Authoring SOPs stayed painfully manual
Strong repository tooling does not remove the labour of turning onboarding packs, clinical pathways or process notes into crisp procedures — the bottleneck simply moves upstream.
SOPHIA: turn raw documentation into governed SOPs — with AI humans can trust.
Carradale Futures' SOPHIA platform already gave organisations a central home for policies, SOPs and compliance evidence. The missing piece was the authoring burden: producing consistent, inspection-ready procedures from messy source material. We designed and delivered an AI module on Azure that drafts SOPs, asks for missing clarifications, suggests improvements, and hands outcomes back into SOPHIA's existing workflows — without forcing teams through yet another disconnected tool.
Carradale Futures
SOPHIA platform
Azure-native integration
Phased, test-driven
Draft
SOPs generated from uploaded source docs
Prompt
Clarification questions when input is incomplete
Improve
Suggestions for steps, controls and gaps
Integrate
Native experience inside SOPHIA
The Client
A governance platform built for inspection-ready operations — now faster to feed.
SOPHIA centralises policies and standard operating procedures with versioning, reminders and proof of compliance — so frontline teams stop hunting through PDFs and leaders get visibility into what is actually followed. While Carradale started in healthcare, the same friction exists anywhere accuracy and audit trails matter; SOPHIA is positioned as the smarter home for governed documentation across regulated sectors.
FatFish partnered with Carradale to integrate AI into that core product story: not a novelty chatbot, but a workflow that respects how SOPs are written, reviewed and published. Read the short overview on FatFish Digital · Explore Carradale Futures.
The Challenge
Make SOP creation intelligent — without trading away precision or control.
Consistency and domain nuance both matter
Automating the wrong detail is worse than doing nothing. The system had to surface questions, propose improvements, and stay inside Carradale's accuracy bar — not spray generic text.
It had to belong inside Azure and SOPHIA
A bolt-on demo would never ship. Carradale needed documented APIs, secure authentication and a path from prototype to production that did not rewrite their platform overnight.
What the module returns
From messy inputs to a reviewable first draft.
Users upload real artefacts — manuals, pathway notes, checklists — into an authenticated flow. The AI analyses the material and returns an ordered response: a structured SOP draft, explicit clarification requests where detail is missing, and improvement suggestions that authors can accept, reject or edit before sign-off.
- Automatic generation from raw documentation
- Prompts when the model needs human judgement
- Improvement suggestions rather than silent guesswork
- Final SOP only once open gaps are closed
AI module · document run
First-draft SOP
Structured steps, roles and controls translated from the uploaded source material.
Clarifications
Targeted questions where the model cannot responsibly infer policy or intent.
Improvement ideas
Optional enhancements — extra checkpoints, sequencing fixes, missing hand-offs.
Final draft
Completed procedure once gaps are resolved, ready for human review and publication.
Illustrative of the module outputs described in Carradale's engagement; live behaviour is tuned per tenant and governance rules.
Our Approach
Prototype with conviction. Integrate without drama.
FatFish deliberately sequenced the work into two phases: prove the AI behaviour in a controlled portal, then embed the same engine behind SOPHIA's native screens — mirroring the story on our Carradale · SOPHIA write-up.
01
Modular delivery
We sliced the programme into coherent increments so Carradale could validate quality early, without committing to a monolithic release.
02
Security first
Authentication, tenant isolation and Azure-aligned hosting were treated as product requirements from day one — not a hardening pass at the end.
03
Test-driven iteration
Representative documents exercised the module continuously: edge cases in clinical language, multi-step pathways, and incomplete inputs.
04
APIs with a paper trail
Documented interfaces between the AI engine and SOPHIA mean future features slot in predictably — integration is an asset, not folklore.
How we shipped it
Prove on Azure. Then productise inside SOPHIA.
The two-phase structure deliberately reduced delivery risk: Carradale could validate outputs and guardrails before we wired the module into live customer journeys. That same modularity is what makes ongoing evolution tractable as SOPHIA adds new sectors and intelligence features beyond core repository capabilities.
Secure test portal
Authenticated uploads to an Azure-hosted module: draft SOPs, clarifications and improvement lists returned for Carradale teams to stress-test before productisation.
SOPHIA integration
The same engine embedded directly into SOPHIA. Authors upload, iterate and publish final procedures without leaving the platform; documented APIs connect engine to existing architecture.
Who it helps
Governance that teams will actually maintain.
Faster path from source material to signed SOPs.
Authors spend cognitive effort on judgement and clinical accuracy — not retyping structure that a model can draft and humans can refine.
Consistency at scale across sites and roles.
When hundreds of procedures must align to the same standards, semi-automated generation plus explicit clarification prompts beats ad hoc Word templates.
An integration that is maintainable, not mythical.
Clear APIs between the AI service and SOPHIA mean new UI flows or batch jobs can reuse the same engine without reverse-engineering prompts from a demo.
Governance tooling that keeps pace with change.
Interactive SOPs, reminders and analytics only help if the library stays current. AI-assisted authoring lowers the cost of staying inspection-ready.
Outcomes
AI-assisted SOP development inside the product customers already trust.
Carradale now ships an integrated capability that turns documents into structured, high-quality SOPs in one place — replacing a slow, error-prone manual loop with something that scales as SOPHIA expands beyond its healthcare roots into other regulated environments.
Singleworkflow
Generate, clarify, improve and publish without exporting content into side channels.
FasterSOP turnaround
Manual rewrites give way to draft-first authoring; humans approve rather than originate every line.
Lowerdelivery risk
Phased delivery with a dedicated test harness meant Carradale could validate quality before native integration.
Alignedwith Azure
The module sits naturally inside the ecosystem SOPHIA already operates on — security and procurement narratives stay coherent.
Risk & governance
Automation that respects regulated documentation.
In governance software, a flashy draft that drifts one clause out of alignment is a failure mode. The module is built so accuracy, domain expertise and auditability stay in Carradale's control — AI accelerates drafting; humans retain accountability for what ships.
Semi-automated, not autonomous
The product is designed around human review: clarifications exist precisely because the model must not silently invent policy.
Tenant-appropriate controls
Authentication and data handling align with how enterprise customers expect SOPHIA to behave in production — especially where records may be sensitive.
Traceable integration
Documented APIs make it clear how prompts, documents and outputs move between services — essential when IT and procurement ask sensible questions.
Quality gates before scale
The isolated portal phase was the governance vehicle: Carradale could prove value and tune behaviour before exposing the module fleet-wide.
Next steps
Shipping AI inside a product customers rely on?
We design modular AI capabilities for regulated platforms — Azure-native engineering, clear APIs and governance that procurement teams can defend. Book an intro call or start with our digital readiness audit.