Vertical SaaS · Workflow & compliance
Calltinuity
A workflow operating system for UK mortgage advisers — turning client conversations into structured, defensible case files in minutes, with the adviser always in control.
Documentation infrastructure for UK mortgage advisers — turning regulated client conversations into structured, auditable case files.
- Client
- Calltinuity
- Industry
- Vertical SaaS · Workflow & compliance
- Year
- 2026
- Role
- Founder · Product strategist, designer & builder
The product thesis
In regulated workflows, the most valuable AI structures, flags and evidences — never decides.
Trust is the feature. Auditability is the moat. The whole product is shaped around a model of AI that is deliberately less capable, not more.
Capture
The product's front door. Light-touch context — case stage, joint application, self-employed, gifted deposit — tunes the extraction without putting cognitive load on the adviser. Notes can be pasted, uploaded as a photo with OCR, or recorded as audio.
By the numbers
- 8–12
- Artefacts per case
- 40–60%
- Adviser week on admin
- ~70%
- Targeted admin reduction
- 10
- Structured tabs per file
- 0
- Editable audit entries
- MVP
- End-to-end flows live
Overview
Calltinuity is documentation infrastructure for UK mortgage and protection advisers. A single workflow surface that ingests a transcript or notes and produces a structured, navigable case file — every section editable, every section explicitly confirmed, every change logged. The AI is bounded by design: it extracts and structures, never recommends a product, estimates affordability, or infers missing financials.
The problem
01
Advisers are drowning in regulated admin.
A single mortgage case generates 8–12 distinct artefacts — pre-call brief, fact-find, follow-up email, document request, suitability rationale, protection log, vulnerability assessment, sign-off declaration. Most are reconstructed from memory hours after the call, in different formats by different advisers. Under FCA Consumer Duty, the gap between what was said and what is in the file is now an enforcement risk.
The insight
02
Advisers don't want an AI that thinks. They want one that remembers.
Talking to advisers, one pattern stood out. Their professional judgement is the product — they don't want it replaced. What they need is a system that captures the call faithfully, structures it, flags risk, and lets them review, edit, confirm, and lock — with a defensible audit trail at every step. This isn't an AI assistant. It's documentation infrastructure.
Workspace
Intentionally restrained: clients, recent calls, the three numbers that matter. Top navigation surfaces every supervisor capability without crowding the adviser's day-to-day flow. Premium product surfaces earn their density — they don't impose it.
The concept
03
Bounded AI. Adviser in control. Audit by default.
A single workflow surface that ingests a transcript and produces a tabbed case file: Brief · Fact-Find · Email · Checklist · Compliance · Protection · Advice · Suitability · Transcript · Audit Log. The AI extracts and structures — it never recommends a product, estimates affordability, or infers missing financials. High-risk patterns surface as flags for adviser review, never as conclusions. Every change is attributed and logged.
The product
04
Calm and professional, not flashy.
Paste notes, pick the case stage, and within seconds see a structured case file. Each tab is a Notion-style document — readable, editable, one-click confirmable. A right-hand status panel tracks file quality, outstanding documents, compliance flags, and next actions. The whole thing locks on declaration and produces an immutable snapshot for compliance review. Notion × Xero in feel, with a Stripe/Linear influence on the marketing surface.
Firm oversight
Firm-wide oversight is a first-class surface, not a settings sub-page. A traffic-light view of every case across the firm — Complete · Review Required · Incomplete · Vulnerability. Risk Radar exposes adviser-level patterns to supervisors only.
The trust model
05
Built around audit, not around automation.
Advice boundary lock enforced in both system prompt and UI. Roles in a separate table with security-definer functions. An append-only audit log under RLS — no updates, no deletes. Vulnerability calibration: when high-vulnerability indicators surface, the AI's tone shifts and adviser acknowledgement is mandatory. Every case is tagged with the active regulatory framework version so historic interpretations remain defensible.
Architectural commitments
Seven decisions, made once.
Each commitment is enforced in both the system prompt and the UI — not a setting, not a toggle, not a subscription tier. They're the spine of the product.
01
Advice boundary lock
No product recommendations, no affordability estimates, no inferred financials. Language calibrated — 'for adviser review', never 'recommended'.
02
Section confirmation
Every section requires explicit adviser sign-off before counting toward file quality. No silent agreement.
03
Override with justification
Disagreement with an AI flag is captured in writing, attributed to the adviser, and stored in the audit log.
04
Append-only audit
RLS prevents updates and deletes. Foreign keys enforce referential integrity. Hashed device fingerprint stored with every action.
05
Vulnerability calibration
High-vulnerability indicators trigger softer AI framing and mandatory adviser acknowledgement.
06
Lock & snapshot
Adviser declaration creates an immutable JSON snapshot of the entire case for regulatory defensibility.
07
Regulatory versioning
Every case is tagged with the active framework version so historic interpretations remain defensible.
How Calltinuity differs
Where the category ships generic, Calltinuity ships regulated.
Generic AI note tools summarise. Generic CRMs store static fields. Calltinuity produces domain-structured, attributed, lockable case files — built for a regulator to audit, not a sales team to ship.
In regulated workflows, the most valuable AI is the one that structures, flags and evidences — never the one that decides. Trust is the feature. Auditability is the moat.
My role
Founder · Product strategist, designer & builder.
- 01Defined category positioning — workflow infrastructure, not 'AI assistant' — and the regulatory thesis that the advice boundary is itself a feature.
- 02Designed the case-file model, the section-confirmation system, the override-with-justification pattern, and the file quality score.
- 03Modelled the schema — calls, call_outputs, audit_log, case_snapshots, user_roles — with RLS data isolation and role-based supervisor access.
- 04Wrote the advice boundary system prompt and designed the vulnerability calibration logic and immutable audit model.
- 05Built the full stack: React + TypeScript front end, Supabase (Postgres + RLS + edge functions), AI gateway integration, OCR pipeline, snapshotting and supervisor dashboards.
- 06Designed the 'Quiet Competence' visual system — Outfit typography, soft neutrals, single brand-blue accent, severity colours used sparingly.
- 07Shaped the supervisor surfaces — Compliance Dashboard, Risk Radar, DSAR exports — as enterprise-grade, not consumer-grade.
- 08Wrote every line of marketing copy — including the deliberate disclaimer that frames the product as documentation support, not advice.
In summary
Status: working MVP. End-to-end flows operational — transcript ingestion through to structured case file, adviser confirmation, lock and immutable audit export. Currently used for design-partner walkthroughs ahead of broader beta. Not yet in commercial production, not yet through formal FCA review, not yet stress-tested by working compliance teams — and framed honestly as such.