Evidence before claims
Important statements are connected to a source, classified honestly and kept separate from assumptions.
Trust and Governance
Genuine proof, clear permissions and human responsibility
Practical AI is useful only when the business can understand the evidence, control important actions and remain responsible for the customer experience.
GDP defines responsibilities, permissions and limitations before capability expands.
Four working principles
Clear boundaries create useful confidence
Permission before action
Access to information does not automatically authorise messages, changes, publication or commitments.
Control grows with capability
The greater the possible consequence, the stronger the testing, approval, access and recovery controls should be.
Human responsibility remains visible
People retain responsibility for consequential judgement, exceptions and the promises made to customers.
Evidence classification
Evidence is clearer when different kinds of knowledge stay separate
GDP does not turn an uncertain observation into a confident claim simply because it sounds persuasive.
Verified fact
Supported by an identified source and checked for the purpose at hand.
Sourced observation
Something directly visible in a named record, page, report or customer feedback source.
Reasonable interpretation
A conclusion drawn from available evidence and clearly presented as an interpretation.
Needs confirmation
A missing, conflicting or time-sensitive detail that should not be guessed.
What that means in practice
An old post tells us what was shared then, not necessarily what you offer today or what you want to do next. A missing public review reply does not tell us whether someone helped the customer privately. And two different addresses might be a mailing address and a shopfront, or reflect a move, rather than an error.
We check the source, date and context before turning a difference into a recommendation. When reviewing customer feedback, we explain which platforms and how much history we could see. We consider all ratings and contrary examples, distinguish an unobserved reply from an unknown reply status, and do not present a small sample as the whole picture.
What we observed, what we think it may mean and what you want remain separate. We bring you neutral questions to resolve what is unclear. Confirming a fact is not permission to change a listing, publish a reply or start paid work.
We can begin with publicly available information. Access to your accounts or private records requires your permission.
Human approval boundaries
Knowing something is not the same as being allowed to act on it
The agreed scope makes responsibility and approval requirements clear.
Before launch, during use and if something goes wrong
An AI assistant should make your business easier to deal with, not leave you wondering what it is doing or who is looking after it.
Before anything goes live, we agree what it should do, what it must leave to a person and how you can stay in control.
Start with a job worth helping with
We begin with your business and the people you serve. What would make their experience easier? Where would your team benefit from some support?
That might mean answering common questions, helping someone understand your services or collecting the details needed for a follow-up. It does not mean handing every enquiry or decision to AI. If a simpler approach would serve you better, that belongs in the conversation too.
We also explain which platforms the solution depends on, what information they need and any relevant limits, costs or data-handling arrangements. Those details are checked for the proposed setup, rather than assumed to be the same for every business.
Test it before customers depend on it
A convincing demonstration is a starting point, not proof that a system is ready.
We agree what needs to work, then test realistic situations. Can the assistant find the right business information? Does it recognise when it does not know? Does it respect the limits on bookings, messages and other actions? Can someone reach a person when they need to?
You review the results and any remaining limitations before approving a live launch. Anyone responsible for overseeing the system also needs clear instructions on using it, recognising problems and stepping in.
Know who is looking after it
Business details change. Platforms change. An assistant that was useful at launch still needs appropriate attention afterwards.
Before launch, we agree who keeps information current, who reviews problems and what ongoing support or monitoring is included. We also agree how to pause or restrict the assistant, keep essential enquiries moving another way and arrange a handover or shutdown if needed.
Support hours, response arrangements and any additional costs are made clear. A system being available around the clock does not mean a person is monitoring it around the clock.
Customers should be able to flag an incorrect answer, raise a concern or ask for a person. Their feedback helps identify what needs checking, correcting or changing.
Privacy and retention
Use only what the work requires
Customer and business information should have a defined purpose, appropriate consent, limited availability and a considered retention period. Information should not quietly become available for unrelated future uses.
GDP begins with least-necessary access and expands it only when a clearly approved capability requires more.
Security proportionate to capability
Security grows with capability
A read-only report does not create the same risk as a system that can contact customers, make bookings or change a public resource. Controls should match the possible consequence.
This may include restricted accounts, narrowly permitted tools, client separation, confirmation of recipients and destinations, activity records and periodic access review.
Business ownership and portability
Your information and deliverables remain business assets
GDP provides clear documentation, copies of commissioned deliverables and available export or handover information. We distinguish information owned by the business from settings or features controlled by a platform.
We do not describe a system as portable when practical recovery or transfer is not supported.
Supplier and platform transparency
You should know what your solution depends on
AI models, messaging services, hosting providers and business platforms can have their own limits, costs, policies and availability. Material dependencies should be disclosed rather than hidden behind GDP branding.
If a supplier changes an important capability, the effect on the business should be assessed before the workflow is presented as dependable.
Activity records, recovery and handover
A useful system should leave an understandable trail
Where the selected platform allows it, important activity should be recorded so the business can understand what occurred, what information was used and whether a person approved the action.
Instructions and configurations should be versioned where practical. Recovery, fallback and handover arrangements should reflect the importance of the workflow rather than being considered only after something fails.
No quiet expansion. New information sources, recipients, actions or authority require their own consideration and approval.
Honest limitations
Useful work without unsupported promises
GDP can improve the quality, consistency and usefulness of business information, customer proof and workflows. We do not guarantee rankings, enquiries or revenue, and we do not claim that every platform will display, distribute or interpret information in the same way.
Unresolved matters and material limitations should remain visible.
Accessibility
Useful systems should remain usable by real people
Public pages and customer experiences should use clear language, readable presentation, meaningful structure and practical alternatives where a particular channel or method cannot be used.
Accessibility needs are considered alongside privacy, security and customer care rather than treated as a decorative final check.
Genuine proof and illustrative advertising
A genuine review must remain genuine when it becomes content. We preserve its meaning, check the source and obtain the required permissions. Authentic footage shows actual people, places and work. Generated material is not presented as proof of an event or result that did not happen.
Clear disclosure of generated material
Generated presenters and illustrative scenes are identified clearly. We check the labelling requirements of each publishing platform before release. A generated presenter is not the reviewer, an employee or a witness to the work.
Customer choice and human support
Customers should know when they are speaking with AI assistance, understand the options available and be able to reach a person. Assistance must not make unsupported promises or treat a customer's enquiry as permission for unrelated marketing.
Proprietary methods
GDP demonstrates practical benefits, safeguards and limitations. Its proprietary methods remain confidential.
See a relevant example before deciding what should proceed
Start with review responses, a video concept or a private customer-assistance demonstration. We agree the preview scope and permitted material before preparing it.
Register your interest in a Free Private Preview. Email GDP through our Contact page.
Register your interest in a Free Private PreviewThe Private Preview is free. Paid production, implementation, ongoing support and third-party platform costs are quoted separately. No paid service begins without your agreement.
Live connections, customer contact and publication require separate approval.