Skip to main content

Responsible AI

Human accountability before automation

Any automated or model assisted capability associated with PHARMORIS must remain bounded by source provenance, human authority, transparent uncertainty and reviewable decisions.

Discuss the AI boundaryReview clinical governance
Authority | Human decisions remain requiredCurrent boundary | No autonomous clinical decisionsEvidence | No unsupported model performance claims

Responsible use

Automation may assist the record, not assume the authority.

Model assisted functions should support clearly bounded tasks such as structuring information or identifying missing fields only where their behaviour is tested and reviewable.

They must not be presented as clinical judgement, approval, prescribing or a substitute for accountable professionals.

01

Bound the task

Define the exact input, output, user and decision context.

02

Show uncertainty

Do not hide missing evidence, low confidence or source limitations.

03

Require review

Keep progression subject to an authorised human decision.

Control path

Source, suggestion, review and recorded decision remain distinct.

The interface and audit record should make clear what came from a source, what was generated or suggested, who reviewed it and what decision followed.

01

Source

Retain the evidence, rights and provenance behind the input.

Observed
02

Suggestion

Label model assisted output as a proposal rather than a fact or approval.

Generated
03

Human review

Require an authorised person to assess relevance, limitations and next steps.

Governed
04

Decision record

Record the human outcome and the evidence relied upon.

Accountable

No go boundary

Stop when the system cannot explain its source, scope or authority.

No buyer facing claim should imply continuous auditing, bias performance, safety scores or model effectiveness without exact and current evidence.

Human decision requiredProceed only inside an evidenced and governed scope.
01

Missing provenance

Do not progress output that cannot be traced to an authorised source.

02

Hidden uncertainty

Do not present a generated suggestion as a supported fact.

03

Autonomous authority

Do not allow a model to approve, prescribe or make a clinical decision.

Classified public record

Open the classified public record

ANALYSIS + EVIDENCE BOUNDARY

Read the argument. Verify the product separately.

Published analysis can explain why governance matters. It cannot evidence security, clinical authority, certification, product readiness or local approval. Use the dated NEUVIOR records for that.

  1. AI Ethics Now / University of Warwick IATLPodcast appearance

    AI and Accountability: When Does a Suggestion Become a Decision?

    Varun Sharma joins AI Ethics Now to examine false confidence, human responsibility and the point where AI assisted suggestions become accountable decisions.

    Hear where an AI suggestion becomes a decision
  2. TechRadar ProContributed opinion

    Five questions to test whether an AI stack is truly under your control

    In TechRadar Pro, Varun Sharma sets out five tests for enterprise AI control across provenance, infrastructure, legal authority and safe exit.

    Test provenance, control and safe exit
  3. NEUVIOR InsightsNEUVIOR analysis

    From pilot theatre to financial grip: how NHS CFOs should evaluate AI in medicines optimisation

    NHS finance leaders do not need another AI pitch. They need a disciplined way to decide whether a technology can release value within the financial rules they are already being held to.

    See how a finance leader should test an AI claim
Read the governance arguments

NEUVIOR Insights and After Publication reflections are company or founder analysis. Publisher records evidence publication, not endorsement, adoption, product readiness, savings or clinical outcomes.

Make the next step useful

Bring the problem that needs a clearer decision.

Tell us what is stuck, what decision you need to make and what a better outcome would change. We will connect you to the most relevant NEUVIOR, PHARMORIS, media or investor route.

Discuss the AI boundarySpeak with NEUVIOR

QUESTIONS · ANSWERED DIRECTLY

The useful answers, in one place.

Browse every question
01What is NEUVIOR?

NEUVIOR is a pharmaceutical company registered in Scotland. It is building capabilities across medicines intelligence, governed technology and longer term development. PHARMORIS is its immediate product priority.

02What does NEUVIOR build today?

NEUVIOR is developing PHARMORIS, a human-governed medicines operations system in controlled, non-production development. It is intended to connect evidence, decisions, accountable work and value review. No independent TRL, deployment, customer adoption or clinical validation is claimed.

03What is PHARMORIS by NEUVIOR?

PHARMORIS is a configurable medicines operations system under controlled development. It is designed to help authorised teams progress medicine opportunities from evidence linked to its source, through an owned ActionCase, a human decision, delivery and review.

04Who founded NEUVIOR?

Varun Sharma is the Founder, CEO and CTO of NEUVIOR and leads the company and PHARMORIS product architecture.

05Where does NEUVIOR have a presence?

NEUVIOR works across Glasgow, London and Mumbai. Glasgow is the company origin and registered Scottish base. London supports commercial activity led by the founder. Mumbai supports market engagement in India and relationships across the pharmaceutical sector. These are different forms of presence. They do not imply three registered offices or regulated facilities.