HarryWestworth

Automations Lead at Knowunity · AI systems, internal tooling & product

He / HimManchester · Berlin

I find the bottleneck, build the system that removes it, then make sure the team never has to think about it again.

I lead automation across Knowunity, moving between AI systems, product engineering and internal tooling for teams supporting more than 25 million students.

Selected work
01

Selected work

A short index of the kinds of problems I help make clearer.

Selected work

Growth · Finance · 2025—26Creator operations without the manual workBuilt a connected creator-operations system spanning sourcing, invoicing, payouts and duplicate-content checks across more than 25 markets.Sourcing surfaces roughly 250 candidates per country and run, while payout automation removes 8–10 hours of weekly finance work.Architecture · Automation · Operations

Creator operations without the manual work

The bottleneck

Creator scouting, invoicing, payout preparation and cross-post checking were separate manual workflows that became harder to scale together.

The system

A connected operating layer that discovers and evaluates candidates, generates and reconciles invoices, prepares payouts for approval and prevents duplicate payment across platforms.

What changed

Growth reviews ranked candidates instead of trawling platforms, while Finance has recovered 8–10 hours every week.

Trust & Safety · 2025AI moderation that removed the review queueImplemented a custom AI content moderation model designed around the realities of a global student platform.Improved model reasoning accuracy from roughly 40% to 99% and reduced manual moderation workload by approximately 80%.Model design · Product · Rollout

AI moderation that removed the review queue

The bottleneck

A growing moderation queue was consuming specialist attention on decisions that could be handled consistently by a model.

The system

A fine-tuned moderation model shaped around Knowunity’s content, risk thresholds and escalation requirements, with a repeatable evaluation loop.

What changed

Reasoning accuracy reached 99%; manual review fell by approximately 80%, while higher-risk cases remained visible to the team.

Customer operations · 2025A support agent that resolves, not just routesBuilt and rebuilt an AI-powered Freshdesk agent with 12 operational tools, autonomous resolution and clear human escalation.Automatically resolves 93% of roughly 350 weekly tickets, moving resolution from a two-day SLA to under one hour.AI systems · Service design

A support agent that resolves, not just routes

The bottleneck

The support team repeatedly answered predictable questions, leaving less capacity for sensitive or ambiguous cases.

The system

A support agent connected to the existing service workflow, using prompt caching, parallel tool use and 12 tools with clear boundaries for escalation.

What changed

Ninety-three per cent of roughly 350 weekly tickets are now handled automatically; resolution moved from a two-day SLA to under one hour.

Student product · 2025—26Mock Exams, from weekend prototype to productTook the original hackathon concept into a feature live across all markets, owning database migrations, backend endpoints, web implementation, user research and iteration.Student research returned 9–10/10 ratings, with each session feeding directly into the product now live across every market.Product engineering · Research · Evaluation

Mock Exams, from weekend prototype to product

The bottleneck

Students needed targeted exam practice that reflected their curriculum, alongside a product loop capable of learning quickly from real use.

The system

Mock Exams combine curriculum-aligned generation, exam-board selection, grading, file uploads and a clearer submission experience.

What changed

The feature moved from a winning weekend prototype to every market, with 9–10/10 research sessions directly shaping what shipped.

Company-wide enablement · 2026AI capability that belongs to the whole companyBuilt product-style AI enablement through hands-on workshops, weekly feedback releases and a Knowledge Hub spanning five internal sources.Finance, Marketing, Sales and People now build their own automations, while recurring feedback ships as weekly improvements.Strategy · Facilitation · Systems

AI capability that belongs to the whole company

The bottleneck

AI adoption depended on one-off announcements and a small number of technical builders rather than repeatable team capability.

The system

Hands-on workshops, internal hackathons, a weekly product loop and semantic search across Linear, Notion, GrowthBook, Slack and Granola.

What changed

Teams across the company now build their own automations and see their feedback reflected in the tools they use.

These systems sit across operations, product and growth. The common thread is capacity: giving teams more time for the work that needs their judgement.

02

The operating loop

How recurring friction becomes durable capacity.

The operating loop

The work is not finished when the automation runs.

  1. 01

    Observe the work.

    Get close enough to see where time, attention and judgement are actually being spent.

  2. 02

    Find the constraint.

    Separate recurring friction from the symptoms and choose the problem with the most leverage.

  3. 03

    Build and measure.

    Take the system end to end, instrument what matters and use feedback to keep improving it.

  4. 04

    Hand it over.

    Make it understandable and durable, then move to the next constraint without creating dependency.

03

Experience

Four roles at one company; each one created by the work that came before it.

Experience

Knowunity · 2023—Present

From student ambassador to leading automation across the company.

01

Automations Lead

Freelance · Hybrid

May 2026—Present

Lead automation company-wide, choosing the highest-leverage operational constraints and shipping the systems end to end.

  • Own and ship the company-wide automations roadmap
  • Built creator sourcing, payout and integrity systems that remove days of manual work
  • Lead practical AI enablement and product-style internal tooling across the company
02

Automations Engineer

Full-time · Hybrid

Nov 2025—May 2026

Expanded from operational automation into end-to-end product engineering while shaping the roadmap that became the Lead role.

  • Owned Mock Exams across database, backend, web, user research and iteration
  • Shipped web product work tied to measurable retention and conversion lifts
  • Created a multi-market localisation pipeline for more than 15 languages
03

Growth Analyst

Full-time · Berlin

Oct 2024—Nov 2025

Combined product analysis, experimentation and automation to ship improvements for students and internal teams.

  • Improved moderation reasoning accuracy from roughly 40% to 99%
  • Built the autonomous support agent
  • Took the study material generator from concept to launch
04

Community Support Assistant

Freelance · Remote

Sep 2023—Oct 2024

Joined from Knowunity’s student community and quickly moved from support into systems and product work.

  • Scaled the UK Ambassador Programme to five times the engagement of any other market
  • Built a knowledge hub for operational procedures
  • Improved trust and safety escalation processes
04

Working principles

A few convictions that shape the way I approach the work.

Working principles

01

Start with the bottleneck.

The technology is secondary. First understand where time, attention or judgement is being lost; then build precisely enough to remove it.

02

Run internal tools like products.

Track use, make feedback easy, ship improvements visibly and give people confidence that the system will keep getting better.

03

Create capability, not dependency.

The strongest systems do not depend on their builder. They give teams more capability, remain understandable after handover and free their creator to move to the next constraint.

05

Profile

Manchester · Berlin · Working across design and technology.

Profile

I joined Knowunity at 16 with no CV. Today I lead automation across the company: setting the roadmap, shipping production systems and measuring what changes after launch. The work rarely stays in one discipline. I move between operational design, AI systems, product engineering and user research—whatever is required to turn recurring friction into a system the team can own.

Areas of focus

  • Automation strategy
  • AI systems & agents
  • Internal tooling
  • Product engineering
  • User research & evaluation
  • Team enablement

Education

Preston College

Level 3 Extended Diploma, Information Technology

2023—2025 · D*D*D · Highest possible grade

Completed while working at Knowunity throughout the course; awarded a scholarship for academic performance.

Penwortham Priory Academy

GCSEs

2018—2023 · 3× A* · 2× A · 2× B · 1× C

A* grades in Computer Science, Mathematics and English Literature.

Recognition

1st Place · Lancashire & Cumbria IoT Skills Competition

The Lancashire Colleges Skills Competition · Mar 2025

Commended · Innovations and Ideas Learner Award

Preston College · Jun 2024

06

Notes from the work

Longer reflections on building, systems and learning.

07Start a conversation

Good work often starts with a clear conversation.