Security & privacy

Your data doesn't need another place to live.

Where practical, Kepler26 builds AI workflows around the enterprise systems, permissions, and governance your organization already trusts.

Secure technical delivery backed by an ISO 27001-certified engineering organization.Kepler26's technical delivery is supported by our technical partner, a secure software engineering and cybersecurity organization.

01Your environment.02Your permissions.03Your governance.04Better workflows.
ISO 27001Certified technical delivery partner
50+Technical delivery organization
Secure softwareDevelopment capability
Regulated industriesEngineering experience
CybersecurityIntegrated when the risk requires it

Technical delivery partner

Enterprise engineering backed by our technical partner.

Kepler26's technical delivery is supported by our technical partner, an ISO 27001-certified software engineering and cybersecurity organization specializing in secure enterprise solutions for regulated industries.

The partnership adds software engineering, solution architecture, secure AI development, cloud, application security, DevSecOps, cybersecurity and identity and access-management depth. Kepler26 remains the client's primary consulting and delivery relationship.

Discuss technical delivery

Preferred architecture

Client-environment first.

Our default question isn't how to move your data into another platform. It's how much we can accomplish inside the environment you've already approved.
01 / UsersYour people
Client-controlled boundaryYour approved technology environment
Microsoft 365SharePointOneDriveTeamsCopilotCopilot StudioAzureChatGPT Business / EnterpriseClaude EnterpriseWrikeVeevaClient databasesClient cloud infrastructureClient API accountsOther approved systems
Kepler26 brings the architectureDesigned AI workflowAgents / integrations / automation / review points
Approved inputsYour data sources
Controlled resultOutputs remain in your environment

Client controls

  • Identity
  • Files
  • Permissions
  • Applications
  • Data
  • Access
  • Retention
  • Governance

Kepler26 provides

  • Workflow analysis
  • AI architecture
  • Agent design
  • Automation
  • Integration
  • Prompt and system design
  • Implementation
  • Testing
  • Governance recommendations
  • Training and optimization

Feasibility depends on the client's licensed capabilities, permissions, approved providers, integration options, and governance requirements.

Implementation models

Different workflows require different architectures.

We start with the lowest-complexity architecture that can support the workflow, information boundary, and level of control required.

01 / Client-native

Client-native

Kepler26 builds the workflow primarily inside technology already owned and governed by the client.

01Client data
02Client systems
03Kepler26-designed AI workflow
04Client systems

Why start here

  • Minimizes unnecessary data movement
  • Uses existing identity and permissions
  • Fits existing IT governance
  • Reduces platform sprawl
  • Keeps institutional knowledge under client control

Secure software development

Security doesn't start after the build.

For custom applications and integrations, security is considered from architecture through deployment and operation. The controls included in an engagement depend on the system, risk and agreed scope.

01

Architecture

  • Security requirements
  • Threat considerations
  • Trust boundaries
  • Information flows
  • Permission model
02

Development

  • Secure coding
  • Code review
  • Dependency management
  • Secrets management
  • Environment separation
03

Testing

  • Functional and integration testing
  • Access testing
  • Failure scenarios
  • Security testing as appropriate
04

Delivery

  • Controlled deployment
  • CI/CD
  • Environment configuration
  • Role-based access
  • Logging
05

Operations

  • Monitoring
  • Access lifecycle
  • Incident handling
  • Maintenance
  • Vulnerability management where applicable

Specialist capability

Security expertise when the architecture needs it.

Not every engagement needs a separate cybersecurity workstream. Kepler26 can bring the appropriate specialists into architecture, engineering and review when the information, integration or risk profile requires it.

Information boundaries

Give AI only what the workflow needs.

Connecting a model to an entire repository because it is technically possible usually creates more exposure, more noise, and weaker retrieval.

Brand X MLR support may need
  • Brand X prescribing information
  • Approved claims and references
  • Brand style guidance
  • Relevant SOPs
  • Historical review decisions
It probably does not need
  • HR or finance files
  • Other clients
  • Unrelated brands
  • General corporate material

Better context boundaries can improve both security and AI performance.

Least privilege

Agents shouldn't have more access than they need.

Project-specific, folder-specific, read-only, time-limited, and role-based access can reduce the consequences of a mistake while making the workflow easier to understand and review.

  • Limited API scopes
  • Named users instead of shared accounts
  • Access expiration when projects conclude

Entire company repository

Relevant brand and project information

Agent autonomy

Automate the work. Keep humans accountable for the decisions.

