- AI (Artificial Intelligence)
- Software that reads, writes, and reasons about text, images, or data — useful for tasks that used to need a human in the loop.
- In 2026, when a Montana business owner says 'AI,' they almost always mean a large language model (LLM) like ChatGPT, Claude, or Microsoft Copilot. These tools take words in and give words out, and they're useful for customer messages, paperwork, summarization, and answering questions about your own business documents.
- LLM (Large Language Model)
- The kind of AI behind ChatGPT, Claude, Gemini, and Microsoft Copilot. Trained on lots of text; good at reading and writing.
- LLMs are the most useful AI category for small business. They reason about text, follow instructions, and can be pointed at your own documents (see RAG below). Examples: GPT-5, Claude, Gemini, Llama, Mistral.
- Microsoft Copilot
- Microsoft's AI assistant built into Outlook, Word, Excel, Teams, and PowerPoint. Sold as Microsoft 365 Copilot.
- Copilot is the right default for most Montana businesses already on Microsoft 365. It can draft emails in Outlook, summarize Teams meetings, build PowerPoint decks, and analyze spreadsheets — all using your business data with enterprise data-protection on by default.
- ChatGPT
- OpenAI's general-purpose AI chatbot. Comes in free, Plus, Team, and Enterprise tiers.
- ChatGPT Team and Enterprise tiers do not train on your business data and are the right pick for client work. The free tier is fine for personal use but should not be used for confidential business information.
- Claude
- Anthropic's AI assistant. Known for long-document handling and careful reasoning.
- Claude is a strong pick when your work involves reading long contracts, transcripts, or technical documents. Available via Claude.ai and via API.
- Gemini
- Google's AI assistant. Built into Google Workspace.
- Gemini is the right default for Montana businesses already on Google Workspace. It works inside Gmail, Docs, Sheets, and Meet, similar to how Copilot works inside Microsoft 365.
- RAG (Retrieval-Augmented Generation)
- Pointing an AI at your own documents so it answers from your business's information, not the public internet.
- RAG is how 'AI that knows about your business' works. We index your documents (contracts, SOPs, customer records, manuals), and when the AI is asked a question, it retrieves the relevant passages and uses them to write an accurate, citation-backed answer. This is how chat assistants for SMBs are built in 2026.
- Agent
- An AI that can take actions on your behalf — send emails, update records, schedule meetings — not just write text.
- Agents are the next step beyond chat. Instead of just answering a question, an agent can read an incoming email, look up the customer, check inventory, draft a quote, and queue it for your review. For Montana SMBs, agents are useful for repetitive multi-step work like quoting, invoicing, and follow-up.
- Hallucination
- When an AI confidently states something that isn't true.
- Hallucinations happen because LLMs predict likely-sounding text. They're rare in well-built systems with RAG and source citations, but always possible. The rule: any AI output that matters legally, medically, financially, or for safety must be reviewed by a qualified human before action.
- Prompt
- The instructions you give an AI. Better prompts produce better answers.
- A good prompt names the role, the task, the inputs, the constraints, and the desired output shape. Most of the value from training is teaching teams to write better prompts for their actual work.
- Fine-tuning
- Customizing a base AI model on examples specific to your business.
- Fine-tuning used to be the default way to customize AI; in 2026 it's rarely the right first move for an SMB. RAG, good prompts, and Copilot/Custom GPTs cover 95% of small-business needs at a fraction of the cost.
- Custom GPT
- A version of ChatGPT pre-loaded with your instructions and documents. No coding required.
- Custom GPTs are the fastest way to give a team a shared AI tool that knows your business. We frequently ship a 'house Custom GPT' as part of training engagements.
- MCP (Model Context Protocol)
- An open standard for letting AI tools securely connect to your business systems.
- MCP, introduced by Anthropic in 2024 and widely adopted in 2026, lets an AI assistant access tools like your CRM, calendar, or accounting software through a consistent, permissioned interface. We use MCP where it removes the need for brittle custom integrations.
- Token
- The unit AI bills you in. Roughly one token ≈ ¾ of a word.
- Tokens matter for cost estimation. A typical short business chat is a few hundred tokens; reading a long PDF can be tens of thousands. Modern models price tokens in fractions of a cent.
- Context Window
- How much text an AI can consider at once.
- In 2026, mainstream models comfortably hold a small book in context (200k+ tokens). For most SMB use cases, context window is no longer the bottleneck — quality of retrieval and prompting is.
- AEO (Answer Engine Optimization)
- Making your business visible in AI answers (ChatGPT, Perplexity, Google AI Overviews), not just blue-link search.
- AEO is the evolution of SEO. It includes structured data, citable content, AI-crawler access, and direct-answer copy. For Montana businesses, AEO is the difference between being recommended and being invisible when a customer asks an assistant for help.
- BAA (Business Associate Agreement)
- The HIPAA contract required when a vendor handles patient information.
- Required for Montana clinics using AI tools that touch PHI. Microsoft (Copilot for Microsoft 365), Google (Workspace), and OpenAI (Enterprise) all offer BAAs on appropriate tiers. We won't deploy AI against PHI without one in place.
- Pilot
- A small, time-boxed test of an AI tool with real users before rolling it out broadly.
- Every engagement we run starts with a pilot — usually two people, 30 days, measured hours saved. If the pilot doesn't pay for itself, we kill it cleanly and try a different workflow.