TL;DR: Hiring consultants in the age of AI is becoming increasingly difficult as generative AI makes it easy for candidates to produce polished, technically correct answers without real-world experience. In consulting, the real signal is not textbook knowledge, but judgment under ambiguity, trade-offs, failures, and messy customer environments. If interview processes still reward memorised answers over applied reasoning, the hiring process itself is now the vulnerability.
Table of Contents
Are we building a generation of lyrebirds?
Hiring consultants in the age of AI has exposed a problem most interview processes weren’t designed to handle.
As generative AI makes it easier for candidates to sound experienced without real-world exposure, the signal we’ve relied on is starting to break down.
There’s a bird in the Australian bush that can mimic almost anything. Chainsaws. Camera shutters. Other birds. Car alarms. The lyrebird is so convincing that even other species get fooled. It doesn’t understand what it’s reproducing. It just plays back what it’s heard, perfectly.
Are we building a generation of lyrebirds?
Not the feathered kind, the kind that sit across from you in interviews and meetings, fluently reciting concepts about Zero Trust architecture, least privilege, or the shared responsibility model. Textbook perfect definitions, incredibly confident, and from what I am seeing, increasingly sourced direct from generative AI fifteen minutes before the call with little to no real-world experience.
How Generative AI is changing technical hiring
For most of the last two decades, technical hiring — the process of assessing what a candidate knows, how they implement it, and whether they can apply it in real environments — has leaned heavily on knowledge-based questions.
Which is why interviews leaned so heavily on knowledge-based questions.
- What does Active Directory do?
- Why do we use groups in Active Directory?
- What it the difference between HTTP and HTTPS?
Over a decade later these questions persist
- What’s the difference between authentication and authorisation?
- Walk me through how conditional access policies evaluate.
- What port does LDAP use over SSL?
These questions made sense when knowledge was hard to acquire. There were limited ways to acquire this knowledge
- You had to read the documentation
- Break things in a lab
- Study for a certification
- Work it out the hard way in production through painful deployments
You acquired some proverbial scars along with the knowledge you were seeking. Having the answer reflected experience because of the path required to get that knowledge.
That’s no longer true. Someone with no hands-on experience can now generate a fluent, structured, technically accurate answer to almost any knowledge question in seconds. Not through study or practice. Just a prompt and a few seconds of patience.
This isn’t a complaint about AI. It’s a structural shift.
The floor of baseline knowledge has dropped out. If everyone can sound like they know the answer, knowing the answer stops being a useful signal. That means not just knowing how something works, but being able to deploy it, troubleshoot it, and adapt it under constraint.
Why AI Mimicry isn’t real consulting competence
Here’s the uncomfortable part. A candidate who has used ChatGPT to prepare for your interview might genuinely sound better than someone with ten years of field experience.
They’ll hit the right keywords. Structure their answers cleanly and avoid the hesitation that comes from actually having done the work.
But real experience is messy.
What does that actually mean?
It looks like:
- Running a project end to end — including the cleanup
- Dealing with midnight incidents and near misses
- Realising the documentation doesn’t apply and working it out anyway
That experience comes with a “well, it depends on the environment” and “we tried that but the legacy integration broke it” and “the vendor said it would work but it didn’t.” Those rough edges, that is the signal you are trying to pick out from the noise.
The mimicry doesn’t come with questions and uncertainty. The seasoned practitioner does. They know that there is always more to the question.
This matters more in consulting
If you’re hiring into a product team or an internal support role, the lyrebird might survive for a while. They can lean on internal docs, pair with colleagues, and gradually build real understanding on the job. It’s far from ideal, but the blast radius is contained.
Consulting doesn’t work that way. You put someone on a Teams call and they land in a customer environment they’ve never seen before. They’re given partial documentation, a vague brief, and usually privileged access to systems that are someone else’s production. They’re trusted. A LOT. Often immediately.
That trust is the product. Consulting companies don’t sell software or widgets. They sell that messy experience and the judgement that comes with it.
When a customer brings in a consultant, they’re paying for someone who can walk into an unfamiliar environment and figure out what’s going on.
Not what the brief says.
Not what the previous vendor documented, or the sales team promised.
