AI
29 posts in this category.
-
The Carcinization of B2B Software: Why Platforms Keep Growing the Same Five Organs
B2B software is converging on one anatomy - MCP, a knowledge graph, workflows, agents and skills. Why it happens and why competition moves under the shell.
-
7 Things You Can Do With Improvado MCP in Claude (With the Prompts)
Seven real workflows for Improvado MCP in Claude Desktop, Cowork and Claude Code, each with the exact prompt, what Claude does, and the result it produced.
-
GPT-6 Astra vs Claude Fable 5.1 (and Mythos 5.1): Which AI Should You Actually Use, What It Costs, and What Happens Next
GPT-6 Astra vs Claude Fable 5.1 and Mythos 5.1: who wins on writing, coding, research and daily tasks, what each costs, the safety drama, what comes next.
-
Kimi K3 vs GLM-5.2 vs DeepSeek V4 Pro: Open-Weight AI on Your Own Marketing Data
Three open-weight models compared on price, license, context, tool calling and data terms, and what it takes to run one of them on your own marketing data.
.jpeg)
-
Google Ads AI Max: What the September Upgrade Changes for Reporting
Google auto-upgrades broad match and ACA campaigns to AI Max in September 2026. The reporting fields that change, and the landing-page join that breaks.
-
Why Part of Our Agent Knowledge Base Runs on Subversion
A file your team reads as documentation and your agents execute as code needs a home that fits both. We moved part of our knowledge base to Subversion for two specific reasons, and kept a Git mirror running the whole time.
-
ChatGPT Ads Measurement: What You Can and Cannot Track
A line-by-line reference on ChatGPT Ads measurement, checked against OpenAI's own help documentation. What Ads Manager reports today, what does not exist yet, and the GA4 tagging mistake that makes paid clicks look organic.
-
The Campaign Launcher Skill for Claude, Ungated
In March I asked people to comment LAUNCHER to get our Campaign Launcher skill for Claude. The thread ran to 743 comments. Here is everything that was behind that gate, linked directly, no comment required.
-
How Long Does Improvado Actually Take to Set Up?
A straight answer on Improvado setup time and time to value, sourced from the actual connection flow, docs, and published customer outcomes, not a demo script.
-
Build vs Buy Data Integration in the Agent Era: It Is a Custody Decision Now
Build vs buy used to be a cost decision. In the agent era it is a custody decision, and the boring layer worth buying now carries three jobs, holding the keys, giving agents MCP access, and keeping analytics consistent on a knowledge graph.
-
The AI Trust Layer Problem: Confidence Is Not Correctness
A trust layer on top of a broken data foundation does not create trust. It laminates the disagreement. Three questions expose any setup, and none of them are about the interface.
-
Every Connector Is a Switch: OAuth Revocation and Your Marketing Stack
When a third-party integration is compromised badly enough, every token it holds gets revoked at once. That second blast takes your syncs, pipelines and reporting down with it. Here is who owns that switch.
-
Context Engineering vs Prompt Engineering: What Your AI Agent Actually Sees
Prompt engineering did not die. It turned into context engineering, and that moved the work from wording to what your agent knows about your business on every call. Here is who owns that and how to audit it.
-
The Kill Radius: What AI Platforms Absorb and What They Make More Valuable
When Claude got its own Slack account, thin layers went on alert. The platform absorbs thin layers and makes everything it cannot reach more valuable. Here is how to tell which side you are on.
-
AI Agent Spend Controls: Why a Prompt Is Not a Guardrail
Our agents spent $1,768 in one night after a prompt told them not to. The failure chain, and the harness-level controls that actually cap what an agent can spend.
-
AI Agent Traffic: The Buyers Your Analytics Cannot See
94% of business buyers used AI in their most recent purchase, and their agents leave no UTMs, no cookies, and no referrers. How to see, segment, and report AI agent traffic.
-
The Hidden Costs of AI Agents: You Pay for Statelessness Three Times
A stateless AI agent gets paid for three times, in tokens, in team hours, and in repeated mistakes. Why organizational memory is the missing cost structure fix.
-
AI Agent Orchestration: A Complete Guide for Marketing Ops Leaders (2026)
Master AI agent orchestration for marketing operations. Learn proven frameworks for coordinating autonomous agents across data pipelines, analytics, and campaign execution.
-
AI Accountability: Who Owns the Miss When the CFO Reads the Wrong Number?
When an AI dashboard hands the CFO a confidently wrong number, who owns the miss? A practical AI accountability framework, RACI matrix, and governance architecture for finance-grade AI reporting.
-
AI Compliance and Corporate Governance: Who Owns the Call When a Model Votes on the Board?
AI compliance is no longer about whether to use AI in regulated decisions, it's about who owns the call when the model is in the chain. Two scenarios, four frameworks, one diagnostic.
-
SaaS Is Not Dead: Why Per-Seat SaaS Pricing Models Are Being Replaced by Outcomes
SaaS isn't dead. Per-seat pricing is. Here's why SaaS pricing models are shifting from headcount to outcomes, and how to evaluate any tool's pricing in 2026.
-
AI Job Loss in 2026: What the Historical Record Actually Says About AI and Employment
AI job loss is real in the short run: NBER projects 502,000 cuts in 2026. But every prior tech revolution ended with more work, higher wages, and shorter hours. Here's the long-run shape.
-
AI Agent Memory Is Not Search: Why Your LLM Context Window Is an Archive, Not a Brain
AI agent memory and search are not the same thing. Here's why a bigger context window doesn't give your agent a brain — and what the missing layer actually looks like.
-
Secure Agents or Shadow Ones: How Prompt Injection, Shared Keys, and Tool Sprawl Quietly Compromise Your Stack
Prompt injection, shared credentials, and tool sprawl turn shadow AI agents into data leaks. Here's the architectural pattern that makes the dangerous things structurally impossible.
-
What Actually Works When You're Running 70+ Autonomous Agents
Most AI agent governance frameworks are committee-shaped. The marketing failure mode is architectural. Here's the 4-dimension runtime pattern that holds at 70+ agents.
-
Marketing Data Integrations Are the Real AI Moat: Why 1000+ Connectors Beat a Smarter Model
Everyone's racing to build smarter agents. We spent 8 years building boring pipes. Here's why marketing data integrations are the unsexy infrastructure that decides whether your AI stack survives the next API change.
-
Why Pharma Marketing AI Needs an Industry-Specific Knowledge Graph (And Why Generic Agents Are a Compliance Liability)
A generic AI marketing agent in pharma is a lawsuit waiting to happen. Here's why regulated industries need an industry-specific knowledge graph — and what one actually encodes.
-
Conversational UI vs Buttons: AI Agents Don't Kill the Click. They Compress It.
Bret Taylor said AI agents will kill button-clicking. The more accurate read: agents absorb navigation clicks and compress UI down to the few clicks that actually own a decision.
-
The AI-Native Marketing Team Is a Hiring Trap
Every CEO wants an AI-native marketing team. Here's why the best operators of AI agents inside our customers are ten-year veterans, not AI-native juniors — and what to hire for instead.