Copilot, ChatGPT or Claude for a property firm

The short answer. If your firm runs on Microsoft 365 and the work you want help with lives in Outlook, Word, Excel and SharePoint, start with Copilot, because it sits inside those apps, it inherits the permissions you already have, and it tends to be the one IT signs off without an argument. For the harder drafting, the long documents and the analysis, Claude is what I reach for and what I use every day. ChatGPT is the other strong general option and a fair choice if that is where your people already are. Most firms of any size end up running two: one broadly, one for the handful of people who push hardest. That is a sensible outcome, not a failure to decide.

A disclosure before the rest of it. Claude is the tool I use daily, so treat the recommendation as a preference with a reason attached rather than a neutral verdict. I am not a reseller, partner or affiliate for any of the three, and nobody pays me to name a tool. The section on where Claude is weaker is there because a recommendation without one is an advert.

The comparison articles ranking for this question are mostly written by resellers of one product or the other, and almost none of them mention the two things that decide it in practice: what your existing licence forces on you, and the state of your SharePoint permissions.

What each one is for

Copilot is AI where your work already is. Claude and ChatGPT are AI you bring work to, and both have spent the last year moving closer to where the work sits.

Copilot's pitch is context. It sees your mailbox, your files, your calendar and your Teams messages, so a question like "summarise where we got to with this landlord" has something to work from. Microsoft states that it "inherits your Microsoft 365 permissions, sensitivity labels, and retention policies" (Microsoft), so a user sees what they were already entitled to see.

Claude and ChatGPT start in their own window and give you a stronger, more flexible general model. You paste in the lease, the notes, the spreadsheet. For long reasoning tasks, awkward drafting and anything analytical, most people who use both will tell you the general models are the better instrument, at the cost of moving work in and out.

That gap has narrowed. Anthropic now ships add-ins for the Office apps: "Claude for Excel, PowerPoint, and Word are generally available on all paid Claude plans. Claude for Outlook is in beta on all paid plans" (Anthropic). There is also a Microsoft 365 connector covering SharePoint, OneDrive, Outlook and Teams, available on every plan (Anthropic).

The distinction that survives is architectural. Copilot is built into the app surface by the company that owns the app. Claude arrives as a task pane add-in and a connector. Both reach your work. One of them is a tenant in the building, the other owns it, and the difference shows up in the governance section below rather than in the drafting quality.

For a property firm, the trade lands in a specific place. Property work is document-heavy and inbox-heavy, and most of the documents already sit in Microsoft. That biases the everyday towards Copilot, with a general model earning its seat on the harder, less frequent jobs. Which general model is where my preference comes in: for reading a hundred-page lease and telling me what is in it, Claude is the one I trust to say "this is unclear" rather than produce a tidy, confident answer with an invented date in it.

What it costs

Check every figure yourself before you commit. They move, promotional rates expire, and the reseller emailing you has an interest in the number.

As at 12 August 2026, on the vendors' own pages:

Microsoft, published in sterling:

Microsoft 365 Copilot Business, the add-on, advertised at "Originally starting from £16.10 now starting from £13.80 user/month, paid yearly", or £19.32 on a monthly commitment. "A Microsoft 365 Business plan is required" (Microsoft 365 Copilot pricing). That discounted rate is promotional and Microsoft's page shows it running to 30 September 2026, after which the £16.10 applies, so budget on the higher figure.
Bundled, Business Standard with Copilot at £18.10 and Business Premium with Copilot at £24.60 per user per month, paid yearly (Microsoft 365 Business plans).
Microsoft 365 Copilot on the enterprise page at £23.10 per user per month, paid yearly (enterprise pricing).
All exclude VAT.

Anthropic, published in dollars:

Pro, the single-user plan: "$17 per month with annual subscription discount ($200 billed up front). $20 if billed monthly".
Team: "$20 Per seat / month if billed annually. $25 if billed monthly", "For teams of 2 to 150". A premium seat with five times the usage runs at $100 a seat annually, $125 monthly.
Enterprise: "Seat price + usage at API rates", with the seat at $20. Anthropic's own help centre is blunt about what that means: "The seat fee only covers access to the platform and doesn't include any usage" (Anthropic).
The pricing page footnote reads "Prices shown don't include applicable tax. Price and plans are subject to change at Anthropic's discretion" (Claude pricing).

Three things a UK buyer should take from that.

Anthropic publishes no sterling price. Every figure on the pricing page is in dollars, and Anthropic's help centre states that "Enterprise plans require a minimum of 20 seats" (Anthropic) on the self-serve route. Billing in another currency means a conversation with their sales team. A twelve-person firm buys in dollars and carries the exchange rate as well as the VAT. Microsoft quotes you in pounds. For some finance directors that settles it on its own.

