

Private AI for business, on accounts your firm owns
Private AI means client data only goes into AI your firm controls, under terms that rule out training on it. We build it as a custom system in your own Microsoft or Google account.
4.9+ from 226 reviews
We've built AI systems used by:
- Asahi
- Unilever
- Rentokil Initial
- Freshworks
- BambooHR
The short version, for private AI.
- What it means
- Client data goes into AI under business terms: not used for training, kept no longer than needed, in accounts the firm controls. Personal ChatGPT and Claude logins don't meet that bar.
- What most firms need
- A business AI account in their own Microsoft or Google tenant. Fully local models only pay off when a client contract demands them.
- What we build
- A custom AI operating system connecting your core software to run multi-step workflows, with staff signing off client outputs.
- Where it runs
- In your own Microsoft or Google account, or in ours in the UK, EU or US. Business AI accounts only.
- Price
- Pricing depends on the size of your firm and how many teams we connect. We go through numbers on the intro call and confirm the quote in your AI opportunity audit within 24 hours.
- Timeline
- Intro call, then your AI opportunity audit within 24 hours. We ship the first live workflow with your initial team within 7 days of the intro call, then expand one team at a time.
- Who owns it
- You own the system outright. It runs in your own accounts with full documentation, and your team or another provider can update it at any point.
- Guarantee
- If you don't get an ROI within 90 days, you get a full refund.
What private AI means, from ChatGPT to your own servers.
Private AI spans five levels, from personal ChatGPT logins to models hosted on your own servers. Each step up gives your firm tighter control over client data, alongside higher setup and running costs.
Level 1
Personal AI accounts
ChatGPT or Claude on a free or personal login
Chats can be used to train models unless the user turns it off (OpenAI, Anthropic). The firm can't see what went in.
Cheap and capable, but the firm has no visibility or oversight, making it unsuitable for client data.
Level 2
Business AI plans
ChatGPT Business or Enterprise, Claude Team or Enterprise, Microsoft 365 Copilot
Not used for training by default (OpenAI, Anthropic, Microsoft). OpenAI can store data in the UK for new Enterprise workspaces. Copilot keeps data in your Microsoft 365 tenant, but UK prompts can be processed in the US or EU until UK processing arrives, expected by the end of 2026.
A licence per person with frontier models and minimal upkeep, though staff still have to copy client data into a chat window and paste answers back out.
Level 3
AI through the API
OpenAI or Anthropic models called from your own systems
Not used for training by default. OpenAI keeps API logs for up to 30 days for abuse checks, and offers zero data retention to customers it approves.
You pay for what you use, and AI can work inside your tools. Someone has to build and run that.
Level 4
Your own cloud tenant
Azure OpenAI in Microsoft Foundry, AWS Bedrock, Google Vertex AI
Runs in your cloud account, in a region you pick. The model maker doesn't see your prompts, and they aren't used for training (Microsoft, AWS, Google). UK regions exist, but newer models are often only offered as Global or Data Zone deployments, which can process prompts outside the UK.
The most control you can get with top models. It needs someone to set up and run the cloud account.
Level 5
Local or on-premise AI
Open-weight models such as Llama, Mistral or gpt-oss on your own servers
Never leaves your building or your private cloud.
You buy and run the hardware, open models usually trail the best hosted ones, and you patch and upgrade them yourself.
We build at levels 2 to 4, in your own Microsoft or Google account or in ours in the UK, EU or US. Fully local models make sense when a client contract requires on-premise hosting; for most firms, the hardware, upkeep and model quality trade-offs outweigh the extra isolation.
Secure AI for business, by the rules your firm works under.
No UK regulator bans AI, and their expectations are consistent: know where client data goes, who can access it and how long it is retained. If a vendor cannot answer those three questions, keep client data out of that tool.
No training
Client data isn't used to train models unless you say so.
Short retention
Prompts and answers are kept no longer than needed, and you know for how long.
Your environment
Data stays in accounts your firm controls, away from third parties.
Trained staff
Staff know which tools are approved and which categories of client data they can hold.
- SRA, for law firmsThe SRA's August 2026 warning notice says client information should only go into AI with contractual, technical and organisational safeguards. Paid and free tools can both fall short, depending on their terms and settings. More in our SRA AI guidance summary.
- ICAEW, for accountantsICAEW's November 2026 note says confidentiality duties still apply to AI, and that putting data into a tool outside a secure environment can be akin to posting it publicly. More in our ICAEW AI guidance summary.
- UK GDPR and DPIAsThe ICO lists AI as an innovative technology that can make a DPIA a legal requirement, and says most AI use will need one. A DPIA asks where data goes, who processes it and for how long.
- Data residencyUK GDPR doesn't demand UK-only hosting, but transfers abroad need safeguards, and some clients write UK or EU hosting into their contracts. Check where prompts are processed as well as where they're stored. Our private ChatGPT alternatives comparison says where each tool stores and processes data.
- Client contracts and NDAsConsulting has no sector regulator, so your client contracts set the rules. Check what they say about third parties handling client data before you pick a tool.
Ready for the full system? See the AI operating system
One system your whole firm runs on.
Most firms have AI built into a few separate apps while staff still copy information between them. An AI operating system connects those systems together so workflows run from start to finish with full client context.
- Connects your existing tools
Outlook
Teams
HubSpot
Xero
SharePoint and the rest of your stack. Your team keeps working inside the apps they already open.
- Shared memory across the firmYour client records, templates, house rules and procedures live in one shared layer that your workflows draw on.
- Built-in human sign-offThe system handles filing, CRM updates, first drafts and chasing. A member of your team reviews and approves client-facing work before it goes out.
Kept private, because it runs in your accounts.
We build a custom AI operating system, one team at a time. It connects the tools you already use, and client data stays where you decide.
Your own accounts
Systems run in your own Microsoft or Google account, or in our UK, EU or US hosting, keeping client data in the region you choose.
Business AI accounts only
Business accounts don't train on your data, and you choose which clients, matters and folders the system can access.
A person signs off
AI handles routine drafting and admin, while your team approves client-facing outputs before sending.
Documented, and yours
We document data flows as we build for your data protection lead. If you stop working with us, you keep the full setup.
Built for your sector.
We build each system around the software, regulations and workflows of your sector. Explore how it works for law, accountancy and consulting firms.
- AI operating system for law firmsConnects Clio, Actionstep, LEAP, iManage, NetDocuments and Microsoft 365. Built around the SRA's rules, with a fee earner signing off.Read more
- AI operating system for accountantsConnects Xero, QuickBooks, IRIS, Karbon, TaxDome, Dext and Microsoft 365. A qualified person signs off, and your MLRO makes the AML decisions.Read more
- AI operating system for consulting firmsConnects HubSpot, Teams, SharePoint, Google Workspace, Notion and Slack. Built around UK GDPR and your client NDAs.Read more
A $24M-a-year business runs on a system we built.
Floor Authority installs flooring for national home builders. Quotes, orders, invoices and reports all run through a custom system we built around how the business operates.
- A year in revenue, running through the system
- $24M
- A year in revenue, running through the system
- Growth in two years, with only a few new hires
- 2x
- Growth in two years, with only a few new hires
- Staff productivity measured by CEO
- +25%
- Staff productivity measured by CEO
“The result has been far better than anything I could do on my own.”

