Your business software is becoming one thing, not ten

AI tools are stopping being something you open, and starting to become the thing everything else runs through.

What that actually looks like

Right now, a lot of small businesses use AI the way people used to use a calculator: open it, do the one thing, close it. Ask ChatGPT to draft an email. Ask Copilot to summarise a document. Separate tools, separate moments.

The shift underway is toward all of that happening in one place: your AI assistant reads your inbox, checks your calendar, drafts the follow-up, and does it without you manually copying information between five different tabs. Not because the AI got smarter overnight, but because the connections between it and your other software got built properly.

The plumbing behind it, in plain terms

The technical piece making this possible is something called MCP: a standard way for AI tools and business software to talk to each other. Before this existed, every connection between "your AI" and "your CRM" or "your accounting software" had to be built one at a time, by hand, for that specific pair of tools.

MCP is closer to a universal adapter, one connection type that works across many tools, instead of a custom cable for every plug.

You don't need to understand MCP to benefit from it. You just need to know that it's why "connect your AI assistant to your actual business systems" is becoming a realistic afternoon's setup rather than a custom software project.

Why the timing actually matters

This is worth acting on now, not eventually. Competitors who get their tools properly connected are cutting hours out of routine work (quoting, follow-ups, admin) while everyone else is still doing it by hand across five tabs.

That gap compounds: the business running on a connected setup gets faster and cheaper to run every month it keeps going, and the business that waits is catching up from further behind each time it finally looks.

But moving fast without thinking creates a second, quieter risk: it happens gradually enough that nobody makes a deliberate decision about it. Staff start using whichever AI tool is bundled into the software they already have, connections get added one at a time without anyone stepping back to ask what's now reading what, and six months later nobody could give a clean answer to "what does our AI actually have access to."

So the real urgency isn't "adopt anything, immediately." It's "get a properly set-up version of this running soon, rather than an accidental, ungoverned one." The businesses that move now with a plan will be ahead. The ones that move now without one just get into trouble faster.

What we'd actually suggest, if you're a small business

• Don't chase "which AI tool is best." The honest answer, increasingly, is that they're converging: Copilot, Gemini, Claude and ChatGPT are all heading toward the same "one assistant, connected to everything" shape.

Pick based on what you already pay for (Microsoft 365 or Google Workspace often already include one) rather than chasing whichever one is loudest this month.

• Know what's connected, on purpose. Every new connection is a new thing that can read your information, and potentially act on it.

That's fine. It's also exactly why "one folder, not the whole drive" is worth deciding deliberately rather than clicking "allow" repeatedly without noticing.

• Keep a human reading anything before it goes out. The orchestration getting better doesn't mean the oversight gets to disappear. A drafted follow-up email is a time-saver. An email that sent itself is a different risk entirely, and nothing about "AI as an operating system" changes that rule.

The technology behind this is real and moving fast. The mistake would be treating "which platform" as the decision that matters, when the actual decision (what it can see, and who checks its work) is the one that was always yours to make.

The current players

As of September 2026, here's roughly who each one suits and what it actually does well.

Claude Cowork

Best fit for teams already using Claude, and for functions like marketing, sales, and legal that Anthropic has built specific bundles around.

Cowork used to be a separate mode you switched into. As of this month, Anthropic folded it directly into the main Claude interface: chat, agent work, and Artifacts all live in one window now, and memory is shared between them, so a task Claude runs in the background can draw on something you told it earlier instead of starting cold. On the admin side, businesses can manage feature access and spend, track usage across the organisation, and connect Cowork to Google Drive, Gmail, DocuSign, and FactSet. Claude Docs and Claude Slides also shipped alongside this, so drafting, editing, and presenting a document or deck can happen inside Claude itself, then export to PowerPoint or PDF when you need to hand it off.

Microsoft 365 Copilot

Best fit for a business already deep in Microsoft 365, especially one with real data sitting in Power BI, SharePoint, or Excel that's currently going unused.

Copilot can now reason directly over enterprise data from Power BI reports and semantic models, and "bring your own knowledge" connectors let it pull from non-Microsoft systems too, ServiceNow, SAP, Salesforce, or a custom REST API, so it isn't limited to what already lives inside Microsoft's own apps. It creates and queries Planner tasks, so task management doesn't need a separate check-in. An adjustable effort setting (Light through Max) lets you trade speed and cost against depth depending on the task, rather than every request running at the same setting. On the admin side, usage analytics in Insights let a business see adoption trends across the organisation, not just individual activity. The honest trade-off, and it's the same one that's always applied here: licensing is the messiest of the major platforms, a per-user add-on with tenant admin involved, so setup takes longer than the others even though the ceiling on what it can do with a business's existing data is high.

ChatGPT for Work

Best fit for businesses that want the widest possible reach into other software, since the value here is breadth of connections more than any one specific feature.

Launched in July 2026 as OpenAI's full productivity suite for Pro, Enterprise, and Edu subscribers, running on GPT-5.6, which brings a 1.5 million token context window on Enterprise plans. It connects to more than 1,400 workplace apps and can carry a multi-step project through to a finished document, spreadsheet, presentation, or dashboard. Workspace Agents let a team set up its own repeatable automations across those connected apps, usable from ChatGPT or Slack, with admins able to limit what an agent can do in each app and see what it's actually been doing. Deep Research can now pull from the web, your own files, and connected apps together, then hand back a cited, editable document instead of a chat answer you have to reformat yourself.

Google Gemini Workspace

Best fit for a business already on Google Workspace that wants AI built into the tools it uses every day, without adopting a separate platform.

Gemini sits inside Gmail, Docs, Sheets, Slides, Drive, Meet, Chat, and Vids: drafting and summarising in Gmail, generating a first draft or reacting to team edits in Docs, building formulas and trackers from a plain description in Sheets, searching your whole Drive in natural language, and producing meeting summaries and action items in Meet, even when the meeting itself was hosted on Zoom or Teams. Workspace Studio lets a team build a reusable skill that encodes its own rules and templates, so Gemini follows house process instead of generic defaults. Pricing is the simplest of the three: Business plans run roughly $7 to $22 per user a month with Gemini already included, so for a business already paying for Workspace, this is often the cheapest way in rather than a new subscription on top.