Book a call →

I turn operational chaos into intelligent systems.

I design AI agents, automation and data architecture for customer-facing operations. The work has saved £190k+ annually, lifted CSAT from 53% → 81%, cut resolution times by 87%, and now runs in teams accross three continents.

Lyubo Voynov

Numbers that survived contact with production.

The measurable outcomes from real operational systems — not pilots, not slideware.

£0k+
Annual savings delivered
0→81%
Customer satisfaction (CSAT)
0%
Faster resolution time
0%
Vendor cost reduction
0%
Complaint reduction (YoY)
0→78
Net promoter score (NPS)
0+
Operational sites managed
0+
Enterprise locations rolled out

End to end — from finding the problem to shipping the system.

Most people own one slice of this. I run the whole loop, which is why the work actually lands.

01

I find high-value problems

SQL, workflow analysis and operational data to surface where time and money actually leak.

Discovery
02

I build the business case

Turning the findings into a numbers-backed argument that wins leadership buy-in.

Strategy
03

I deliver cross-functionally

Aligning engineering, product, finance and CX so the solution ships and sticks.

Delivery
04

I implement AI systems

Agents, automation and workflow redesign built for measurable ROI — not demos.

Build

An autonomous agent that works the queue like a teammate.

A simplified view of one system I architected: it reads tickets, reasons in context, resolves what it can, and escalates the rest with a one-line brief — looping until the queue is clear.

Runtime loop

New tickets

email · chat · forms

Reasoning core

classify · pull context · decide

Resolved
answered and closed
Awaiting customer
set to pending, watched
Escalated to a human
with a one-line summary
0
Resolved by the agent
0
Escalated with context
0
Awaiting the customer
How the reasoning layer is built stays under the hood The loop is the what. The training, routing and decisioning is the how — and it's tailored to each stack.
Let's talk about yours →

Evidence, not adjectives.

A few systems I've owned end to end, and why each one mattered to the business.

JustPark × Zendesk Flagship

AI-powered CX transformation

53→81%
CSAT
87%
Faster resolution
£90k
Annual savings

End-to-end redesign of the customer-experience function — triage logic, agent support and escalation rules rebuilt around an AI reasoning layer.

Why it matters — proved an AI agent could carry real ticket volume without trading away quality. The template for everything since.
Read the Zendesk customer story
Operational redesign

£190k+ annual savings

£190k+
Saved / year

Identified workflow inefficiencies through SQL analysis, secured stakeholder buy-in, and delivered an automated operational redesign.

Why it matters — recurring savings. It compounds every year it runs.
Enterprise rollout

Multi-site rollout at scale

100+
Locations
2,500+
Sites

Delivered a multi-million-pound rollout across 100+ locations and 2,500+ operational sites.

Why it matters — the systems hold up under real scale, not just in a single team.
Risk monitoring

Risk-prevention system

£100k+
Losses prevented / yr

Designed a monitoring framework that catches operational failures before they turn into revenue loss.

Why it matters — this one protects the downside instead of chasing upside.

Most organisations don't have an AI problem. They have a workflow problem.

AI only creates value when processes, data quality and decision pathways are designed correctly. My work is finding those opportunities and building the systems that turn them into measurable business outcomes — then handing teams something they trust.

What I actually work with.

The tools and disciplines behind the systems above.

AI

OpenAI & Anthropic ModelsAgentic workflowsPrompt engineeringAI evaluation

Data

SQLAnalyticsReportingData quality

Automation

APIsWorkflow automationZendeskGoogle Workspace

Delivery

ScrumPRINCE2Product discoveryStakeholder management

Have a workflow worth fixing?

Whether it's manual overload, scaling pains, or a CX function that needs an upgrade — tell me about it. Quickest path is to grab a slot.

Send a message
I read every one and reply personally.

✓ Message sent. I'll be in touch soon.

Something went wrong — email me directly at voynov.lyuboslav@gmail.com