In this guest article, Jon Block, Founder & Principal Consultant, Syllepsis delves into why every team in your organisation – not just engineering – should be using AI coding agents and why it’s one of the biggest competitive blind spots in business today.
Your finance team should be using AI coding agents.
So should marketing. And operations. And product. If that sounds absurd, you’re in good company – most organisations feel exactly the same way. But whilst everyone else dismisses this as relevant only to technical companies, a handful of organisations are quietly building competitive advantages that compound by the month.
Because here’s the thing: AI coding agents aren’t just development tools anymore. They’re becoming the universal operating system for knowledge work. And the fact that this still sounds radical is exactly why early movers will build such significant advantages.
What Changed
We talk a lot about Software 3.0 at Syllepsis – Andrej Karpathy’s term for programming in natural language. But most people still think that means “engineers typing to ChatGPT instead of writing Python.”
No.
Software 3.0 means that structured English instructions become your codebase. And once you accept that premise, everything shifts.
Because if instructions written in English are now your source code, then suddenly every department in your company should be writing code. Finance writes code when they specify how monthly reports should be generated. Marketing writes code when they define how product documentation gets created. Operations writes code when they automate workflow approvals.
They just don’t realise it yet.
Why AI Coding Agents, Specifically?
You might be thinking:”Why not use chatbots or workflow automation platforms?”
Because those tools are brittle. They’re black boxes. And they can’t scale the way knowledge work actually needs to scale.
AI coding agents – platforms like Cursor, Windsurf or VS Code (with AI extensions) – offer something fundamentally different:
File organisation. Your entire automation infrastructure lives in one navigable workspace – context documents, prompt libraries, testing frameworks and output directories.
Version control. These platforms integrate directly with GitHub, meaning your finance team can collaborate on prompt engineering the same way engineering teams collaborate on software – with branches, pull requests, peer review and rollback capabilities.
Testing frameworks. Wrap your English instructions in automated tests that validate outputs, cross-reference data sources and flag anomalies. This is production-grade automation.
System integration. Through protocols like MCP (Model Context Protocol), AI coding agents connect to databases, APIs, SharePoint and internal systems whilst maintaining the context and instructions that govern those integrations.
And here’s the critical bit: it’s all repeatable.
This Is About Repeatability, Not Magic
This isn’t about asking an AI a question and hoping it gives you a good answer.
This is about writing structured English instructions that can be executed again and again, identically, with carefully controlled inputs and carefully managed instruction nuances, wrapped in testing frameworks and producing consistent outputs that can be automatically verified – doing this in minutes rather than hours or days.
You’re not having conversations with AI. You’re building automated systems that happen to be controlled through natural language.
How This Actually Works
Finance: Monthly Board Reporting
Your finance team produces monthly board reports synthesising data from your accounting system, CRM and various Excel trackers. Currently it’s about manual exports, copy-pasting, reconciliation and strategic narrative writing – taking hours or days each month.
With AI coding agents:
- Prompt templates in version-controlled files define exactly how to extract and process each data source
- Automated tests validate figures, sections and charts
- Execute a single command – the agent pulls data, processes it, generates narratives and visualisations, outputs a complete draft with human checkpoints
- Improvements go through peer review before being merged
- Next month, everyone benefits
The finance team isn’t writing Python. They’re maintaining structured English instructions with engineering rigour.
Marketing: Product Documentation Pipeline
Your product marketing team maintains documentation across websites, training platforms, internal wikis and customer PDFs. Traditionally this involves manual translation from technical specs, separate versions for each channel and extensive effort maintaining consistency.
With AI coding agents:
- Product information lives in structured markdown files
- Prompt templates define how to transform this for different audiences and formats – Testing prompts verify brand voice, technical accuracy and required elements – When features update, run the documentation pipeline – generate everything simultaneously – Version control tracks evolution of both product information and prompts – The entire suite updates in minutes with automatic validation
No traditional code. Just systematic, repeatable processes controlled through structured English instructions.
Who’s Already There
I wonder whether somewhere right now there’s a startup where everyone uses an AI coding agent as their primary work environment.
Not just the engineers. Everyone.
The finance person managing runway and board decks. The operations lead coordinating onboarding. The marketing manager producing content. All working with version-controlled prompts, testing frameworks and systematic automation – operating at 10x the speed of traditional companies.
I suspect those companies exist. I suspect they’re going to be terrifyingly competitive. And I suspect traditional organisations won’t understand what hit them.
(If you know of organisations doing this, I’d genuinely love to hear about it in the comments.)
The Memo-to-Email Moment
If this sounds revolutionary, that’s because it is. But we’ve been here before.
Go back far enough and internal communication was governed by strict hierarchies. Only senior people could send memos to the entire company because those memos had to be typed by the typing pool, photocopied and physically distributed. If you were junior, you simply couldn’t communicate broadly. The infrastructure didn’t allow it.
Then email arrived.
Suddenly, anyone could contact anyone. Hierarchies flattened. Information flowed differently. It fundamentally democratised communication in ways that transformed how organisations operated.
This is that same kind of shift.
Right now, automation and systematic workflow design are largely the province of engineers and technical specialists. Everyone else makes do with manual processes or waits for IT to build them tools.
But if instructions are now written in English and if platforms exist that let anyone manage those instructions with engineering-grade rigour, then suddenly everyone can automate their own workflows. Finance can engineer their reporting. Marketing can engineer their content pipelines. Operations can engineer their approval processes.
Not by becoming software engineers. By becoming Context Engineers and Prompt Engineers – managing knowledge and instructions with the same systematic discipline that code requires, but in their own domain language.
That’s the transformation unfolding right now.
Just like 30 years ago when email flattened organisational hierarchies, the organisations that recognise what’s happening early will define the next era of productivity. They’ll be the ones operating at speeds their competitors can’t match.
But understanding the paradigm shift is only half the battle. The real question is: how do you actually make this transition? What does it mean for your career? For your organisation?
That’s what we’ll explore in Part 2.