“I don't think any of the other firms that we talked to could have built this.”

Our guarantee
If you don't get an ROI within 90 days, you get a full refund.
Book an intro4.9+ from 226 reviews
Next steps to your AI operating system.
We need an intro call, a one-hour kickoff and a short handover session with each team as their workflows go live.
- 1
Today
Intro call
We look at how your firm operates, which processes take the most manual effort and where tasks stall between systems.
Book an intro - 2
Within 24 hours
AI opportunity audit
We send a written breakdown of where an operating system will save the most time, which team to start with, and the pricing.
- 3
Within 7 days
First deployment
We ship the first live workflow with your pilot team, delivering measurable time savings while we build out the wider system.
- 4
Then
One team at a time
We connect the first team's tools and automate their core workflows, then expand the same system to the next department.

“They think ahead, always deliver on time”
Frequently asked questions.
Have another question? Ask us on the intro call.
How an AI operating system worksTake the free AI readiness assessmentIt's AI your firm can put client data into without losing control of it. That means business terms that rule out training on your data, short retention, and accounts the firm controls rather than staff's personal logins. It can be a business AI plan, AI in your own Microsoft or Google tenant, or models on your own servers.
Often it's still ChatGPT or Claude, on a business plan rather than a personal login. ChatGPT Business and Enterprise, Claude Team and Enterprise, and Microsoft 365 Copilot don't train on your data by default, and admins control who uses them. What matters is the account, the settings and what your team is allowed to put in.
Yes. OpenAI, Anthropic and Microsoft don't train on data from their business plans or APIs by default, and Azure, AWS Bedrock and Google Vertex AI don't use your prompts to train models. Personal accounts are different, as chats can be used for training unless the user switches it off.
Probably not. Local models keep data on your own hardware, but you buy and run the servers, the open models usually trail the best hosted ones, and you patch them yourself. Most firms get what their regulator asks for from business AI accounts in their own tenant. Go local if a client contract demands it.
Yes. The SRA's 2026 warning notice says client information should only go into AI with proper safeguards: a secure environment, no training without permission, and data kept no longer than needed. Personal AI logins don't meet that. It also expects you to stay responsible for the work, protect client confidentiality and supervise how AI is used. We draft an AI policy for your firm that covers this, and that your own COLP can adapt and own.
Yes, in the right tool. ICAEW says confidentiality duties still apply when you use AI, and that a tool outside a secure environment is close to making data public. Use firm-approved business accounts that keep data within the firm, and have a qualified person sign off the work.
Usually. The ICO treats AI as innovative technology and says most AI use is likely to be high risk, which makes a DPIA a legal requirement. It asks where data goes, who processes it and how long it's kept. Because the system runs in your accounts and we document it as we build, your data protection lead has those answers.
Pricing depends on the size of your firm and how many teams we connect. We go through numbers on the intro call and confirm the quote in your AI opportunity audit within 24 hours. We ship the first live workflow with your initial team within 7 days of the intro call, then expand one team at a time. If you don't get an ROI within 90 days, you get a full refund.

Put AI to work on client data, in accounts your firm owns.
Book an intro and we'll send your AI opportunity audit within 24 hours. It shows where AI can save your team time.