Case Study 06: Enterprise Implementation
Compliance-First Learning Ecosystem
A regulated lender had twelve federal requirements and a spreadsheet
Read the full engagement report
- 01RequirementTwelve named federal obligations: TILA, RESPA, ECOA, HMDA and eight others, written down before any course was scoped.
- 02ObjectiveWhat the employee must be able to do, stated at a level that can be observed.
- 03AssessmentThe check that the objective was met. Compliance courses were gated on it.
- 04RecordWritten by the platform, timestamped, and not editable by the person it covers.
42 courses built on this chain. 0 findings at federal audit.
01: ANALYSIS
The compliance risk was not the training. It was the evidence.
- The organisation
- A financial services startup in rapid headcount growth, in a sector where a federal examiner can ask on any given day who was trained on what, and when.
- What existed
- A legacy platform built for a headcount the firm had already passed, and twelve manual spreadsheets standing in for an audit trail. There was no learning function behind either.
- What was actually at stake
- Twelve federal requirements including TILA, RESPA, ECOA and HMDA. Under examination the question is not whether training happened but whether it can be evidenced, and a spreadsheet is evidence only until someone edits it.
- Twenty-four interviews
- Across six departments, producing 47 documented requirements. The interviews mattered as much for adoption as for scope: the people who would later be asked to change had already been asked what they needed.
02: THE PROBLEM
Buying a platform is the easy half. Getting it used is the engagement.
- Evidence
- Twelve spreadsheets maintained by hand, each one editable by the person it exonerates. An examiner does not ask whether the training happened. They ask you to prove it, and a document that can be changed after the fact proves nothing.
- Capacity
- A legacy platform sized for a headcount the firm had already passed, with no learning function behind it to notice. Every month of growth widened the gap between who needed training and who could be tracked.
- Adoption
- A platform nobody opens carries the same audit exposure as no platform at all, at greater cost. Most of the risk in this engagement sat after go-live, not before it.
Why the change model was chosen with the vendor, not after it
Treating adoption as a rollout task at the end of the project would have put the change work after the money was already spent, when the only lever left is asking people harder. Selecting the change model alongside the platform meant the two could be evaluated against each other: a system that fights the way people already work cannot be rescued by communications later.
03: THE DESIGN
Evidence had to be a by-product of work, not one more thing to do.
The constraint every decision was tested against
The firm had already demonstrated what it would not sustain. Twelve spreadsheets were not a tooling failure, they were the predictable result of asking busy people to maintain a record that produced no value for them. Any design that placed the burden of proof on the learner or the manager would have decayed the same way, more expensively. So the brief inverted: enrolment, completion and the audit trail had to be generated by systems people already touched for other reasons, and the curriculum had to trace back to a named federal requirement so that any course could be defended under examination rather than merely counted.
01
ADDIE for the curriculum
Forty-two courses across compliance, onboarding and professional development, each traced from a documented requirement to an assessable objective, so every course could answer why it existed.
02
ADKAR for the people
Awareness and desire were built before knowledge and ability, and reinforcement was scheduled rather than hoped for. In a startup absorbing four times its headcount, the constraint is attention, not capability.
03
Automation as audit evidence
Integration with the HRIS, Asana and Jira meant enrolment and completion were produced by the systems people already used, so the audit trail was a by-product of work rather than an act of record-keeping.
04
Executive visibility
Real-time dashboards for the C-suite. Sponsorship is sustained by evidence that the thing is working, and withdrawn quietly when leaders have to ask.
The traceability chain, requirement to record
- Requirement
- A named federal obligation. TILA, RESPA, ECOA, HMDA and eight others, each written down before any course was scoped, so the curriculum answered to regulation rather than to whoever asked loudest.
- Objective
- What the employee must be able to do, stated at a performance level that can be observed. Not what the course would cover, which is a description of content and cannot be assessed.
- Assessment
- The check that the objective was met. Compliance courses were gated on it, because under examination a completion record that only proves attendance is a record of attendance.
- Record
- Written by the platform from the systems already in use, timestamped and not editable by the person it covers. This is the row an examiner actually reads.
