Back to Workflow100DATA PROTECTION, CONFIDENTIALITY & AI

We will keep your confidential dataout of AI models.

Your clients trust you with their information. We take that responsibility seriously. Our commitment covers client records, internal documents and sensitive business information, from development and testing through to ongoing support.

We design, build, test and support your software. AI is one of the tools we use where it helps, under our direction and review. We use fictional examples and briefs cleared of confidential information with AI, and build the finished tool to process your live records without sending them to an AI model.

Discuss your data requirements

Confidentiality comes first.

Keeping confidential data out of AI is a requirement throughout discovery, development, testing and ongoing support. Before accessing real information, we agree what is confidential, who may access it and where it may be processed.

We use fictional examples and briefs cleared of confidential information when AI supports our development work.
Access to real client records, internal documents and sensitive business information agreed in advance.
The same confidentiality boundaries apply to fixes, logs, screenshots and ongoing support.

01 / WE BUILD THE TOOLNo confidential data shared with AI

We define the workflowUsing cleared briefs and fictional records.
We design & developWe use AI selectively to support our work.
We review & testWe check the code, permissions and results.

The reviewed software is deployed.
An AI connection doesn’t have to come with it.

02 / USING THE TOOL No data sent to AI

Your business recordsFrom the systems and files you approve.
Your purpose-built toolApplies agreed rules and calculations.
Useful work, completedReports, updates, tasks and documents.

An example of a workflow designed without AI processing live data. Hosting, integrations and any other data destinations are agreed separately.

WHAT THIS CAN LOOK LIKE

Less manual work.
No AI needed in the workflow.

Many time-consuming tasks can be handled with structured data, templates, calculations and clear rules. These are examples we could build around your processes.

SURVEYING & CONSULTANCY

Assemble a report from structured records.

Starts with: Inspection answers, selected photos and an approved template.
The tool handles: Populate sections, calculate totals and flag missing information using agreed rules.
Your team gets: A consistent draft for a qualified person to review.

CORPORATE & PROFESSIONAL SERVICES

Keep tasks and documents moving.

Starts with: Client records, document statuses and dates your team has approved.
The tool handles: Check required fields, track deadlines and route tasks to the right person.
Your team gets: Clear worklists and reminders, with decisions left to your team.

WHOLESALE & DISTRIBUTION

Turn an order into the next steps.

Starts with: Structured order details, an approved price list and stock records.
The tool handles: Calculate prices, check availability and prepare documents through approved integrations.
Your team gets: A draft invoice and fulfilment tasks, without retyping each line.

Rule-based automation has limits. Interpreting arbitrary documents, generating free-form advice or making professional judgements may need a different approach or human input. We make that distinction when scoping the work.

CONFIDENTIALITY BY DESIGN

Agree the boundaries.
Then build around them.

Keeping data out of AI is one requirement. The wider system still needs appropriate protection. These are the arrangements we work through with you before development and launch.

What development tools may receive

Agree which process details and source material can be shared. Use fictional records, and keep confidential material, credentials and client data out of AI prompts, attachments, screenshots and logs.

Where the live data goes

Map hosting, integrations, storage locations and service providers. An agreed environment might be your existing systems or separately approved hosting; “no AI” does not mean “no third parties”.

Who can access it

Define staff permissions, developer and support access, and any confidentiality or data processing agreements needed before access is granted.

How it is checked and maintained

Include code review, security checks and tests of calculations, permissions and outputs. Agree logging, retention, backups and how changes are reviewed.

Clear about what
“without AI” means.

“Not used for training” is a different promise.

An AI service may still receive or retain information even when it does not use it to train a model. Keeping your confidential data out of AI means verifying that it is not sent to an AI service at all.

Your policy applies during development too.

Some organisations restrict AI-assisted development itself, including sharing business rules or source code. We discuss those restrictions before starting. Fictional sample data does not override your policy.

Controls are agreed for your project.

This describes an approach we can design for, not a blanket security or compliance guarantee. Your sector, contracts and applicable data protection requirements still matter. Any later AI feature would need a separate discussion and approval of its data flows.

START WITH THE PROCESS, NOT THE RECORDS

Tell us what the tool needs to do.

For an initial conversation, a description of your workflow and restrictions is enough. Please don’t send client files, confidential records or passwords through the enquiry form.

Talk through your requirements

We can start with fictional examples and agree how any necessary access to real data will be handled.