Boost My Business AI Innovation Limited · Trade Licence CL11954 · DIFC, Dubai · gaurav@boostmylocalbusiness.ai

Security and governance posture

Written for the person whose job is to say no.

This page describes the standard we apply to every enterprise and mid-market build: how data is bounded, how access is controlled, how actions are approved and recorded, and how you exit. It is a posture statement, not a marketing page. The full security pack, with named suppliers, data-flow schedules and policies, is provided during engagement under confidentiality.

Boost My Business AI Innovation Limited · Trade Licence CL11954 · DIFC Innovation Hub, Dubai International Financial Centre.

Approval gate Policy enforced · Not bypassable by the system
Active
Held · Awaiting a named approver

Send proposal to external counterparty

Prepared by the document agent. External send requires approval from a person holding the Commercial Approver role. The system cannot approve its own work.

PolicyAnything that sends externally, commits money or changes a record that matters stops here first.
Illustrative view. Gates are policy, enforced in the platform, logged in the action record.

Data residency

The four boundaries.

A client-owned cloud does not automatically mean all processing stays there. If a system sends prompts, documents, voice or metadata to an external processing endpoint, that information has left the boundary, even when the application and database remain inside. So we refuse to make one blended residency claim. Every architecture we propose discloses four boundaries separately.

BoundaryThe question it answersOur standard
1. StorageWhere does data at rest live?In your environment or required region, stated per data type, with encryption at rest.
2. ProcessingWhere is data actually processed, including by AI models?Stated per processing path. Where the jurisdiction requires it, processing is routed to services that can be hosted inside that jurisdiction. No blanket claims.
3. Logging and monitoringWhere do logs, traces and telemetry go?Disclosed and bounded. Logs follow the same residency rules as the data they describe.
4. Support accessFrom where can support staff reach the system?Disclosed, role-limited and logged. Access can be restricted to named individuals where required.

Every proposal includes a written data-flow schedule: what data enters the system, where each copy is stored, what service processes each field or document, which country or region may receive it, which subcontractors can access it, retention, deletion, and which actions require human approval.

The ownership model

Your environment, your keys, your billing.

For private deployments, the environment is provisioned inside your own cloud account, in your required region. This is not a roadmap item. It is the architecture our AXON product runs in production today, one isolated environment per client.

IsolationA dedicated environment for your organisation alone. No shared tenancy, no pooled data, no cross-client anything.
KeysYou hold the encryption keys. We design so that revoking us is possible and survivable.
BillingCloud and model consumption is billed to you directly, at cost, with no markup and a meter you can open at any time. Our margin is the build and the management, never a hidden spread on your usage.
IdentityYour company sign-in, role-based access, service identities and least-privilege permissions. Access rights map to your organisation chart, not ours.
ExitDocumented export, handover and transition assistance in every contract. You own your data and your configuration. A system that traps you is a system we consider badly built.

Approval rules

Six autonomy levels. You choose, per workflow.

Autonomy is not a feature to maximise. It is a dial to set deliberately, workflow by workflow, in writing. Every system we deliver operates at a defined level with a defined escalation path. Try the dial.

The system may

The system may not

In practice:
LevelWhat the system may doExamples
0 · ObserveRead and report onlySummaries, analytics, opportunity detection
1 · RecommendPrepare a decision for a personSuggested reply, risk flag, proposed next step
2 · DraftCreate work, but not send or commitDraft email, report, CRM update, document
3 · Act with approvalExecute after an authorised person approvesSend a proposal, change a booking, create a supplier record
4 · Act within limitsExecute routine actions inside explicit rules and thresholdsConfirm a standard appointment, route a ticket, update an approved field
5 · Extended autonomyComplete a sequence with supervision, limits and exception handlingA narrow, mature workflow with proven controls and a shutdown route

Financial movement, legal commitments, medical decisions, staff decisions, access changes and material customer remedies stay at levels 1 to 3 unless specialist governance and formal approval support more.

Operating controls

What runs around every system.

Action logEvery important action is recorded: what was done, by which agent, on whose approval, with what data. The log is exportable for audit and carries retention rules your compliance team defines.
Approval gatesHard gates before anything that sends externally, commits money, or changes a record that matters. Gates are policy, not politeness: the system cannot bypass them.
Confidence and escalationOutputs carry a confidence score. Low-confidence cases route to a named person. The escalation path is designed with you and documented.
Quality evaluationAn evaluation harness measures whether the system is still correct, whether quality has drifted, and what the failure rate is. Performance is measured, not assumed.
Monitoring and alertsCost controls, caps, anomaly alerts and a dashboard your team can read without calling us.
The stop controlA single control that halts the system. It belongs to you.
Incident responseA documented incident process with defined severities, escalation and communication. Incidents are reported to you, not managed around you.
EncryptionEncryption in transit and at rest as standard. Key arrangements per the ownership model above.
Change controlControlled releases with regression testing. Domain rules are version controlled, so you can see what changed, when, and why.

Disclosure

What we tell you, and when.

Suppliers and subprocessorsFully named in contracts and security documentation, with locations and roles. We do not name technology suppliers in marketing, and we do not hide them from your reviewers.
IntegrationsStated honestly in three lists: live and proven in production, standard and supported, and custom subject to technical review. A connector that exists is not the same as a connector we have run under load.
ClaimsAny figure we publish must be traceable to a source file: the deployment involved, the baseline, the measurement period and the calculation. If we cannot evidence a claim to that standard, we do not make it.
JurisdictionWe operate from the DIFC, under the first data protection regulation written specifically for autonomous systems, and we run the required assessments on our own products first. We are a technology company registered in the DIFC. We are not a regulated financial services firm.
The security packPolicies, data-flow schedules, subprocessor list, architecture diagrams and completed security questionnaires are provided during engagement under confidentiality. Ask for the pack in the first meeting. We bring it unprompted to the second.

For your security review

Send us the questionnaire.

The fastest way to test everything on this page is to send us your standard security and vendor questionnaire, or to bring your CISO to the briefing. We would rather meet the hard questions in the first meeting than in the third.