Enterprise on Claude is not a flat per-head cost. Seat plus consumption is a different budgeting exercise from Copilot's fixed price per user, and a document-heavy firm is exactly the profile that consumes. Admin spend caps exist, so use them.

For ChatGPT, read OpenAI's current pricing directly. Business and enterprise tiers are priced per user with different admin controls at each level, and the numbers have changed more than once. I am not quoting a figure I have not checked today.

The number that matters is not the headline price anyway. It is the price multiplied by the people who will use it in month three. Ten seats that get used beat forty that do not.

The two things that quietly decide it

Your existing licence. Copilot sits on top of a Microsoft 365 subscription, and the SKU you already hold constrains what you buy. Microsoft's Business plans are "Available for up to 300 users" (Microsoft), so a growing firm crossing that line moves onto enterprise licensing and a different price. Firms on Business Basic have no desktop apps, which changes what an in-app assistant is worth. And a firm that grew by acquisition often has three licence types across four offices, which is worth untangling before anyone buys AI on top of the mess.

Your SharePoint permissions. This is the one nobody warns you about, and it applies to all three. Every one of them reaches your documents through your own access rights, which is a genuine safeguard and also the catch: if a decade of loose sharing means half the firm technically reaches the folder with the partner drawings, the salary review or the sensitive client file, an AI assistant will find it and surface it in an answer. It has broken nothing. It has revealed what was already true. Property firms with long-lived shared drives are exactly the profile where this bites. Audit sharing before you roll out, not after someone asks an innocent question and gets an uncomfortable answer.

What they say about your data

All three say business-tier content is not used to train their models. Read the terms for the specific tier you are buying rather than the marketing page, because the tier is what decides it.

Microsoft states that with Microsoft 365 Copilot, "Prompts, inputs, and responses are never used to train the models" (Microsoft).

Anthropic puts it in the commercial contract rather than a brochure. The Commercial Terms of Service state: "Anthropic may not train models on Customer Content from Services" (Anthropic Commercial Terms). That is the line to hold them to.

OpenAI makes an equivalent commitment for its business and enterprise tiers; read it on their site as it stands when you buy.

Three practical points that matter more than any of those statements.

The tier is what counts, and Pro is not a business tier. A small firm putting five people on Claude Pro is on the consumer footing, where Anthropic's position is that training happens if "You choose to allow us to use your chats and coding sessions to improve Claude" (Anthropic). The setting sits with the user, which means it sits with whoever clicked through the onboarding. On Team or Enterprise you are under the commercial terms instead. If you are handling client documents, buy the team tier, not five individual subscriptions.

Safety review outlives your retention settings. Anthropic states that content flagged by trust and safety systems is held for up to two years, and classification scores for up to seven (Anthropic). Every provider does something along these lines. Do not promise a client that nothing persists.

UK data residency is not on offer first-party. Anthropic's regional compliance page lists Europe, the US, Canada and Asia-Pacific, with Europe and Canada marked as coming during 2026, and no UK region (Anthropic). Firms that need in-region processing are pointed at running Claude models through AWS Bedrock or Google Cloud, which is a heavier deployment than handing someone a login. If residency is a hard requirement of a client contract, sort that out before you choose a tool, not after.

Where Claude is weaker than the brochure suggests

I recommend it, so this section carries more weight than the last one.

Office add-in activity does not reach your audit logs. This is the big one for a regulated firm. Anthropic's own documentation states that the Microsoft 365 add-ins "do not inherit custom data retention settings your organization may have set, and activity is not included in Enterprise audit logs, the Compliance API, or data exports. Chat history is stored locally in your browser, not on Anthropic's servers" (Anthropic). So a firm buying Enterprise for the audit trail does not get audit coverage of what people do inside Excel and Word. Copilot activity lands in Microsoft Purview. If your compliance position depends on being able to show what was asked and when, that difference is the whole argument.

SharePoint access is tenant-wide or nothing. Individual users are still limited to their own permissions, and Anthropic is clear that "Users cannot bypass SharePoint sharing settings or folder permissions" (Anthropic). But the same page states that "SharePoint search requires Sites.Read.All permission. Site-specific permissioning (using *.Selected permissions) is not supported". A firm that wants to scope AI to one site collection while a migration finishes cannot do it. Copilot allows finer scoping.

