The Chartered Governance Institute UK & Ireland (CGI)

Overview

We built a 10-step AI production pipeline that processes 1.4 million words of source content into a structured 200-hour modular learning programme across five governance sectors. Each step enforces strict content authority rules. Two CGI staff members now operate the entire system independently.

Alongside the pipeline, we deployed 10 standard assistants across finance, operations, marketing and other teams, giving the wider organisation immediate productivity gains. Combined, the pipeline and standard toolkit deliver 1,500+ hours of annual productivity.

Modernising a qualification at scale

The Chartered Governance Institute sets professional standards for company secretaries and governance officers. The Foundation Programme is CGI's entry-level qualification, delivered across five governance sectors. CGI needed to restructure the entire programme for RQF Level 4 delivery.

The challenge was scale. The source material comprised 1.4 million words across 65 files in Word and PDF formats. Converting this into structured, AI-readable content while maintaining content authority and regulatory compliance was beyond any manual approach. Subject matter expert availability was limited, slowing review cycles. No existing tooling could process material at this volume while preserving accuracy.

10 pipeline steps with strict rules

We started with 10 standard assistants deployed across all teams, covering finance, operations, marketing and AI guidance. These gave every user immediate productivity gains from day one, separate from the pipeline work.

We then built the production pipeline: 10 distinct steps from market research through to final content build. Step 1 analyses competitor programmes. Steps 1.5 and 1.75 convert evidence into structural recommendations and module syllabuses. Steps 2 through 2.75 handle unit planning, topic planning, gap research and case study standardisation. Step 3 develops full topic content with reference appendices. Step 3.5 runs a quality gate checking case quality, date currency, maths integrity, governance relevance and upstream fidelity. Step 4 builds final storyboards mapped to learning platforms.

Every step enforces four governing rules: Authority of Source (content extracted, never invented), Traceability (every piece maps to the competency framework), Specialisation (each assistant has one job) and Anti-Hallucination (strict search-and-retrieve with citations). We built the pipeline collaboratively with CGI's education team through weekly sessions so they designed each step rather than receiving a finished product.

1.4 million words structured

The full 10-step pipeline is operational. All 1.4 million words of source content have been converted, indexed and structured. 22 foundation knowledge base files and 43 sector research files are AI-readable. The system produces structured six-part learning activities with checkpoint assessments for each topic.

Two CGI staff members operate the pipeline independently using a detailed operator guide. They run the canonical development route (topic planning through content checking through final build) without external support. The quality gate at Step 3.5 enforces pass/block decisions: content only reaches the build stage when every required check clears.

Combined, the pipeline and standard toolkit deliver 1,500+ hours of annual productivity across the organisation. CGI's Foundation Programme is on track for phased release across all five governance sectors.

From one course to any course

We are turning the 10-step pipeline into a fully agentic workflow. Today, each step requires a human operator to trigger the next. The agentic version will chain steps automatically: when Step 2.1 flags coverage gaps, Step 2.5 fires without a manual handoff. Quality gates at Step 3.5 still enforce human review at critical points, but the routine transitions between steps disappear.

The bigger shift is scope. The current pipeline is built for one specific course. The agentic version will work with any type of course or qualification programme. CGI will be able to point the system at new source material and produce structured learning content at the same quality standard, regardless of subject area.

This moves CGI from a one-time restructuring project into an ongoing content production capability. New programmes, updated regulations, refreshed case studies: all can flow through the same pipeline with the same governing rules.

What we learned

Topic-level granularity kept AI conversations focused. An early module-level version had eight integration issues, all resolved by the topic-level redesign. Processing one topic at a time through the full chain produced consistently better output than trying to handle modules in bulk.

Subject matter expert availability was limited. SMEs reviewed content in evenings, slowing approval cycles. We adapted by building the pipeline to produce self-contained outputs that could be reviewed independently, removing the bottleneck of needing everyone in the room at once.

For industrial-scale content production, strict extraction rules matter more than AI sophistication. The four governing rules (Authority of Source, Traceability, Specialisation, Anti-Hallucination) prevented quality drift across 1.4 million words. Without those guardrails, accuracy would have eroded at scale.

*This engagement was delivered through the smartAI programme, our reseller partnership with smartimpact. Read the smartimpact case study to learn more.