AI for small business, in plain terms

What AI agents can and cannot do for a small or mid-sized business, how to keep your data and your approvals, and what we learn doing the work.

The Future of Work in a Small Business: People Decide, Agents Do the Steps
Strategy
The Future of Work in a Small Business: People Decide, Agents Do the Steps
In a small business, the future of work is not whole jobs disappearing. It is people deciding and approving while agents do the repeatable steps, with each job redesigned task by task and each agent owned by a named person. Here is what that looks like in a twelve-person company, and where to start.
Taylor Gendron · October 27, 2026 · 6 min read
Make Your Company Queryable: Answer the Exact Question, for People and AI Assistants
Guide
Make Your Company Queryable: Answer the Exact Question, for People and AI Assistants
Customers now ask exact questions, and more of them ask an AI assistant first. Your products, prices, stock, hours, policies and service area should answer those questions directly, in a form a person or a machine can read. Here is what that takes.
Taylor Gendron · October 27, 2026 · 6 min read
Make Your Company Agentic: From Tools You Operate to Outcomes You Approve
Strategy
Make Your Company Agentic: From Tools You Operate to Outcomes You Approve
Once your facts answer exact questions, software can act on them: follow up a quote, draft a review reply, reorder stock, send the Monday report. What changes is who does the work and who decides. Start with one channel, and keep approvals and a log.
Taylor Gendron · October 27, 2026 · 5 min read
Work Backwards From Intention: Start With the Customer's Exact Question
Strategy
Work Backwards From Intention: Start With the Customer's Exact Question
Most websites, offers and sales flows are built forwards from the org chart or the product list. Customers arrive with something to get done and an exact question about it. Start there, and build the page, the offer and the flow backwards.
Taylor Gendron · October 27, 2026 · 5 min read
Understand Your Market: Count First, Then Choose Where to Compete
Guide
Understand Your Market: Count First, Then Choose Where to Compete
Before you spend on ads, outbound or a new website, map the market honestly: who fits, how many there are, where they are, who already wins them, and the narrow place where a small business can win first.
Taylor Gendron · October 27, 2026 · 5 min read
The Adjacent Possible: Your Next Product Is One Step From What You Already Have
Strategy
The Adjacent Possible: Your Next Product Is One Step From What You Already Have
The biologist Stuart Kauffman called it the adjacent possible: the next moves that sit one step from what already exists. For a small business, the best new products, services and channels reuse the capabilities, customers and data it already has. Here is how to see them and test one cheaply.
Taylor Gendron · October 27, 2026 · 5 min read
SKU Morphology: The Shape of Your Catalogue Decides What Sells
Guide
SKU Morphology: The Shape of Your Catalogue Decides What Sells
How you split products and services into SKUs, variants, attributes, bundles and price points decides whether customers, search engines and AI agents can find, compare and buy them. A practical look at the shape of a catalogue, and how it has to change as you grow.
Taylor Gendron · October 27, 2026 · 5 min read
The Tectonic Plates of Technology: From the Dot-Com Web to Robotics
Strategy
The Tectonic Plates of Technology: From the Dot-Com Web to Robotics
The web, the cloud, mobile, SaaS, AI and now robotics did not replace each other. They stacked. Each made one expensive thing nearly free, and each time the value moved to whatever stayed scarce.
Taylor Gendron · October 24, 2026 · 7 min read
From Chatbot to Digital Brain: What Changes for a Small Business
Strategy
From Chatbot to Digital Brain: What Changes for a Small Business
A chatbot answers questions. A Digital Brain does a piece of work from your own facts and asks you before anything goes out. Here is the difference in plain terms, and how to tell which one you are being sold.
Humonix Team · October 26, 2026 · 4 min read
Grounded AI: Why Your AI Should Answer From Your Facts, Not Its Guesses
Guide
Grounded AI: Why Your AI Should Answer From Your Facts, Not Its Guesses
An AI that sounds sure is not the same as an AI that is right. Grounding means it answers from your own price list, policies and pages, shows where each answer came from, and hands off to a person when it does not know.
Humonix Team · October 26, 2026 · 4 min read
Service as a Software: Buy the Outcome, Not Another Seat
Strategy
Service as a Software: Buy the Outcome, Not Another Seat
For years, small businesses bought software by the seat and did the work themselves. AI changes what you can buy: the finished work, at a fixed price, with you approving what goes out. Here is how to think about it.
Humonix Team · October 26, 2026 · 4 min read
