When former AI director at Tesla, Andrej Karpathy, coined the term ‘vibe coding’ this year – “where you fully give in to the vibes; embrace exponentials and forget that the code even exists” – he may not have expected the term to bleed into analyst reports or enterprise strategy.
Yet Gartner’s recent research article, ‘Why vibe coding needs to be taken seriously’, estimates that 40% of new enterprise production software will be created with vibe-coding techniques and tools by 2028.
It’s clear why novices have flocked to vibe coding. With easy-to-use AI tools, vibe coding is fast and simple and realising fully fledged projects beyond the user’s own technical expertise can be empowering. Hobby builders or weekend coders are sharing their success stories, like the father and daughter duo who crafted a website together in six hours.
Businesses are also eager to embrace vibe coding. Its champions highlight time-to-market advantages and the potential to help bypass a perceived skills shortage among developers. Swedish AI unicorn Lovable, for instance, claims users can create working products in minutes with GenAI. It has reported over $100m (£74m) in annualised revenue since its launch in November 2024. So some high-profile vibe coding wins are causing a lot of buzz.
But, equally, there are cautionary tales. Just look at software company Replit, which faced disaster when its AI model deleted the business’ entire database, with vibe coding the culprit.
Good guardrails and governance are essential
Despite the hype, vibe-coding is not enterprise-ready. There are simply too many risks and too much volatility for anything complex or built to scale. Businesses that depend on reliability need structure and planning, not “vibes”.
Enterprise software depends on robust financial systems, well-maintained critical infrastructure and vast amounts of high-quality data. Tech stacks must be well-governed and run by skilled employees who know what they are doing.
Some businesses are proceeding with vibe-coding initiatives, tempted by the perception of a shortcut to lower costs and headcount, keeping some humans “in the loop” where needed.
This is premature. AI-powered coding does create a speed advantage, but if the code is junk, then it also lacks the structure and safeguards that enterprises need.
LLMs excel at pattern-matching and generating plausible output, but they struggle to grasp context.
Even if some output looks like it follows best practice, LLMs cannot understand the intent behind it. This is why Replit’s AI saw no issue in deleting all of that data. Governance checks could have stopped the disaster from unfolding long before it was triggered.
This absence of governance and guardrails is a problem. Enterprises need to ensure all of their output aligns with business goals as well as ethical standards and regulations. Whether it’s protecting customer data or vetting third parties, vibe coding simply isn’t built to handle this level of responsibility.
And although AI can generate code quickly, fast doesn’t always mean good. While 76% of developers currently use or plan to use AI tools in their workflows, only 43% say they trust the results.
Blindly following AI hype shows a worrying lack of foresight or consideration for what humans bring to the table: strategy, nuance and accountability.
Enterprises should focus on building cohesive IT strategies designed for the long haul. There should be room for adaptability, allowing organisations to trial new tools or take advantage of the best that AI has to offer.
But the fundamentals are more important. Organise your data and update your software development lifecycle to include data engineers and end users. Invest in employees with sharp minds who are eager to evolve.
The temptation to churn out software faster with fewer people has caused short-term thinking in businesses – and short-term thinking limits long-term gains.
Raymond Kok is CEO at Mendix, a mobile application development company.
When former AI director at Tesla, Andrej Karpathy, coined the term ‘vibe coding’ this year – “where you fully give in to the vibes; embrace exponentials and forget that the code even exists” – he may not have expected the term to bleed into analyst reports or enterprise strategy.
Yet Gartner’s recent research article, ‘Why vibe coding needs to be taken seriously’, estimates that 40% of new enterprise production software will be created with vibe-coding techniques and tools by 2028.
It’s clear why novices have flocked to vibe coding. With easy-to-use AI tools, vibe coding is fast and simple and realising fully fledged projects beyond the user’s own technical expertise can be empowering. Hobby builders or weekend coders are sharing their success stories, like the father and daughter duo who crafted a website together in six hours.
