Microsoft
Microsoft 365 and Azure
- Cowork
- Copilot Cowork
Generally available since June 2026 for Microsoft 365 Copilot customers, billed by usage. Admins choose who gets it.
- Code
AI in business, explained
Types of AI in business:
chat, cowork, code and bots
Most businesses meet AI as a chat window. There are now four kinds of tool in everyday use, and the biggest change is bots: AI that carries out routine work across your systems, with a person in charge. Here is what each type does, what Microsoft, Claude and Gemini offer, and how to deploy bots safely in the UK.
The four types
From answering questions to doing the work.
Each step hands the AI more of the task. Chat helps a person work faster. Bots take a routine off the team entirely, and bring the decisions back to a person.
Chat
Ask a question and get an answer or a draft. You do every step, and you decide what to use.
Good for: Research, writing, summarising documents and meetings, and thinking a problem through.
Cowork
Give it a goal and it works through the steps in your files, email and apps, checking with you before it acts.
Good for: Preparing reports, briefings and documents that pull from several sources.
Code
AI that writes, tests and fixes software alongside a developer.
Good for: Internal tools, integrations between systems, scripts and the automations behind bots.
Bots
AI agents that run a business routine on a trigger or a schedule, across your systems, and hand decisions to a person.
Good for: Repeated, multi-step work: inbox triage, order checks, prospect research, reconciliations and weekly reporting.
The main platforms
What Microsoft, Claude and Gemini offer for each type.
All three cover every type. The right choice usually follows the systems you already use and the work you want to change, not the brand.
Claude
Anthropic
- Code
Gemini
- Cowork
Product names, features and licensing change often, and what you can use depends on your plan and admin settings. Links go to each vendor's own page and were checked on 1 November 2026. Naming a product here does not imply a partnership.
Bots
AI that works like a digital employee, with a person in charge.
A chatbot answers questions. A bot does work. Something starts it, such as a new email, a form or a schedule. It plans the steps, uses your systems through approved connections, and prepares the result. A person approves anything that changes a record, contacts a customer or moves money.
From one bot to a team of agents
The newest systems use several agents working together. A coordinating agent breaks a goal into tasks and hands them to specialists, one to research, one to draft, one to check, then combines the results. Teams of agents can take on longer routines, but every extra agent is another set of permissions, costs and failure points to manage. We start with one bot doing one routine well.
How bots connect to your systems
Through each system's API, through the platform's own connectors, or through the Model Context Protocol (MCP), an open standard that lets AI tools use a system's functions within the permissions you grant.
What bots do well in a UK business
Inbox and service
Reads every enquiry, sorts it, drafts the reply and logs the job, escalating anything urgent.
Orders and quotes
Checks each order against the accepted quote and flags price or delivery exceptions before release.
Prospecting
Researches buying signals, scores accounts with reasons and drafts outreach for a rep to approve.
Finance routines
Matches invoices, chases overdue payments in your tone and prepares month-end exceptions.
Management reporting
Pulls figures from your systems every Monday and writes a briefing on what needs attention.
Deploying bots safely in the UK
The question is always: is it safe, and where does our data go?
The answer depends on how the bot is hosted, which identity it uses and what it is allowed to do. No hosting choice makes a system compliant by itself. Compliance comes from how it is configured, documented and run.
Vendor-hosted (SaaS)
- Getting started
- Fastest to start
- Control over data
- The vendor runs it. You rely on their regions, retention settings and data terms.
- Best for
- Pilots, standard use cases and teams without in-house cloud skills.
Hosted in your own cloud tenant
- Getting started
- More set-up and running effort
- Control over data
- You choose the region, such as Azure UK South (London) or UK West (Cardiff), the network boundary and the logging.
- Best for
- Sensitive data, regulated work and bots that touch core systems.
Hosting a bot in a UK region is not the same as keeping every step in the UK. The AI model it calls may run elsewhere unless you choose a model deployment in a UK region, and each provider has its own retention terms. Check both before you rely on a residency claim. Azure lists its UK South and UK West regions on its regions page.
Give every bot its own identity
Run bots under their own accounts, with only the access they need, managed through your existing sign-on. Microsoft's Entra Agent ID does this for agents in Microsoft environments.
Approve before it acts
Require a person to approve any change to a record, message to a customer or payment, and make the system enforce it rather than relying on an instruction in the prompt. Our guide to what an AI agent should be allowed to do covers how.
Log it and assess it
Keep a record of every action and tool call. UK GDPR still applies to bots: complete a data protection impact assessment where the ICO's criteria apply, which new technology processing personal data often meets.
Questions to ask before you deploy a bot
- What exactly may the bot read, prepare and change?
- Which actions need a person to approve them, and is that enforced in the system, not just the prompt?
- Where are prompts, files, outputs and logs processed and stored, including by the AI model provider?
- Which identity does the bot use, and does it have only the access it needs?
- What happens when it is unsure, an integration fails or it makes a mistake?
- How will we measure whether it saves time or improves accuracy?
- What will it cost to run each month, including AI usage?
Where we help
From advice and training to automations and bots.
We help you choose which types of AI fit your business, train your team on chat and cowork tools, build the code and automations behind bots, and run them with you once they are live. Scope and price are agreed before any paid work.
Free initial conversation