A year ago, software developers were having heated debates about whether AI could ever really code. The arguments were familiar: machines don't understand what I've learned over 20 years. There's a human level of intelligence in design and architecture that can't be replicated. Vibe coding is dangerous because people are shipping things they don't understand.

A year later? It's pretty much accepted that machines can code better than humans. They make solid architectural decisions. They're faster, more accurate, and they can produce and test code at speeds no human can match.

But here's the thing nobody talks about enough: vibe coding still produces garbage. The difference isn't whether you use AI—it's whether you know what good looks like.

The Pattern That's About to Repeat

Consulting is maybe six months to a year behind software development on this curve. And consultants are starting to have the exact same debates developers had: surely an LLM can't produce the nuanced analysis I can. It doesn't understand the client context. It can't synthesise information the way an experienced consultant does.

They're wrong in the same way developers were wrong.

An LLM with the right input—strong opinions baked in, clear guidance on approach, institutional knowledge embedded in the context—will produce reports faster, better, and cheaper than any human can. The written output, the PowerPoint, the PDF analysis: it comes out better than what most consultants produce manually.

But vibe consulting? Where you just dump your problem into an LLM and expect a great solution? That produces exactly the same kind of slop that vibe coding does.

The Real Shift Isn't What You Think

For expert consultants, AI tools are a 100x multiplier. For consultants who don't understand what good work looks like, AI just produces garbage more quickly.

Same pattern. Same outcome. Same uncomfortable truth: if you can't tell the difference between good and bad output, you don't know if you're producing slop.

The human skills needed have fundamentally changed. Software developers went from memorising syntax and typing code to essentially becoming product managers—steering direction, understanding context, making judgment calls about what to build and why. The ability to debug line by line or learn library after library stopped being the core skill.

Consultants are facing the same transition. You no longer need to be brilliantly fluent in Excel or exceptionally good at writing reports from scratch. You do need to recognise the difference between good analysis and bad. You need to understand the bigger picture well enough to steer the AI toward solving the right problems.

A poor product manager can ship beautifully-built software that nobody needs. A poor consultant can produce polished reports that answer the wrong questions. The failure mode isn't bad execution anymore—it's bad direction.

What Actually Matters Now

The consultants who'll thrive are the ones who understand this isn't about whether to use AI. That question is already answered. It's about whether you have:

• Strong opinions on what good work looks like
• Clear frameworks for how problems should be approached
• The judgment to know when output is excellent versus when it's plausible-sounding nonsense
• The ability to steer toward the right problem, not just produce smooth deliverables

The same controls that smart developers put in place—strong architectural opinions, clear quality standards, expert review of AI output—consultants need to build the same systems.

The consultants who think their 20 years of experience makes them immune to this shift are making the same mistake developers made a year ago. The consultants who think AI means they can skip learning what good work looks like are making the mistake vibe coders made.

The ones who'll win? They're already baking their expertise into AI workflows and producing work that used to take teams of people.

The industry is about to learn what software development already knows: AI doesn't replace expertise. It exposes whether you had any in the first place.