Best AI tools for business

What are the best AI tools for a business, and what does the choice depend on?

The best AI tools differ by situation: the choice follows first from the task, the software you already have, data and privacy, admin, cost and exit, and only last from the brand. For a choice that affects more than a handful of licences, the Vendor Review helps reach a well-founded decision.

Do vendors pay for a place in the best AI tools comparison?

No. The comparison of the best AI tools is based on the vendors' own official product, pricing and privacy pages, as of October 2026. There are no affiliate, commission or partner relationships with the vendors named, and it contains no original benchmark figures.

How is a new AI tool trialled when choosing the best AI tools?

The advice is to have one team use a new AI tool on real work for a few weeks, with an agreed question the trial is meant to answer, not everyone at once and not based on a demo. For a choice from the best AI tools that affects more than a handful of licences, candidates are tested on the same task with their own examples.

Source: article by Dave Willemstein. Updated: October 2026.

How do you choose an AI tool?

“What are the best AI tools?” is an understandable question, but the honest answer is that it isn’t about the brand. The best AI tool for a law firm that runs entirely on Microsoft is a different one from the best tool for a marketing agency that works in Google Workspace. That’s why this page compares on criteria rather than producing a ranked list.

An honest caveat first: this comparison is based on the vendors’ own official product, pricing and privacy pages, as of October 2026 (re-checked against the figures first published in September 2026). I don’t publish my own benchmark numbers here. And I have no affiliate, commission or partner relationships with any of the vendors named; nobody pays for a mention.

These are the criteria I go through for every choice:

• Task: which work needs to be faster or better? Choosing a tool before you’ve named the task is looking for a solution to a problem you haven’t defined yet.
• Existing software: if you’re on Microsoft 365 or Google Workspace, AI is often already in your plan. That saves an extra contract and an extra login.
• Data and privacy: is your input used for training, where is the data stored, and can you sign a data processing agreement?
• Admin: single sign-on (SSO), user management, audit logs and configurable retention periods.
• Cost: licence per user, any usage on top of the licence, and the time setup and training cost.
• Exit: can you take your conversations, files and settings with you if you switch?

These criteria match the way I approach larger selections in AI vendor selection in eight steps: criteria and weighting first, names only after that.

1. 1. Which task?Text, customer queries, meetings, data or a recurring process
2. 2. What do you already have?Microsoft 365 or Google Workspace: start with Copilot or Gemini
3. 3. What data goes in?Business or personal data: only a business-tier plan with a data processing agreement
4. 4. Who manages it?Need SSO, user management and logs? Then a team or enterprise plan
5. 5. Only then: which brand?Compare two or three candidates on the same task with your own examples

Best AI tools by use case

Below are the most common business use cases, with the tools that fit logically. All product facts are as of October 2026.

Text and content

For writing, summarising, translating and rewriting, the general-purpose AI assistants are the most widely used: OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini and Microsoft Copilot. The differences sit less in “can it write?” and more in the environment around it.

• ChatGPT Business offers projects, shared GPTs, deep research and integrations with Google Workspace, Slack, GitHub and Microsoft 365, among others. How to put it to business use is covered in ChatGPT for business.
• Claude Team offers projects, connectors and search across your organisation’s own sources. More on that in Claude for business.
• Gemini sits inside the Google Workspace plans, in Gmail and, depending on the plan, in Docs, Meet and other apps.
• Microsoft Copilot (formerly Microsoft 365 Copilot) works in Word, Excel, PowerPoint, Outlook and Teams. The different Copilot variants are covered in Microsoft Copilot for business.

Whichever assistant you choose, the quality of the output depends heavily on how you phrase the request, a skill often called prompt engineering.

Customer service

In customer service, it’s rarely about a standalone chat tool, but about an AI agent running inside your helpdesk. One example is Intercom Fin: according to the vendor it works inside Intercom’s own platform and also on an existing helpdesk such as Salesforce, priced from $0.99 per “Fin outcome” (as of October 2026). Other helpdesk vendors offer comparable features.

The key criterion here isn’t the tool, but your knowledge base. An AI agent answers based on the articles and procedures you give it. If those are outdated or incomplete, your customer gets a confident wrong answer quickly. Also set out in advance when a conversation gets handed to a staff member.

Meeting notes

For summarising meetings you often don’t need to buy anything new. Microsoft Copilot works in Teams, Gemini sits in Meet depending on your Workspace plan, and ChatGPT Business has a recording mode. Specialist meeting-notes tools exist as well.

The choice depends on where you meet, who else is listening in, and where the recordings and transcripts are kept. Because you’re capturing participants’ spoken contributions, UK GDPR and EU GDPR apply directly here, including what you tell participants beforehand.

Data and analysis

For analysing spreadsheets and exports, the general-purpose assistants offer an analysis feature: ChatGPT Business has data analysis, and Copilot works inside Excel. That’s strong for exploring, summarising and a first chart.

For recurring reporting, an assistant usually isn’t the right place. There you want a fixed source, fixed definitions and a BI environment that shows the same number every month. An AI assistant helps you ask questions of that data, but it doesn’t replace the data foundation. Why that foundation matters so much is covered in the data foundation under AI.

Automation

For automating processes you end up with workflow tools. Examples include n8n, which you can self-host (Community Edition) or take as a hosted service with storage in Frankfurt, and Zapier, which connects AI to thousands of apps. With tools like these, an AI step becomes part of a process: classifying an incoming email, pulling data out of a document, preparing a draft.

