Availability: In development
Care plan assurance
Cross-check assessments, needs, risks, outcomes and interventions, and identify missing or inconsistent care-plan content.
- For:
- Registered managers
iCura Guardian AI
Continuous Assurance. Explainable Intelligence. Better Care.
Guardian AI is the artificial-intelligence assurance layer within iCura. It is designed to analyse operational evidence to identify potential quality, safety and compliance issues requiring professional attention.
Availability: In development
Guardian AI is in development and is not yet available to customers. Speak with us about its development and timescales. Status reviewed 15 October 2026.
The idea
Assurance as an ongoing activity
Quality assurance works best when information is reviewed regularly over time, rather than gathered in a rush before a deadline or inspection.
Guardian AI is designed to support that ongoing habit: helping teams return to their operational information, notice what has changed and keep a record of how they responded. How often information is refreshed depends on how a deployment is configured.
Operational evidence
The evidence Guardian AI is designed to analyse
Guardian AI is designed to work across the operational evidence held in iCura, looking for potential quality, safety and compliance issues that need professional attention.
- Care plans
- Assessments
- Safeguarding
- Medication
- Workforce
- Visits
- Incidents
- Documents
- Governance
Assurance model
From requirement to human review
Every finding follows the same path: it starts from a defined requirement, is explained with reference to the evidence, and ends with a decision made by a person.
Proposed design. Guardian AI is in development; this model describes how it is being designed to work, not a released feature.
Stage 1: Requirement
The quality, safety or compliance expectation being checked, such as a policy or standard.
Stage 2: Evidence
The relevant operational records held in iCura.
Stage 3: AI analysis
The evidence is examined against the requirement.
Stage 4: Finding
A potential issue or gap that may need attention.
Stage 5: Risk
An indication of how significant the finding may be, to help prioritise review.
Stage 6: Explanation
Why the finding was raised, linked back to the evidence behind it.
Stage 7: Recommended action
A suggested next step for staff to consider — not an instruction.
Stage 8: Human review
An authorised professional checks the evidence and decides what, if anything, to do.
Within iCura
The intelligence and assurance engine underneath iCura
Guardian AI is not a standalone chatbot. It is designed to reason across iCura’s operational domains and surface its findings in the iCura modules teams already use, with people deciding what happens next.
Stage 1
Operational data
Care, workforce, medication, visit and governance records in iCura.
Stage 2
Guardian AI
The intelligence and assurance engine underneath iCura.
Stage 3
Findings and recommendations
Explained, evidence-linked findings with proposed responses.
Stage 4
iCura modules
Surfaced where teams work: iCura 360 for governance and compliance, Core, WorkOps, eMAR and Operational Intelligence.
Stage 5
Human action
Authorised staff review, approve and act.
Core capabilities
Three capabilities at the heart of Guardian AI
Together these give Guardian AI its distinct identity — a continuous assurance engine rather than an AI compliance assistant. All three are in development.
Guardian Assurance Graph
Connects people, needs, risks, care plans, interventions, staff competency, visits, incidents, regulation and evidence, so gaps can be found across records.
Guardian Compliance Score
A dynamic assurance and readiness score from individual service user to organisation level. An internal measure to guide attention — not a regulatory rating.
Corrective Action Engine
Turns findings into actions with owners, priorities, deadlines and escalation, approved by authorised staff.
Guardian Assurance Graph
Connecting the evidence, not just reading documents
Rather than analysing documents one at a time, the Guardian Assurance Graph is designed to build relationships between people, their needs and risks, the care delivered, the staff involved, and the regulations and evidence that apply. That lets Guardian AI ask questions that span records.
Connected relationships
- Person
- Need
- Risk
- Assessment
- Care plan
- Intervention
- Medication
- Staff competency
- Visit
- Outcome
- Incident
- Review
- Regulation
- Evidence
Questions it is designed to ask
- “This need exists — where is the required evidence?”
- “This risk increased — was the care plan reviewed?”
- “This medication changed — were the relevant risk assessment and staff instructions updated?”
- “This incident occurred twice — was an intervention changed?”
- “This staff member is scheduled tonight — are they competent for the needs of everyone they will support?”
Illustrative example
How a finding could be raised and resolved
An illustrative scenario showing the intended design. It is not a real case and does not describe a released feature.
A service user has epilepsy.
Guardian AI recognises the care requirement.
Checks the assessment.
Checks the epilepsy risk assessment.
Checks the care plan.
Checks the emergency protocol.
Checks the medication plan.
Checks the competency and training of assigned staff.
Finding: Finds that two assigned staff members do not have current epilepsy competency evidence.
Shows the supporting records and explains the finding.
Creates a proposed corrective action.
Human decision: The registered manager reviews and approves.
iCura 360 updates the assurance status.