The real world is noisy.
Customers don’t always know what they need. They know:
- What hurts
- What an auditor flagged
- What another vendor told them six months ago
Half the job is untangling that.
Figuring out what the actual problem is — underneath the stated problem, and underneath the language shaped by previous vendors.
You can’t fake that. Someone can give you a technically correct answer about identity governance. But they can’t sit in a workshop, notice that the customer’s real issue isn’t their access reviews, it’s that three departments have different definitions of “contractor”, and redirect the conversation before the project heads off a cliff.
That’s consulting. A well-prompted response doesn’t get you there.
GenAI is a tool, not the job
This isn’t an anti-AI argument. It is not even a warning about it.
Generative AI is genuinely useful in consulting work.
A good consultant uses it to get up to speed on an unfamiliar technology in hours instead of days. They use it to draft policies, summarise logs, pressure-test their thinking, or generate a first-pass architecture to react to.
That’s real, practical value.
But the value comes from what the consultant does with the output, not the output itself.
If you’re paying consultants to read Microsoft documentation back to your team, the problem is not related to the documentation. Customers already have access to the docs.
Reading documentation back to a customer is not consulting. Applying documentation to a messy, political, half-understood environment is.
They’re paying consultants because the docs don’t have a clear section for their environment, constraints of their politics. Good consultants use GenAI to rapidly orient themselves.
We’ve written before about this thinking as ‘rubber ducking with AI‘, using it to test ideas, not outsource judgement.
This can look like:
- Using AI to sanity-check an unfamiliar identity flow
- Drafting a first-pass policy to discuss
- Pressure-testing a proposed change or design
- ‘Rubber duck troubleshooting’ to validate an approach
That chain of reasoning is what customers are paying for.
It’s communication, risk management, and contextual judgment layered together. It rarely has anything to do with the shiny tools. GenAI allows consultants to spend more time delivering their experience.
How to interview consultants in the age of AI
If knowledge questions are now trivially gameable, the filter has to shift. Not away from technical depth, but toward something harder to fake: the ability to apply knowledge in constrained, ambiguous, real-world conditions. The kind of conditions consultants actually work in.
The difference shows up in how candidates reason, not how they recall.
Here are a few approaches that hold up better.
What does a strong consultant actually sound like?
A strong candidate doesn’t just answer the question. They work the question.
They’ll rarely give you a clean, immediate answer. Not because they don’t know, but because they know it depends.
They don’t give you a single answer. They’re trying to get enough context to adapt the answer.
- “Before I answer that, what environment are we talking about?”
- “We tried something similar, but it didn’t behave the way the documentation suggested.”
- “I’d want to understand what problem we’re actually solving before recommending anything.”
If you change the scenario halfway through, they don’t get thrown. They adjust:
- “In that case, I’d probably change approach…”
- “If that system can’t support it, we’d need to either accept the risk or design around it.”
- “That constraint changes the trade-offs quite a bit.”
They’re comfortable changing their position, because they’re reasoning in real time, not recalling something they’ve rehearsed.
They also talk about things not going to plan. Not polished failures. Real ones.
- What broke
- What they missed
- What they’d do differently now
And importantly, those answers evolve under pressure. If you dig deeper:
- The story gets more specific, not more vague
- The details get clearer, not rehearsed
- The trade-offs become sharper
They don’t retreat into theory. They stay grounded in what actually happened.
A strong consultant will also challenge the premise without being combative.
If you give them a flawed scenario, they’ll test it:
- “Are we sure that’s the real requirement?”
- “That sounds like a solution to a different problem.”
- “What’s driving that decision?”
They’re not trying to win the interview. They’re trying to understand the problem.
And when they don’t know something, they don’t bluff.
They’ll say:
- “I haven’t worked with that directly, but here’s how I’d approach it.”
- “I’d start by validating X, then confirm Y with the customer before making a call.”
They describe a process, not an answer.
That’s the difference.
The lyrebird gives you an answer that sounds right. A strong consultant shows you how they think, how they adapt, how they question, and how they recover when the textbook doesn’t apply.
Scenario-based questions with trade-offs
Don’t ask what conditional access is, ask what happens when a conditional access policy conflicts with a legacy application that the business refuses to modernise.