Why mandatory and elective were kept visibly separate
Compliance, onboarding and professional development were built as three tracks rather than one catalogue. The temptation in a growing firm is to mandate everything, which reads as seriousness and behaves as noise: once professional development is compulsory, a completion rate stops distinguishing the training the regulator asks about from the training the company merely likes. Keeping the mandatory set small and gated is what allowed the completion figure to mean something.
04: PLATFORM SELECTION
Five platforms, twelve weighted criteria, eight weeks.
How the decision was defensible
Five enterprise platforms were scored against twelve weighted criteria including compliance capability, scalability, API depth and five-year total cost of ownership, across fifteen demos with six stakeholder groups. The selected platform was neither the cheapest nor the most expensive. Writing the criteria and their weights before seeing the demos is what makes a vendor decision reviewable afterwards, and what stops the loudest stakeholder from becoming the deciding one.
- Phase 1
- Discovery and requirements. Twenty-four stakeholder interviews across six departments, twelve federal requirements documented, and a learning roadmap aligned to a three-year growth plan.
- Phase 2
- Evaluation and selection. Five platforms against twelve weighted criteria, fifteen demos, security assessment, contract negotiated and board approval secured.
- Phase 3
- Configuration and content. Compliance workflows configured, HRIS, Asana and Jira integrated, and 42 courses developed producing 156 assets.
- Phase 4
- Pilot and launch. Four weeks with 25 users across three departments, content revised against the feedback, then company-wide launch reaching 94 per cent adoption within 30 days.
05: OUTCOME
Zero findings, and an L&D function that outlasted the project.
Measured
Zero compliance violations at federal audit and 100 per cent completion on mandatory training. Time to productivity for a new hire fell from 21 days to 8. Twelve manual tracking spreadsheets were retired and audit preparation time fell by 85 per cent. Verified annual savings of 127,000 dollars, composed of 45,000 in eliminated manual tracking, 52,000 in onboarding efficiency and 30,000 in audit preparation.
What I would attribute and what I would not
The compliance and audit-preparation results are directly attributable: the trail either exists or it does not, and it did not before. The onboarding figure is a fair claim but not a controlled one, since the firm was also formalising other parts of induction over the same period. The savings figure is the finance team’s, calculated on their model, not mine. Naming which numbers are clean matters more than the size of any of them.
06: EVALUATION
The only case here where the levels were measured, not designed.
Why this table reads differently from the others
Every other case on this site states an evaluation plan, because the artefact was built as a portfolio piece and the measures are instrumented or intended. This one ran inside a real organisation for twelve months, so the levels below were taken rather than proposed. That makes the caveats matter more, not less: a measured number invites a reader to stop asking where it came from.
| Level | Evidence | Status |
|---|---|---|
| 1: Reaction | Taken as pilot feedback rather than a satisfaction score. Four weeks with 25 users across three departments, and the content was revised against what came back before the company-wide launch. Reaction that changes the material is worth more than reaction that is counted. | Measured |
| 2: Learning | 100 per cent completion on mandatory training, where completion was gated on assessment rather than on opening the course. That proves the assessment was passed. It does not prove the knowledge holds a quarter later, which no measure here was designed to test. | Measured |
| 3: Behaviour | 94 per cent adoption within 30 days, and time to productivity for a new hire falling from 21 days to 8. The adoption figure is clean. The onboarding figure is fair but uncontrolled, since the firm was formalising other parts of induction over the same period. Neither one shows an employee applying TILA in a live conversation with a borrower, which is the behaviour that actually matters and the measure this engagement did not build. | Measured |
| 4: Results | Zero findings at federal audit, twelve manual tracking spreadsheets retired, and audit preparation time down 85 per cent. Those are directly attributable: the trail either exists or it does not, and it did not before. The 127,000 dollars in verified annual savings is the finance team’s figure calculated on their model, not mine, and it is quoted as theirs. | Measured |
07: REFLECTION
What I would do differently.
On sequencing
The pilot ran with 25 users across three departments and was the right size, but it ran after content development rather than alongside it. Piloting three courses at the end of Phase 3 would have tested the format before 42 of them existed in it.
On what compliance training cannot do
One hundred per cent completion is an audit outcome, not a learning one. It proves everyone was reached; it does not prove anyone can apply TILA at the point of a real conversation with a borrower. The honest next measure is a behavioural one taken at the desk, and it was outside this engagement.
The full engagement report carries the vendor matrix, the phase detail and the sponsor's account.