Every AI company, explained in plain words — and put to work in your business, on your accounts, under your control.
AI is on every business owner’s mind right now, and most of what you hear about it is either hype or fear. Here is the plain version. There are a handful of serious AI companies. Each one sells a business product with a contract, rules about your data, and an off-switch. Your staff are very likely already using the free personal versions — Microsoft’s research found 78% of AI users bring their own AI tools to work — and that, not AI itself, is where the risk sits. Henschel Technologies explains the choices, sets up the ones that fit your business, secures them, and trains your people to use them well.
Four companies matter for a business your size. They are not interchangeable: one lives inside the programs your team already opens every morning, and the other three are separate services that connect to your business in their own ways. Under each company you will find how it is charged — and, because the level matters more than the logo, what each level really gives you and what a business has to have to be protected. We don’t publish prices: the vendors change them, and the right setup depends on what your business already runs. For pricing, please contact us — tell us what you have, and we will quote the real numbers in writing before anything is ordered.
What it is. The only AI in this group that lives inside Outlook, Word, Excel, and Teams. It works from your own business data — your mailbox, files, chats, and meetings — through Microsoft Graph, and it only ever surfaces information the person asking already has permission to see. Microsoft states in its data commitments that your prompts, responses, and the data it reads are not used to train its foundation models.
Best at. The daily grind of an office that already runs on Microsoft 365: summarizing long email threads with links back to the source messages, drafting replies you review before sending, coaching on a draft’s tone and clarity, recapping meetings with decisions and action items, and answering questions from your own files with the sources shown.
How it fits. It is an add-on to the Microsoft 365 plan each person already has — no new system to learn. A basic chat version, Copilot Chat, is included free with eligible Microsoft 365 subscriptions; it answers from the web and from files you hand it, but it does not reach across your mailbox and file storage the way the paid version does.
How it’s charged. Per user, per month, on an annual plan. It is an add-on — each person also needs their own Microsoft 365 plan underneath it — or it comes bundled with Microsoft 365 Business Premium. One license, one person; there is no sharing.
Protected means. Copilot runs inside your company’s own Microsoft 365, under your admin console and your sign-in rules — turn a person’s account off, and their Copilot is off. A personal Microsoft account is a different thing entirely: consumer terms, and no control for the company at all.
What it is. The AI most of your staff have already met, bought properly. ChatGPT Business is a shared workspace for your team under a business agreement — OpenAI states that conversations on its business plans are excluded from training by default.
Best at. A familiar, capable writing and thinking partner for the whole office: first drafts, brainstorming, working through documents and spreadsheets you give it, and shared team assistants built for a single job. It connects to tools like Microsoft 365, Google Drive, and Slack.
How it fits. It runs as its own service with its own sign-in, tied to your company through single sign-on. It does not sit inside Outlook or Word the way Copilot does — your team goes to it. The strongest controls (automatic joiner/leaver sync, detailed audit logs, custom data retention) sit in the Enterprise tier, which is priced by quote.
How it’s charged. Per user, per month, billed annually on the Business plan (monthly billing is also offered). The Enterprise plan is priced by quote.
The personal level, plainly. A personal ChatGPT account runs under consumer terms: OpenAI may use those conversations to improve and train its models unless the person opts out — and the company has no console, no record of what was pasted in, and no way to cut it off when someone leaves. The account, and its history, walk out with them. The Business plan is the company level: a workspace of your own, an admin console, single sign-on, and member management — with no training on your data written into the plan terms.
Protected means. The company plan. Signing a personal ChatGPT account into Outlook does not turn it into company ChatGPT — it stays personal, whoever installed it.
What it is. A second serious engine, widely rated for careful writing and long documents. Anthropic’s commercial terms say it may not train its models on customer content from its business services.
Best at. Long, careful work: reading and summarizing big documents, drafting in a natural voice, and reasoning through complicated material. Through Anthropic’s Microsoft 365 connector, Claude can read a signed-in user’s Outlook mail and calendar, OneDrive, SharePoint, and Teams — only what that user can already open, fetched when asked, not copied away and stored.
How it fits. A separate service and console, not buttons inside Outlook — though Claude models can also be chosen inside Microsoft 365 Copilot itself. Its strongest controls (automatic joiner/leaver sync, audit logs, custom retention) are Enterprise-only, so the tier choice matters.
How it’s charged. Per seat, per month, billed annually. Team seats come in two sizes — Standard, and Premium with much higher usage — and the two can be mixed. Enterprise charges per seat plus usage on top, and is priced by quote.