Here the question shifts from “which tool?” to “which process?”. How to choose a suitable process is covered under automate business processes with AI, and how to build such a chain with checkpoints under AI workflows.

New AI tools to watch

The biggest shift in 2026 is that assistants are moving from “answering” to “carrying out tasks”. A few developments I’m tracking, as of October 2026:

• Claude Cowork. According to Anthropic, you give it a goal and Claude works independently across your files and tools, with a built-in browser for web tasks. It’s available on paid plans.
• Copilot Cowork. Microsoft announced on 16 June 2026 that Copilot Cowork is generally available for customers with a paid Copilot licence. It carries out longer tasks end to end across several tools. Usage is billed on top of the licence per task, according to Microsoft.
• ChatGPT Work and workspace agents. According to OpenAI, ChatGPT Business also includes ChatGPT Work and workspace agents, which let you run your own workflows.

Features like these are promising, but need different management than a chat window. A tool that edits files or fills in forms by itself deserves the same questions as a new colleague: what permissions does it get, who checks the work, and how do you spot a mistake? What AI agents can and cannot do is covered in what AI agents can and cannot do.

My advice for any new AI tool: have one team use it on real work for a few weeks, with an agreed question the trial is meant to answer. Not everyone at once, and not based on a demo.

AI tools for professional services firms

Accountants, lawyers, consultancies and other professional services firms work with confidential client information. That makes three criteria weigh more heavily than for other organisations:

1. Training and storage. Choose a plan where your input isn’t used for training by default, and where you know where the data sits. ChatGPT Enterprise offers data storage in Europe (the EEA and Switzerland) as a region for new workspaces. For Claude, data on the business plans sits in the US according to Anthropic; European regions only exist for custom deployments via the cloud platforms AWS Bedrock and Google Cloud Vertex, and, according to Anthropic, soon also via Microsoft Foundry. Microsoft Copilot falls under the EU Data Boundary, with the exception of the Anthropic models Microsoft uses as a sub-processor. In Google Workspace you can set a data region (US or Europe) that also applies to Gemini; what’s covered differs by edition.
2. Retention and logs. Can you set how long conversations are kept, and can you trace afterwards who did what? Those features usually sit in the enterprise plans, not the team plans.
3. Professional rules. Your professional body or regulator may set its own requirements for using AI with client data. Check your tool choice against those before rolling it out.

For services firms it’s tempting to start with the tool with the nicest demo. It’s better to start with the work that costs the most hours and delivers the least. Which applications pay off the most there is covered on the AI for professional services firms page. And make sure a short internal AI policy is in place before rollout.

Comparison table

The table summarises the verified features, as of October 2026. Prices are per user per month billed annually, excluding VAT. For ChatGPT, Claude, Microsoft Copilot and Intercom these are US list prices in dollars, taken directly from the vendors’ global pricing pages; OpenAI and Anthropic state that prices can vary by country and currency. For Gemini and n8n these are the euro prices from their EU pricing pages. “Ask the vendor” means I couldn’t establish it unambiguously from the official pages.

Tool	Type	Good fit for	Business plan and price guidance	Trains on business data	Data storage in Europe
ChatGPT (OpenAI)	General-purpose assistant	Writing, research, analysis, custom GPTs	Business: $20 ($25 billed monthly), from 2 users. Enterprise: price on request	Not by default	Enterprise: yes, Europe (EEA and Switzerland) as a region. Business: no
Claude (Anthropic)	General-purpose assistant	Long documents, writing, code, delegating tasks	Team: $20 ($25 billed monthly), 2 to 150 users. Enterprise: $20 per seat plus usage at API rates, minimum 20 seats	Not by default	Storage in the US; EU regions only via cloud platforms (API)
Gemini (Google Workspace)	Built into office suite	Organisations on Gmail, Docs and Meet	Part of the Workspace plans, from €6.80 (Business Starter, limited Gemini features); scope differs by edition	Not without permission outside your domain	Data region US or Europe configurable from Business Standard; scope differs by edition
Microsoft Copilot (formerly Microsoft 365 Copilot)	Built into office suite	Organisations on Word, Excel, Outlook and Teams	Copilot Chat at no extra cost on eligible plans. Copilot Business: $18, up to 300 users, discounted from an original $21 (promotional pricing, confirmed live as of October 2026). Enterprise licence: $30	Not for training the base models	EU Data Boundary, except the Anthropic models
Intercom Fin	Customer service agent	Support teams with a good knowledge base	From $0.99 per Fin outcome, plus Intercom seats or your own helpdesk	Ask the vendor	Ask the vendor
n8n	Workflow automation	Processes with AI steps, technical teams	Starter from €20 a month billed annually; Community Edition self-hostable	Depends on the connected model	Hosted: Frankfurt. Self-hosted: your own choice

Use the table as a starting point, not a final verdict. For a choice that affects more than a handful of licences, build your own scorecard with your own weighting and test the candidates on the same task with your own examples. If you want an independent view for that, a Vendor Review helps you reach a well-founded decision. If you want to know first whether your organisation is ready, the AI Readiness Assessment is the logical starting point.

A tool is only worth anything once people actually use it in their daily work. Why adoption starts there, and not with the licence, is covered in AI adoption starts with the work. More on how AI tools fit into broader process automation is on the AI automation hub. Not sure which tool fits your situation? Book a thirty-minute call; you’ll hear what I’d do by the end.