Autonomy should be assigned task by task. The more consequential the action, the clearer the human approval point should become.

01

AI can often do the mechanics

Bounded automation lets AI perform a clearly defined, inspectable task with approved inputs and rules while people review exceptions and outcomes.

  • Search and retrieve
  • Summarize and extract
  • Classify and compare
  • Organize and draft
  • Identify inconsistencies
  • Flag potential issues
02

People review consequential work

A person should normally review work before it changes an external, important, or regulated state.

  • Send external communications
  • Publish or distribute materials
  • Update important records
  • Move important documents
  • Submit deliverables
  • Modify client systems
03

Authorized people own decisions

AI can prepare evidence and checks. It should not quietly become the accountable decision-maker.

  • Final medical approval
  • Regulatory approval
  • Promotional claim approval
  • Permission changes
  • Deletion of sensitive information
  • Other consequential regulated decisions
AI can accelerate review. It should not quietly become the accountable decision-maker.

Data classification

Different information deserves different handling.

This example is a conversation tool, not a universal policy. Kepler26 adapts workflow boundaries to each client's classifications and governance.

01

Public

Examples

  • Published papers
  • Public congress information
  • Prescribing information
  • Public websites

Example handling

Broadest approved AI usage, subject to source quality and intellectual-property requirements.

PHI and PII

Don't give a workflow personal data it doesn't need.

Literature surveillance, congress monitoring, competitive intelligence, claims support, annotation, MLR preparation, brand planning, knowledge management, and content development often do not inherently require identifiable patient information.

Client isolation

Client information stays separated.

Custom systems should prevent one client's documents, embeddings, prompts, outputs, logs, databases, or agent memory from becoming available to another client.

Client ABrand ABrand B
Client BBrand C
Client CBrand D

Safer prototyping

Prototype with dummy data.Deploy with real data.

Where practical, early development can use synthetic documents, anonymized examples, public prescribing information, published research, fictional brands, or fabricated test data. Production can then move into the client's approved environment.

Clients do not necessarily need to provide large confidential datasets just to begin.

Traceability

AI workflows should leave a trail.

Appropriate workflows should make it possible to understand who initiated an action, what it accessed, when it ran, what it generated, which system it used, whether review occurred, and what final action was taken.

We design around the observability available in the chosen client technology stack.

Provider governance

Not every AI tool belongs in a regulated workflow.

An approved-provider framework turns abstract policy into practical architecture decisions before a workflow reaches production.

QuestionWhy it matters
Who provides the model?Vendor accountability
Is customer data used for training?Confidentiality and permitted use
What is retained?Data lifecycle
Where is information processed?Jurisdiction requirements
Who are the subprocessors?Supply-chain visibility
How is access controlled?Unauthorized-access risk
What is logged?Auditability
Can data be deleted?Lifecycle control
What happens when models change?Validation and governance

Kepler26 can help translate these requirements into workflow, provider, integration, permission, and review decisions.

Permission readiness

AI can expose permission problems that already exist.

Connecting AI to SharePoint, OneDrive, Teams, or another enterprise repository can make existing information dramatically easier to discover. That makes access boundaries part of AI readiness.

  • Who can access what today?
  • Are historical folders over-shared?
  • Do permissions reflect current roles?
  • Are client and project boundaries enforced?
  • Should every repository be indexed?
Starting engagement

AI Workflow & Readiness Assessment

Identify high-value workflows, determine the information each one needs, assess available systems, and recommend a practical implementation architecture.

  • Identify intended AI workflows
  • Map required information sources
  • Review access boundaries
  • Identify unnecessary data exposure
  • Recommend project-specific knowledge boundaries
  • Define human approval points
  • Identify provider requirements
  • Document a recommended implementation architecture
Assess AI readiness

Before development

AI Workflow Architecture & Security Review

Give business, technology, privacy, security, legal, and procurement stakeholders a concrete architecture to evaluate before implementation.

  1. 01Current-state architecture
  2. 02Proposed solution architecture
  3. 03Workflow map
  4. 04Data-flow diagram
  5. 05Application boundaries
  6. 06Model and provider inventory
  7. 07API and integration map
  8. 08Authentication approach
  9. 09Permission model
  10. 10Agent scopes
  11. 11Human approval points
  12. 12Retention considerations
  13. 13Audit and logging requirements
  14. 14Security and privacy considerations
  15. 15Deployment architecture
  16. 16Implementation plan
  17. 17Engineering estimate

How we approach our own operations

Say what is in place. Do not imply what is not.

