AI news for business owners: September 2026
Meta Muse, Instinct, custom GPTs retiring, Cowork merging into Claude, GPT-6 and Opus 5.5, and what each means for a small team.
Learning center.
Plain answers to the questions teams ask before they put AI into real work: what an MCP server is, what an agent is, how to roll Claude out, what AI consulting costs, and how AI fits commercial real estate. Each page answers the question first and mentions us last, if at all.
What changed this month
Past months, April to August 2026, are in the AI news archive.
The AI in real estate field guide
A four-part guide to moving AI beyond chat and into the files, systems, and recurring work that commercial real estate teams rely on. Each part stands alone, and the full guide is a free 18-page PDF.
Download the field guide
Claude in CRE: Skills
Turn the underwriting steps your team repeats into reusable instructions Claude can follow consistently.
Read the field guide
Claude in CRE: Connectors
Connect Claude to the CRM, property systems, and local market data where the real work lives.
Read the field guide
Claude in CRE: Cowork
Let Claude work directly inside local files so underwriting models stop bouncing between tabs and downloads.
Read the field guide
Claude in CRE: Scheduled Tasks
Build recurring market and pipeline routines that run whether or not you are sitting at your computer.
Read the field guideCommercial real estate
How commercial real estate brokers use AI
Prospecting from public record, BOVs, comps, CRM hygiene, and market updates, with what not to hand to a model.
MCP server for commercial real estate
What it does for a brokerage, what data can sit behind it, and the quiet failures to design against.
What is a broker opinion of value (BOV)?
What goes in one, how it differs from an appraisal, and what Florida law says it may not be called.
What Florida property appraiser data tells you
Just value, DOR use codes, qualified sales, $100 deeds, portfolio sales, and what the roll never carries.
AI lease and rent roll abstraction
The fields to pull, how to check them, and where a person still has to read the lease.
AI in CRE underwriting: what it can and can't do
Normalizing T-12s and rent rolls, drafting memos, and why the value conclusion stays with the analyst.
AI tools for commercial real estate
A map by job to be done: assistants, data platforms, abstraction, underwriting, CRM, and public record.
Can Claude connect to CoStar?
There is no official connector, and CoStar's terms restrict AI access. What to automate instead.
Claude in CRE: Skills
Package the underwriting steps your team repeats into a playbook Claude follows.
Claude in CRE: Connectors
Connect Claude to the CRM, property systems, and local market data.
Claude in CRE: Cowork
Work directly inside approved local underwriting files.
Claude in CRE: Scheduled Tasks
Recurring market, pipeline, and property routines.
Claude, MCP and agents
What is an MCP server?
The Model Context Protocol for a business reader and a technical one: tools, resources, auth, and the security boundary.
What is an AI agent?
Agent vs chatbot vs automation, the loop an agent runs, and where a person has to stay in it.
Personal AI agents like Meta Muse and Instinct
What consumer agents that act across your email, calendar and browser mean for a business, and the policy to set.
Claude Skills vs MCP
Skills package how Claude does a task. MCP is how it reaches a live system.
What is Claude Cowork?
Claude doing multi-step work in your files and tools, and what changed when Cowork began merging into Claude in September 2026.
How to write a SKILL.md that gets used
A sharp description, a lean body, and reference files loaded on demand.
MCP server design best practices
Tool boundaries, read vs write, schemas as contracts, approval and audit.
MCP vs function calling
How a model calls tools, what MCP adds, and where agents fit.
How RAG works: retrieval and grounded answers
Chunking, embedding, ranking, and keeping answers tied to sources.
RAG evaluation for internal knowledge bases
Measure retrieval quality and grounding before trusting it.
Context engineering for large codebases
CLAUDE.md, AGENTS.md, retrieval, and definitions of done as infrastructure.
AI code review with judge agents
A judge-model loop that reviews pull requests without replacing the reviewer.
Rolling AI out to a team
How to roll out Claude to a team
One workflow first, then Projects, Skills, connectors, a data policy, and training.
AI training for employees: what sticks
Train on the team's own tasks, not a feature tour, and measure what people keep using.
Claude prompts for small business owners
Copy-paste prompts from our free workshops, and how to turn a good one into a Skill.
AI governance for small and mid-size businesses
A two-page AI policy, including a rule for personal AI agents like Meta Muse.
How to run an AI workflow audit yourself
Map inputs, decisions, tools, outputs, and failure modes first.
The three phases of AI deployment
Audit, pilot, controlled rollout.
Buying AI services
How much does AI consulting cost?
The pricing models in the market, what drives the number, and the costs outside the consultant.
How to choose an AI consultant
Questions to ask, what good deliverables look like, and the red flags.
Claude vs ChatGPT for business teams
A dated comparison of the things a team actually decides on.
Zapier vs AI agents
Deterministic automation vs judgment steps, and when to combine them.
Have a question that is not answered here?
Bring the workflow you are thinking about. The first call is free and we will tell you whether AI is the right tool for it.
Book a call