Gentle prompts to keep behaviour on track
Real-time reminders and tips that reinforce desired behaviours and learning.
CatalyzU is a seed-stage, venture-backed (supported by Future Africa, Techstars, Jobtech Alliance, & Meta) workforce intelligence and training platform that uses CVs, job descriptions, company documents and skill-gap data to identify what employees need to learn and why. From there, CatalyzU builds contextualised training, delivers it through WhatsApp and measures its impact against business KPIs.
Its flagship concept, the Training Strategy, turned that intelligence into an end-to-end AI workflow. Administrators could move from workforce data to a tailored training programme in one journey. I designed that original experience.
Four months later, engineering constraints and MVP priorities led the team to deprioritise the Training Strategy in favour of a Course Library. My challenge was turning that direction into a Course Library that still reflected the intelligence behind the Training Strategy.
I redesigned the administrator journey from strategy generation to recommendation-led discovery, deciding what work could move to the platform, what administrators still needed control over, and how the intelligence behind each recommendation could remain visible.
This case study is about what survived the pivot, what didn’t, and how I translated a much bigger product vision into a more feasible MVP.
PROBLEM
CatalyzU's Training Strategy turned workforce data into tailored training, but each bespoke strategy took 20–30 minutes to generate and and wasn't reliable enough for the MVP. We needed to simplify the journey without reducing the value behind it.
SOLUTION
I redesigned the administrator’s role from building the training strategy to reviewing, verifying and deciding what to deploy. This moved the heavy lifting to the platform while keeping administrators in control of the decisions that mattered.
The platform prioritised courses around diagnosed skill gaps, giving administrators a clear starting point while the toggle let them explore the full library when needed.
Each course shows the skill gaps it addresses and the KPI it aims to influence, helping administrators understand why this course matters before reviewing the content.
Administrators can preview the learning experience their workforce will receive and see which company materials shaped it. This turned contextualisation from an AI claim into something they can verify.
RESULTS
The recommendation-led Course Library was adopted for the MVP and is now live in pilot with 2 enterprise clients. While the experience is significantly simpler, we deliberately preserved the intelligence behind the original Training Strategy, distilling its strongest elements into workforce-led recommendations, skill-gap and performance alignment, and company-specific contextualisation.
DESIGN CHALLENGE
The original flow asked administrators to do a lot of the work themselves: providing inputs, reviewing hypotheses, adding context and tailoring their choices before waiting for the Training Strategy to be generated.
As we simplified the experience, the question became how much of that work CatalyzU could take on while preserving what made Training Strategy valuable, and still giving administrators the context, proof, and control they needed to make an informed decision.
RESPONSE 1 : WHO DOES THE WORK?
Instead of asking administrators to provide more inputs, we could draw on information CatalyzU already had, from client onboarding and workforce baseline assessments to company SOPs and processes. I used this to rethink where the work should happen.
I then redesigned the interaction around a different split of responsibility:
The platform:
Analyses existing workforce and skill-gap data
Identifies priority training needs
Prepares and contextualises relevant courses
The administrator:
Understands why a course was recommended (based on skill gaps)
Reviews what employees will receive
Deploys to workforce / team / employee
The new flow removes the friction and cognitive load of finding the right solution. Administrators start with a curated selection, can see why each course is recommended and what problem it addresses, and still have the agency to make the final decision.
RESPONSE 2: How does the administrator trust work they didn't do themselves?
Contextualisation used to be something administrators actively did, providing additional information (SOPs, product specs, playbooks, and internal processes) after a strategy was generated.
Moving that work to the platform removed the effort, but raised another question: if administrators were no longer involved, how would they know their organisation’s information had actually shaped the training?
Even if administrators no longer had to contextualise the training themselves, they still needed to see how their company context was being used. I made the contextualisation something administrators could verify.
The platform:
Draws on company documents, SOPs and processes
Uses organisation and role-specific context to shape the training
Applies that context before the administrator reviews the course
The administrator:
Sees which materials were used
Previews the contextualised lessons employees will receive
Checks that the training reflects their organisation before deployment
There was also a specific request from stakeholders to preview the learning exactly as employees would experience it on WhatsApp. I brought this into the course overview, giving administrators visibility into both the content itself and how company context had been applied.
RESPONSE 3: How do they understand why this training matters?
In the original Training Strategy, hypothesis approval sat alongside an in-depth skill-gap analysis of the selected team or cohort. Administrators could see where the workforce was falling short, then use that information to evaluate the training hypotheses being proposed.
Old flow · Step 1: Understand the workforce gaps
Administrators first reviewed the selected team or cohort’s skill-gap analysis to understand where support was needed.
Old flow · Step 2: Review and approve hypotheses
They then used those identified gaps to evaluate and approve the training hypotheses that would shape the Training Strategy.
But that context was temporary. Once the hypotheses were approved and the Training Strategy generated, the skill-gap analysis was no longer visible alongside the resulting training.
When we removed hypothesis approval from the new flow, I used the opportunity to make that context a permanent part of the Course Overview instead.
I kept the workforce need connected to the training, giving administrators the context to evaluate a course before getting into the curriculum.
AFTER
Skill gaps and business KPIs now sit above the curriculum in the Course Overview, keeping the workforce need and performance goal visible as administrators review the training.
The MVP gave up live, bespoke strategy generation on demand. Courses could be prepared and contextualised in advance, but administrators had fewer opportunities to configure and approve the strategy than the original vision allowed.
What we gained?
No 20–30 minute generation wait.
Less configuration and cognitive load for administrators.
A build the team could reliably deliver within the MVP timeline.
A more dependable experience for demos and the first pilot.
Training Strategy hasn’t been abandoned. Live generation has been deferred until the product can support it reliably.
I had to work out which parts of the original flow were central to CatalyzU’s value, and what we could afford to leave behind for the MVP.
Find the smallest version that still matters. Working within an MVP meant learning where to compromise and where not to.
Be willing to let previous work go. Removing interactions I had already designed made room for a clearer, more feasible experience.
Use AI to explore decisions, not make them. Working in high fidelity with UX Pilot made exploration faster, but also taught me not to mistake polish for a good decision.
decision points before deployment
generation wait removed from the admin journey
platform handles more upfront, reducing repeated admin work