For this public website, analytics stays off until a visitor opts in, gated resources use passwordless access, application secrets are supplied through deployment environment variables, and the contact form asks visitors not to send confidential information.

Those website practices are not certifications and do not define a client implementation. Engagement-specific providers, access, retention, review, and documentation requirements are agreed for the work being designed.

Questions stakeholders ask

Security, privacy, and implementation FAQ.

01Is Kepler26 ISO 27001 certified?

Our technical partner is ISO 27001 certified. The certification applies to our technical partner and the scope of its certified information-security management system. Kepler26 does not represent the Mody & Associates or Kepler26 legal entity as independently ISO 27001 certified.

02Who actually builds Kepler26 solutions?

Kepler26 leads the engagement and assembles the technical team required for the solution. Depending on the project, that can include software and AI engineers, solution architects, cloud specialists, cybersecurity specialists, DevSecOps engineers, and identity and access-management experts from our technical delivery organization.

03Are you only an AI strategy consultancy?

No. Kepler26 works from workflow diagnosis and solution architecture through software development, integration, secure deployment, adoption, and ongoing optimization. The work can start small without ending at a recommendation or prototype.

04Can you support larger implementations?

Yes. The delivery model can scale from a focused workflow automation to a larger custom system or enterprise integration. Scope, architecture, staffing, and delivery stages are defined around the implementation rather than forcing every project into the same team structure.

05Will your technical partner work directly with us?

Kepler26 remains the primary engagement relationship and coordinates the technical resources required for delivery. Specialists from our technical partner can participate directly in architecture, security, and implementation discussions when useful while Kepler26 retains solution ownership and client leadership.

06Where will our data live?

Data location is determined by the approved architecture for the engagement. Where practical, information remains in client-controlled systems, regions, providers, and access boundaries. Team location and data-processing architecture are separate considerations and should not be treated as the same decision.

07Does Kepler26 need to host our files?

Not necessarily, and often no. Where practical, Kepler26 designs workflows around systems the organization already controls, such as Microsoft 365, SharePoint, approved AI platforms, cloud environments, and internal knowledge repositories.

08Will our employees need another AI platform?

Not always. One of our first questions is how much can be accomplished using technology the organization already owns and has approved.

09Can you work inside our Microsoft environment?

When the intended workflow, available capabilities, client approval, and appropriate access support it, workflows can be designed around Microsoft 365, SharePoint, OneDrive, Teams, Copilot, Azure, and related systems. Feasibility is confirmed for each implementation.

10Can you work with our enterprise ChatGPT or Claude environment?

Where the organization's plan, permissions, available integrations, and governance policies support the intended workflow, Kepler26 can design around approved enterprise AI environments.

11Do you need access to all our data?

No. We generally prefer the opposite approach: determine the minimum information required for the workflow and limit access accordingly.

12Can AI approve MLR materials automatically?

AI can accelerate quality checks, evidence review, consistency checks, annotation, claims analysis, and preparation for review. Final approval decisions should remain with appropriately authorized human reviewers.

13What about PHI or sensitive personal information?

If a workflow doesn't require restricted personal information, it should not receive it. Workflows involving PHI, PII, or other restricted information require additional architecture and governance review.

14Will you use our confidential information to train your own models?

Kepler26 does not design client engagements around pooling confidential client information into a shared training dataset. Data handling requirements, model-provider terms, retention, and permitted uses should be explicitly defined for each implementation.

15Can our IT or security team review the architecture?

Absolutely. Kepler26 expects security, privacy, IT, legal, or procurement stakeholders to be involved when the workflow warrants it. We can document proposed data flows, providers, permissions, integrations, and human-control points for review before production deployment.

16What happens when the engagement ends?

Access should be reviewed and removed when it is no longer required. Project-specific retention or deletion requirements should be agreed as part of the implementation rather than assumed.

Bring the right people in early

Bring enterprise stakeholders in before production.

Complex implementations often involve IT, information security, enterprise architecture, procurement, legal, privacy, compliance, medical, regulatory and MLR stakeholders. Kepler26 can document and walk those teams through the workflow, data sources, providers, access model, AI boundaries and deployment architecture before production.

Discuss your requirements Security review and architecture documentation can be incorporated into the engagement.