Planned capability map
Guardian AI capability roadmap
Guardian AI is in development, and every capability below is planned — none is yet available to customers. Capabilities are grouped into three phases; phases describe sequence, not committed dates.
Phase 1Core assurance, explainability and human approval
- Availability: In development
Risk and safeguarding intelligence
Detect emerging risks, unresolved safeguarding concerns and inconsistencies across records, for professional review.
- For:
- Safeguarding leads
- Availability: In development
Compliance intelligence
Continuously compare operational evidence against policies, controls and regulatory requirements.
- For:
- Quality and compliance teams
- Availability: In development
Inspection readiness
Maintain a live readiness status and identify missing evidence before an inspection.
- For:
- Registered managers
- Availability: In development
Evidence mapping
Show which records support each compliance finding.
- For:
- Quality and assurance teams
- Availability: In development
Explainable findings
Explain why something was flagged, the evidence used and the recommended response.
- For:
- All reviewers
- Availability: In development
Human approval
Route significant or high-risk findings to authorised staff before any action is taken.
- For:
- Authorised approvers
- Availability: In development
Document intelligence
Read policies, assessments, uploaded documents, forms and evidence so their content can be checked.
- For:
- Quality and compliance teams
Phase 2Domain assurance, scoring and corrective action
- Availability: In development
Medication assurance
Analyse MAR/eMAR records, medication plans, missed doses, audits and related risk evidence.
- For:
- Medication leads
- Availability: In development
Workforce assurance
Match training, competency, DBS checks and credentials, staffing and skills against the needs of the people supported.
- For:
- Workforce coordinators
- Availability: In development
Visit assurance
Compare commissioned and planned care with actual visit evidence, and identify delivery gaps.
- For:
- Operations teams
- Availability: In development
Incident intelligence
Detect incident patterns, recurrence and changes in severity, and links between incidents and care plans.
- For:
- Quality and assurance teams
- Availability: In development
Quality trend detection
Identify deteriorating patterns across services, aiming to surface them earlier than conventional KPIs.
- For:
- Provider leaders
- Availability: In development
Core capability
Guardian Compliance Score
A dynamic assurance and readiness score, from individual service user through service and location to organisation level. An internal assurance measure, not a regulatory rating.
- For:
- Provider leaders
- Availability: In development
Core capability
Corrective Action Engine
Turn findings into actions with owners, priorities, deadlines and escalation workflows.
- For:
- Registered managers
- Availability: In development
AI evidence packs
Assemble timestamped inspection and audit evidence packs for people to check before use.
- For:
- Quality and compliance teams
Phase 3Prediction, cross-service insight and organisation-wide oversight
- Availability: In development
Predictive risk
Forecast likely compliance, staffing, medication or service-delivery problems so they can be reviewed early.
- For:
- Provider leaders
- Availability: In development
Cross-service intelligence
Detect patterns across multiple locations and provider groups.
- For:
- Group and regional leaders
- Availability: In development
Automated low-risk actions
Complete pre-approved, low-risk administrative tasks automatically, within governance boundaries set and monitored by the organisation.
- For:
- Governance leads
- Availability: In development
Executive Guardian Command Centre
An organisation-wide live view of risk, compliance, assurance and intervention priorities.
- For:
- Executives and boards
Explainability
What explainability means here
These are design goals for Guardian AI. They describe what we are working towards, not a list of released features.
Source links
Findings should point back to the records they are based on.
Context
Related information should be shown alongside a finding, not stripped away.
Timestamps
People should be able to see when the underlying information was recorded.
Limitations
Known gaps in coverage or confidence should be stated, not hidden.
Review records
Who reviewed a finding, and what they decided, should be recorded.
Human oversight
People remain accountable
AI-generated information can be incomplete or incorrect. Authorised people must review relevant evidence and decide what action to take.
Significant or high-risk findings are designed to be routed to authorised staff for approval before any action is taken. Any automation is limited to pre-approved, low-risk administrative tasks within boundaries the organisation sets and monitors.
If something looks wrong, staff should rely on the underlying records and their organisation’s usual procedures, and tell us so we can investigate.
Intended use
Limits and intended use
- Guardian AI does not guarantee regulatory compliance or any inspection outcome, and we make no claims about prediction accuracy.
- It is not intended for clinical decision-making, and we make no claims about clinical suitability.
- It does not replace safeguarding procedures or the decisions of qualified professionals.
- The Guardian Compliance Score is an internal assurance measure to help prioritise attention. It is not a regulatory rating and does not predict inspection results.
- Predictive features are intended to prompt earlier human review, not to make decisions.
- Current status: Guardian AI is in development and is not yet available to customers. Speak with us about its development and timescales.
Discuss Guardian AI
Tell us about your assurance processes and what would help your teams. We’ll explain where Guardian AI is in its development and what that means for you.