Sensitivity labels are undocumented. Anthropic says delegated permissions "inherently respect Microsoft 365 data loss prevention (DLP) policies", and says nothing anywhere about Purview sensitivity labels, whether they are read, honoured, or applied to output. Microsoft names label inheritance explicitly. If your firm runs labels seriously, ask the question and get the answer in writing.

Your Global Administrator has to sign off. Even a single Pro user connecting to the firm's SharePoint needs tenant consent: "A Microsoft Entra Global Administrator still needs to grant one-time tenant consent for your Microsoft organization" (Anthropic). No quiet trial against real firm data. Which is correct, and worth knowing before you promise a partner a two-week pilot.

Browser support is Chrome only. Claude's browser extension is Chrome, and Anthropic states it "is not supported on other Chromium-based web browsers" (Anthropic). A Microsoft-shop property firm standardised on Edge will notice.

None of that changes what I use. It does change what I would tell a compliance officer, and a firm that buys on the strength of the drafting quality and discovers the audit gap in month four has been sold to rather than advised.

How I would decide it in a property firm

A short version, in order.

Are you on Microsoft 365 and is your work in it? If yes, Copilot is the default starting point, because adoption is the hard part and an assistant living inside Outlook gets used.
Is your SharePoint tidy? If no, fix that first. This is a genuine prerequisite, not a nice-to-have. It applies whichever tool you pick.
Buy a pilot, not a rollout. Ten to fifteen seats across your busiest desks for a quarter. Measure usage weekly. Anyone not using it after four weeks loses the seat to someone on the waiting list.
Add a general model for the heavy users. Claude is my pick, on a Team plan rather than individual Pro subscriptions, so you are under the commercial terms and you get single sign-on. The people doing long-form analysis, lease work and complex drafting will notice the difference, and they are the ones who bring the good ideas back. If your team already lives in ChatGPT, that works too and the argument is not worth the disruption.
Check the governance gap before you widen it. If your compliance position needs an audit trail, read the section above again and decide where the general model is allowed to touch firm documents.
Train before you scale. Firms that buy licences and send an email get single-digit adoption. Firms that run a session on people's own work get the opposite. This is the step that decides whether the spend was worth it, and it is the one that gets skipped.

The tool question is genuinely less important than it looks. The habits transfer. Someone who learns to brief AI well on Copilot is good at Claude within a day. What does not transfer is a team that never got shown how, which is why the licence choice is the smallest decision in this article.

Frequently asked questions

Which is better for a small property firm? Copilot, in most cases, because it works inside the apps your team already lives in and adoption is the thing that fails. The exception is a firm whose work does not sit in Microsoft, or one wanting a small number of people doing heavy analytical work, where a general model gives more for the money.

Where does Claude fit? As the second tool, for the people doing the heavy reading and writing: lease work, long reports, analysis, anything where the quality of the draft matters more than the convenience of where it appears. It is what I use every day. Buy it on Team rather than individual Pro seats, so your content sits under the commercial terms.

Do we need both? Not to start. Many firms arrive there, with Copilot broadly and a handful of Claude or ChatGPT seats for the people who push hardest. Start with one, learn what your team does with it, and add the second when someone makes a specific case.

Is Copilot included in our Microsoft subscription? Microsoft 365 Copilot is a paid add-on. A separate, more limited Copilot Chat is described by Microsoft as "available at no additional cost for all Microsoft Entra account users with an eligible Microsoft 365 subscription" (Microsoft). What is included has changed more than once, so check your own tenant rather than an article.

Is it safe to put client information into any of them? On an approved business tier with the policy written down, firms do it routinely. On a personal free account, no. The difference is the tier and the policy, not the brand.

What if IT has only cleared Microsoft? Then start with Copilot and stop fighting it. The skills transfer, and a tool your team is allowed to use beats a better one they are not.

How do we know it was worth the money? Decide the measure before you buy. Hours back on a named job, response times on enquiries, how long a report takes. Firms that skip this end up arguing about a licence renewal with nothing but opinions.

If you want help picking, and more usefully getting your team to the point where the licence pays for itself, book a session.

Sources. Microsoft: Copilot pricing (UK); Copilot enterprise pricing (UK); Business plans with Copilot (UK); Copilot product page. Anthropic: Claude pricing; Commercial Terms of Service; Enterprise plan; Claude for Microsoft 365; Microsoft 365 connector; connector security guide; working across Office apps; model training and your data; data retention; regional compliance; Claude in Chrome. OpenAI: ChatGPT pricing. All prices quoted as published on 12 August 2026. Microsoft prices exclude VAT; Anthropic prices are in US dollars and exclude applicable tax.

Copilot, ChatGPT or Claude for a property firm: which one to buy