The ladder.
Not a career history. A ladder. Every rung was climbed to get to the next one, and the whole thing points somewhere.
The past is not a list of employers; it is where the stripes came from. Nobody needs the dates. They need to understand that the ground has been covered. Ten rungs: what each one was, and what it left behind.
The through-line
Read the rungs and the same job appears at every altitude. Someone turns up with a problem, or an idea, that they cannot yet put into words. It gets turned into something that exists and works.
“He has the ability to elicit requirements from fog.”
A project manager wrote that down years ago, better than we can say it. That is the whole business. The fog has not changed. What has changed is that the thing on the other side of it now takes weeks instead of years.
The rungs
10The machine itself
The 1980s
IBM, MS-DOS, Lotus 123, dBase. Clipper and Turbo Pascal. Compiled code, printed manuals, software you could go into a shop and try before you bought it.
An understanding of how software is actually made, from a time when nothing was abstracted away. Most people building with AI today sit one layer above the machine. This is several layers below that, and it shows up every time something breaks in a way the documentation does not cover.
The hired gun
The 1990s
Contractor, not consultant. Process modelling and re-engineering across the UK, Europe, Japan and the US. Systems architecture for public bodies where the security considerations were real.
The ability to walk into an organisation cold, work out what is genuinely wrong, and say so in a sentence anyone can understand. Back then that was called doing the job. It has since acquired a longer name.
Enterprise scale
Around the turn of the century
Private banking. Retail and logistics supply chains joining Marks & Spencer to Federal Express. A bespoke CRM and ERP estate for Imperial Tobacco, renewed eight times over six years.
Scale is mostly other people. Five system integrators, five contracts, five agendas, one deadline. The technical problem is rarely the hard part.
Programmes in trouble
Early 2010s
SAP finance frameworks. The total replacement of a plc's front and back office. Tier 1 ERP data migration at Network Rail, with a million pounds taken out of supplier contracts on the way through.
How to pick up something that is failing, find the two things actually causing it, and land it anyway.
Data, and the first AI rung
2012 onwards
Ask Verizon: five customised virtual support avatars serving over 120,000 users, driven by phoneme and viseme extraction. Then Virgin Trains, where the BI platform turned a monthly view into a daily one, automatic delay repay became a European first and is now a government requirement for every UK operator, and automated gates went into 64 stations.
Data is worthless until it changes a decision. And the questions worth asking are usually the ones nobody thought to ask, which is a lesson that turns out to be the entire premise of agentic systems.
Product, not project
2018
Tableau Technology Partner, first in the UK.
The difference between delivering something and taking it to market. They are not the same discipline and most engineers never learn the second one.
Building inside a regulated business
2019 to 2024
Went into a bank on contract, stayed as an employee to build the applications and data practice from nothing. Enterprise data warehouse, Dynamics CRM and ERP, KYC and AML, a core banking integration, and the bank's own data policy.
Governance as a design constraint rather than a box to tick afterwards. This is the most transferable thing on the ladder, because it is exactly what separates an AI demo from an AI system anyone will sign off.
Building an AI function
2024 to 2026
Director of Technology. Built the AI practice from scratch, hired the engineers, wrote the strategy, architected private RAG and GraphRAG environments and multi-agent orchestration, and sat in front of a group board explaining what was worth doing and what was not.
What it actually takes to move AI from experiment to revenue line, including the parts that are nothing to do with technology.
Hands back on the tools
2026
Six production systems, roughly 2,000 commits. Multi-tenant platforms, graph-backed semantic matching, voice synthesis, a markdown editor built and shipped in nine days. Alongside them a method: competing design agents, a visual prototype gate before any code, golden-master verification so correctness is proved rather than hoped, and tiered model routing so it does not cost a fortune.
Capacity is no longer the constraint. Judgement is. Which means the years underneath this rung are worth more now than they have ever been, not less.
Inteevo
Now
Agentic engineering. AI architecture. Mentoring. And the thing that has run through every rung: taking an idea someone cannot quite articulate and turning it into something real.
“I open the curtains and show you what it can actually do. Not the spin.”
André Jacyshyn
Agentic engineering. Systems where agents do the work, with the verification and guardrails that make them trustworthy.
AI architecture. The structure that lets a working demo become a system people rely on. This is where most projects stop, and it is the reason Inteevo exists.
Mentoring. Teaching teams to work this way rather than doing it for them. Leaving somewhere better than you found it is the part of this job that is actually satisfying.
Innovation from a standing start. For people who have an idea and cannot yet describe it. Fog to working system.
AI is an enabler. An amazing tool in the right hands, capable of taking something from obscurity into the mainstream. Treated as a crystal ball, or as some kind of unicorn, it is not.
Not a sceptic, not an evangelist. A demonstrator.
Intelligence, Evolved. The practice exists because the distance between a working demo and a dependable system is architecture, and architecture is where most projects stop. André designs it, builds it, and teaches teams to do the same. The part he enjoys most is leaving a team able to do things they could not do before.
Engagements designed to monitor people, manipulate them, or quietly cut headcount while calling it something else. And twelve-week procurements with six stakeholders who cannot decide; life is short and so are good build windows.
Bring us the fog.
A short, no-obligation call about the problem. If we are not the right fit, we will say so and point you somewhere better.