized that having a centralized repository of client-specific insights—what I consider ‘institutional memory’—is indispensable. It helps our AI avoid suggesting ideas or strategies we’ve previously dismissed or accepted.

A Client Brain is the Solution

Creating a ‘client brain’ means systematically recording all critical decisions, feedback, and client idiosyncrasies. It’s not a substitute for human intuition but rather a framework that aids in applying that intuition consistently across teams and tasks.

In my experience, effective SEO requires multiple hands—strategists set the path, content leaders draft plans, writers create, and analysts evaluate performance. Without shared context, each transition becomes a potential point of drift, where critical details can be lost.

What a Client Brain Is

A client brain is essentially a detailed, organized knowledge base that AI refers to before launching into any task. I like to think of it as carving out the soul and memory of an account, enabling AI to understand both stable brand tenets and evolving client experiences.

Not all knowledge is created equal. Some parts remain constant, like brand identity and audience details, while others evolve, such as campaign outcomes or technical limitations. It’s vital to separate these layers to ensure clarity and usability.

In practical terms, I categorize this into two layers: the ‘soul,’ which includes static brand knowledge, and the ‘memory,’ which documents dynamic experiences and learning moments from working with clients.

The Technical Anatomy of a Brain

I aim for simplicity when building a client brain. It starts as a collection of plain-text Markdown files, eliminating the need for complex systems or special software. The organization into a ‘soul’ and ‘memory’ folder structure helps keep everything manageable.

Building Core Logic of the Soul

To initiate this process, I recommend creating a directory named ‘brain/soul’ and populating it with core files: company profile, style guide, audience, keyword map, and a ‘never do’ list. Each serves to capture essential client insights succinctly.

brain/soul/
├── company-profile.md
├── style-guide.md
├── audience.md
├── keyword-map.md
└── never-do.md

Memory Captures Decisions, Patterns, and Logs

Within the ‘brain/memory’ folder, I record decisions we made and their rationales, identify recurring patterns in our work, and maintain a chronological log of notes and client interactions. It’s invaluable for maintaining institutional knowledge over time.

brain/memory/
├── decisions/    — choices made and why
├── patterns/     — things that worked or didn’t, by task type
└── log/          — chronological notes by date

Documenting the logic behind decisions is as critical as the decisions themselves. It ensures AI can align with evolving strategies, adapting as contexts change over time.

Building the Brain Step-by-Step

Step 1: Pick the Right Starting Client

It’s crucial to start small, choosing the client where losing context costs the most time. Long-running accounts with a distinct brand voice and a history of ideas are ideal candidates.

Step 2: Block 90 Minutes and Write the Soul Together

I gather the account lead and strategist, focusing on crafting five vital files in straightforward language. It’s about capturing the unspoken knowledge that guides our best decisions.

Step 3: Decide Where the Brain Lives

For solo practitioners, a local folder suffices. Teams, however, benefit from a shared location. Options include Google Drive, Notion, or even a version-controlled system, as long as it serves as a trusted central repository.

Step 4: Set Ownership Rules

I find friction helpful for changes to the soul. The account lead reviews changes, ensuring consistency. Memory changes should be easy for anyone to add, thereby capturing fresh insights on-the-go.

Step 5: Schedule Maintenance

Regular brain maintenance is crucial to prevent rot. Tasks include consolidating duplicates, updating entries, and resolving conflicts. A stale client brain can create more harm than good if left unchecked.

How AI Agents Read the Brain

In practical use, AI tools benefit from having access to the client brain, which helps maintain consistency in outputs and aids collaboration. Whether through comprehensive or selective file loading, the integration should preserve context and avoid unnecessary readjustments.

Version A: Load Everything

A straightforward method is to have AI read all files in the brain folder before starting a task. It incurs some cost, but is often more efficient than repeatedly reexplaining account details.

Version B: Route by Task Type

Selective loading simplifies tasks by having AI access only the necessary files based on the specific task type. It’s a balanced approach many agencies are adopting to optimize efficiency and relevance.

Version C: Vector Retrieval

For agencies managing numerous clients, vector retrieval provides a sophisticated solution. It involves using metadata tags for entries, allowing AI to fetch relevant content effectively while ensuring accuracy and specificity.

Using the Brain Across Different AI Platforms

I ensure the consistency of AI outputs by integrating the client brain into various AI workflows, be it Claude Code or Cowork. The emphasis remains on the AI engaging the soul files at the start of tasks, securing alignment and coherence.

Where This Breaks and How to Fix It

Even a well-maintained client brain can encounter issues like drift or fabrication. Remedy these by ensuring the style guide includes clear examples and that memory entries are frequently reviewed for accuracy and relevancy.

Trust in the client brain derives not from its structure but from its content. A reliable source underpins every memory entry, bolstering confidence and effectiveness.

How to Get Started This Week

To implement this system efficiently, I’ve found it useful to start with a single client, gather the team for a focused session, and conduct a test. This initial phase acts as a proof of concept, validating the utility of a client brain in enhancing SEO tasks.

The real payoff from AI doesn’t come from speed alone but from the context it provides. With a client brain, the gaps usually lost in transition are preserved, ensuring the work is not just faster, but smarter.

Inspired by this post on Search Engine Land.