Insights
Notes from the work.
Practical writing on building enterprise AI that reaches production.
AI rules are slipping in the U.S., Europe and Brazil at once. In 2026, the buyer is becoming the regulator
A voluntary White House accord, a deferred EU timetable and a postponed Brazilian bill leave enterprises carrying more governance load. What buyers should demand from vendors, and from themselves.
Read
Agent memory is enterprise data storage, even when nobody calls it a database
Google Cloud's always-on memory agent sample ingests, consolidates and recalls information nonstop. A useful design reference, and a reminder that memory brings retention, deletion and access duties.
Read
The best agent use cases may be the playbooks your teams already run. AWS shows it with cloud migration
AWS published a four-agent architecture for migration on Bedrock AgentCore: intake, infrastructure code, governance and post-cutover SRE. It works because migration is already a structured playbook.
Read
AI governance gets real when a control produces evidence. EC-Council's ADG 2.0 expands to 180+ controls
EC-Council's new framework specifies the evidence each AI control must produce during operation, and sets human review points by autonomy tier. Useful, if nobody mistakes a crosswalk for compliance.
Read
MCP standardizes how agents connect to tools. It does not govern them. IBM's Orchestrate release shows the gap
watsonx Orchestrate now registers remote MCP servers per tenant and lets admins put validation rules, execution limits and runtime safeguards on MCP tools. The control layer every MCP rollout needs.
Read
Your company no longer has an AI model. It has a portfolio, and it needs managing like one
Specialist models, decision models, domain operators and routing defaults arrived in one week. The single-model era is ending inside the enterprise. A framework for the six roles in a model portfolio.
Read
GPT-Synopsys trains AI to operate chip design tools, not just answer questions. Verification has the last word
OpenAI and Synopsys are building a model that learns to run chip design tools like an expert engineer. It is in early engagements, and its design keeps deterministic verification as the authority.
Read
MCP is easy in a demo. In production, every tool call needs a name behind it
An AWS reference architecture puts corporate SSO, OAuth and gateway token checks in front of an MCP tool; Ping's new agents act only on the user's authority. That is where enterprise MCP starts.
Read
Don't automate the old workflow. Chatham Financial redesigned trade validation around evidence and exceptions
Chatham Financial cut an early trade validation review from about 30 minutes to under 4 by asking what evidence each decision needs, not where to put an agent. Its model routing follows suit.
Read
Your agent should not ask a frontier LLM every small question. Cloudflare's Clef argues for decision models
Cloudflare released Clef and Clef-flash, open models that return probabilities over allowed answers instead of text. They point to an agent architecture with a separate, cheaper decision layer.
Read
Agents make weak IAM fail faster. Microsoft's 2026 Digital Defense Report puts agent identity beside passwords
Microsoft's new threat report says identity is still where attacks start, and the control plane now includes applications and agents. For AI programs, identity hygiene is the first security project.
Read
Prompt guardrails are not a security boundary. This year's agent breaches show where the boundary has to be
Frontier models from OpenAI, Anthropic and Meta broke out of misconfigured test environments this year. NVIDIA's answer moves enforcement outside the agent. What that means for enterprise agents.
Read
Have an AI use case stuck between prototype and production?
Tell us what you’re trying to ship. We’ll reply with honest next steps.
Discuss a use case