RLHF vs RLVR: How AI Models Learn From People and From Checks
Research
RLHF vs RLVR: How AI Models Learn From People and From Checks
A language model is shaped twice: first by reading a huge amount of text, then by feedback on what it writes. Two methods do much of that second step. One learns from people's preferences, the other from answers a program can check. Here is how each works, where each stops, and why the difference matters when an agent does your reconciliations.
Humonix Team · November 18, 2026 · 8 min read
Synthetic Data: What It Is, and When Your Own Records Matter More
Research
Synthetic Data: What It Is, and When Your Own Records Matter More
More and more of the data used to train AI is written by AI. That can be useful, and it has real risks. Here is what synthetic data is, how labs use it, what researchers have warned about, and when a small business is better off with its own real records.
Humonix Team · November 18, 2026 · 8 min read
Frontier or Small Model? Choosing the Right AI Model for the Job
Strategy
Frontier or Small Model? Choosing the Right AI Model for the Job
Every few weeks a new model tops a chart. For a business, the useful question is different: which model is right for this task, at this cost, with this data? Here is a way to think about frontier, small and open models as choices, and how to test them on your own work.
Humonix Team · November 18, 2026 · 8 min read
AI, Energy and Compute: What Drives the Cost, and How to Use Less
Research
AI, Energy and Compute: What Drives the Cost, and How to Use Less
AI runs on electricity and chips, and both show up in what you pay. Here is a plain look at where the energy goes, the difference between training a model and using one, what the published numbers say and do not say, and four practical ways to use AI more efficiently.
Humonix Team · November 18, 2026 · 8 min read
Data vs Compute: Scaling Laws in Plain Language, and Why Your Records Are the Edge
Research
Data vs Compute: Scaling Laws in Plain Language, and Why Your Records Are the Edge
A few research results explain a lot about why AI models keep improving, and what they need to improve. In plain language: what scaling laws say, where compute stops being the whole story, why data quality matters, and why a business's own records are becoming its advantage.
Humonix Team · November 18, 2026 · 8 min read
AI and Intellectual Property: What a Business Should Check
Strategy
AI and Intellectual Property: What a Business Should Check
Who owns what AI writes? Can you be sued over how a model was trained? What happens to the documents you paste in? A plain-language checklist for businesses in Canada (including Québec) and the US, with the sources, and a clear note that this is general information and not legal advice.
Humonix Team · November 18, 2026 · 9 min read
Physical AI and Humanoid Robots: Demos, Deployments and What Is Hype
Research
Physical AI and Humanoid Robots: Demos, Deployments and What Is Hype
Humanoid robots make great videos. What are they doing in actual workplaces in November 2026? A plain look at what physical AI means, which deployments are documented and by whom, what the companies themselves admit, and what it could mean for a trades, distribution or field-service business.
Humonix Team · November 18, 2026 · 8 min read
LLMs vs World Models: A Fair Look at a Debate That Matters for Agents
Research
LLMs vs World Models: A Fair Look at a Debate That Matters for Agents
One of AI's best-known researchers says language models are not the road to human-level intelligence and is building something else: models that learn how the world works from video. Others say language models are still improving and already learn world structure. Here is each side, fairly, with sources, and what it means if you want an agent to act in the real world.
Humonix Team · November 18, 2026 · 8 min read
Reliability Is the Most Valuable AI Feature
Strategy
Reliability Is the Most Valuable AI Feature
Demos show what AI can do once. A business needs what it does every time. Here is why consistency, verification, approvals and an audit trail matter more than raw capability, what research says about agent consistency, how to test an agent on your own work, and what to ask a vendor.
Humonix Team · November 18, 2026 · 8 min read
When Everyone Has a Copilot: Agent-to-Agent, MCP and A2A Explained
Strategy
When Everyone Has a Copilot: Agent-to-Agent, MCP and A2A Explained
Soon a customer's assistant may ask your business for a quote, and your assistant may answer. Two open protocols, MCP and A2A, are how that is being built. Here is what each does, what is settled and what is not on trust, identity, consent and payment, and what a small business should do now.
Humonix Team · November 18, 2026 · 8 min read

The Humonix newsletter

What we learn putting AI agents to work for small businesses: what ran, what broke, what it cost. Published on Substack.