The personal level, plainly. On free and personal plans, training on your conversations is on by default under Anthropic’s consumer terms, unless the person opts out. And a personal account signed in on a work machine is still a personal account — the company controls nothing about it. The Team plan is the company level: single sign-on, central billing, and admin controls, with no training on your content under Anthropic’s commercial terms.
Protected means. A company plan with an admin console, a no-training commitment in the contract, and sign-in control. Anything less is a personal tool being used at work — not protection, whoever installed it.
What it is. The newest of the group. Grok Business launched at the end of 2025 with a shared team workspace, central billing, usage reporting, and a Google Drive connection that answers only from files the user is allowed to see. xAI’s enterprise terms state that business content is not used to train its models and is deleted within 30 days.
Best at. A second opinion and research partner, strongest where its connection to live public conversation matters. It has the shortest business track record of the four, and no built-in connection to Outlook or Microsoft 365 comparable to Copilot or Claude’s connector was available at the time of our research.
How it fits. A separate service on business terms. Its business packaging was being folded into a new subscription ladder in late 2026 — so we flag it plainly: packaging changing, and the exact plan is confirmed at quote time.
How it’s charged. Per seat, per month on the business and Teams plans. Enterprise is priced by quote. And the caution above stands: xAI renamed its packages in October 2026, so we confirm the exact package and terms the day we set you up.
The levels, plainly. Personal Grok accounts on X are consumer accounts — the company has no console over them, no record, and no off-switch. The business plans add the company layer: a shared workspace with seat management, consolidated billing, and usage analytics, and no training on business content under xAI’s enterprise terms. Only on Enterprise: custom single sign-on, SCIM directory sync, advanced audit controls, and custom retention — with the isolated Vault data plane and customer-held encryption keys described as an Enterprise add-on.
Protected means. The same test as the other three: a company plan, an admin console, a no-training commitment, and sign-in control. A personal account plugged into a work machine meets none of them.
Strip away the marketing and AI earns its keep on ordinary office work — the reading, drafting, and note-taking that eats the day.
A forty-message thread becomes five bullet points with links back to the messages that matter. Replies get drafted in your tone from the thread itself — and a person reads and sends every one. Copilot never sends on its own.
Join a Teams meeting ten minutes late and ask what you missed. When it ends, the recap lists who said what, the decisions made, and the action items — instead of somebody’s half-remembered notes.
First drafts of proposals, letters, and reports from a short prompt or an existing file. Long documents summarized faithfully, with the source one click away so the answer can be checked.
“What did we quote them last spring?” “Where is the current policy on that?” The AI searches the files and mail the asker is allowed to see, and answers with citations — not guesses.
Ask a spreadsheet a plain question — the trend, the outliers, the total by month — and get the analysis, the formula, and the chart, explained in words.
Small assistants, built in Copilot Studio, each with one job and only the information that job needs: answering the routine questions your team gets daily, opening a support ticket, fetching the right document. Each has a named owner and a spending cap.
The risk is not AI. The risk is AI nobody set up. When an employee pastes a customer email, a quote, or a contract into a personal AI account, that information has left your building. There is no contract covering it, no log of it, and no way to get it back. On consumer AI services, training on what users type is often on by default unless the user opts out — and when that employee leaves, they keep the account and its history of your business.
IBM’s Cost of a Data Breach Report 2025 put numbers on it: 20% of breached organizations had a breach involving shadow AI — staff using AI the business never approved. Heavy shadow-AI use added an average of $670,000 to breach cost compared with little or no shadow AI. And 97% of organizations reporting an AI-related breach lacked proper AI access controls, while 63% had no AI governance policy at all.
A business deployment changes every line of that picture:
We look at what you already have — your Microsoft 365 plans, your sign-in security, your file sharing, and who does the kind of work AI helps with. No licenses bought yet, nothing switched on.
Which tool for which people, in what order, at what price — and what we will deliberately not buy. You approve it before a dollar moves.
Permissions clean-up first, then labels and sign-in security, then the licenses — starting with a pilot group of five to ten people who live in email and documents. Copilot Chat, the free protected option, can be turned on for everyone as the approved alternative from day one.
Short sessions per group, a shared library of prompts that work for your business, and the one habit that matters most: check the answer against its source before you rely on it.
A monthly rhythm: usage reports reviewed, seats moved away from anyone not using theirs, helper spending checked, and new AI tools your staff have started using spotted and dealt with. A named AI administrator holds the settings — never an all-powerful account. Purpose-built helpers come last, once the basics are bedded in.
Every stage has an owner, a report, and a way back. Any helper can be blocked the same day. That is the difference between AI your business runs and AI that just happened to it.
We will look at your systems, tell you plainly which AI fits and which does not, and put the plan in writing — whether you buy anything or not.
Have a computer problem? Call 855-IT-PROBLEM. The owner answers.
