There's a failure mode spreading through consulting, engagement and social impact work, and it isn't artificial intelligence. It's artificial credibility. Reports, strategies, RAPs, engagement plans and procurement policies are being produced in minutes now, and they look the part: right structure, right vocabulary, right level of polish. Polish was never the hard part of this work. What's missing is the part that was always hard: context, relationship, and an honest account of what's true on the ground. There's a name for this online, AI slop, and it has started turning up in our sector wearing the clothes of strategy.

Fluency isn't understanding

A model can pattern-match its way to language that reads as culturally competent. It has never watched a community argue its way toward consensus, and it has never stood on a proposed infrastructure corridor noticing what the environmental report conveniently left out. It wasn't there, and it has no Country to be there on. Connecting with Country is not something you extract from training data. It's a relationship, built over years, with real consequences when you get it wrong. The risk was never that organisations would start using AI. It's that they'd lose the ability to tell fluency from understanding, or confidence from competence, a distinction this sector has struggled with long before AI turned up.

Looking right and being right are not the same problem

We've reviewed enough AI-drafted RAPs, strategies and procurement frameworks now to know the pattern. They read well until someone who actually knows the subject opens them: invented references, boilerplate recommendations that don't touch the specific community or supply chain in question, cultural framing that collapses under one direct question. The dangerous version isn't the document that's obviously bad. It's the one that creates the impression the hard work has already happened, when none of it has. A hundred and fifty pages generated in an afternoon can feel like rigour to a client who doesn't know better. It isn't. Length is not evidence of thinking. A well-formatted recommendation is not a tested one, and the difference between those two things is the entire job.

The cost doesn't show up in the document - It shows up after

This isn't a quality issue that stays contained to a PDF. Projects stall because the strategy underneath them was never grounded in what's actually true. Communities disengage because the consultation felt transactional, because it was. Procurement decisions weaken because the recommendations behind them were never stress-tested against anyone who'd have to live with the outcome. Reconciliation Australia sends it back. The panel rejects it. The community calls it out in the room. And the client pays for it twice: once to produce the document, and again later to fix what the document got wrong the first time.

We use AI - We don't outsource the thinking

Tika EQ uses AI daily, to move faster through information, pressure-test our own thinking, sharpen a draft, clear the repetitive work that doesn't need a person doing it. What it does not do, and will not do, is the work that actually constitutes this job: listening, building relationships, walking Country, testing ideas with community, understanding the politics and history sitting underneath any brief we're handed. Good consulting was never about producing a document. It's about helping people make a better decision than the one they were about to make. AI can accelerate that. It cannot do it for you, and it cannot be held accountable when it's wrong.

The question that actually matters

Stop asking a consultant whether they use AI. Everyone does now, or will shortly, and the answer tells you nothing. Ask what happened after the AI stopped typing. Who challenged the assumptions it generated. Who tested the recommendations against the people who'd actually be affected by them. Who spoke with community, and understood the politics well enough to know which recommendations would survive contact with reality. Who's prepared to put their name to the advice and stand in the room when it's questioned. That's where the value in this work has always sat, and no amount of generative capability moves it.

AI can generate words. It cannot generate trust, and it cannot take you from intent to impact on its own. That still gets earned the same way it always has: one relationship, one hard conversation, one piece of tested advice at a time.