Ask what they’d recommend, and what they’d sacrifice. There’s no clean answer. That’s the point.
Failure-based questions
Ask about a time something broke. Not a hypothetical, a real one.
- What went wrong?
- What they missed?
- What they’d do differently?
AI can generate a plausible-sounding failure story, but it falls apart under follow-up questions.
- What did the error look like in the logs?
- Who pushed back on your fix?
- How long did it take to resolve?
These are details that only lived experience can fill in.
The customer-behind-the-customer question
Describe a scenario where the customer has asked for one thing, but the underlying problem is something else. A customer says they need a SIEM.
What they actually need is visibility into what’s happening on their network because an insurer asked them a question they couldn’t answer.
The candidate who jumps straight to product comparisons is telling you something. The one who asks “why do they think they need a SIEM?” is telling you something different.
Second-order thinking
Present a proposed change and ask what else it affects.
We’re going to enforce MFA across all accounts next Monday. What happens?
The AI-prepped candidate will list the benefits. The practitioner will ask about service accounts, break-glass procedures, the accounts payable team who haven’t enrolled yet, the third-party vendor with a shared mailbox that’s been quietly humming along since 2017 or the director with a personal device for “a reason”.
Process over product
Ask how they would figure out the answer to something they don’t know. Everyone encounters unfamiliar problems. The difference between a capable consultant and a lying lyrebird is what happens next.
- Do they know where to look?
- Do they know what questions to ask the customer?
- Do they know how to ask questions when they are not getting the answers they want?
- Do they know when to stop and escalate before making things worse?
The real skill is knowing when the textbook is wrong
Generative AI is excellent at producing the correct answer for a general case. It’s poor at knowing when the general case doesn’t apply. That gap, between what should work and what actually works in a specific environment, with specific constraints and specific humans involved, is where consulting experience lives.
A candidate who tells you “The documentation says X, but in practice I’ve seen Y because of Z” is telling you something an AI can’t.
They’ve been in the room. They’ve dealt with the fallout. They’ve had the conversation at 4pm on a Friday when a change broke something that wasn’t in scope. They’ve fixed that problem when the documentation only tells half the story.
That kind of judgment doesn’t come from prompting a model.
It comes from doing the work, getting it wrong, and learning what matters when things are messy and someone else’s business is on the line.
Don’t blame the lyrebird
None of this is the candidate’s fault. If the hiring process rewards polished recall, people will optimise for polished recall. That’s rational.
The problem isn’t that candidates use AI to prepare. The problem is that our filters haven’t caught up.
If your interview process can be defeated by twenty minutes with a chatbot, the process is the vulnerability, not the candidate.
Where this lands
Hiring in consulting has always been an imperfect signal. Generative AI hasn’t created that problem, it’s exposed how much we were already leaning on shallow indicators and gut feeling.
Knowledge questions felt rigorous because they were hard to answer without study. Now they’re easy to answer without understanding.
The consultants who deliver value aren’t the ones who know the most facts. They’re the ones who can walk into a room, listen to what a customer is actually saying, cut through the noise left behind by previous vendors, and figure out what needs to happen next, including the parts that are uncomfortable to say out loud.
GenAI can accelerate that person. It can’t replace them completely. And it definitely can’t simulate them across a table in an interview, if you know what to look for.
What can you do about this?
The shift isn’t dramatic. It’s just overdue. Look for judgment over recall. Rough edges over polish. And conversations that a language model can’t rehearse for, because they depend on things that actually happened, in environments that actually existed, for customers with real problems and real constraints.
The lyrebird is impressive. Beautiful, even.
But it doesn’t know why the kookaburra laughs.
You want the person who does.
Key takeaways
- Hiring consultants in the age of AI requires moving beyond traditional knowledge-based interview questions.
- Generative AI can produce technically correct answers, but it cannot replicate real consulting judgment developed through lived experience.
- The strongest consultants demonstrate reasoning, trade-off analysis, stakeholder awareness, and decision-making under ambiguity.
- Scenario-based, failure-based, and reasoning-led interview techniques are more effective than recall-focused technical questioning.
- Real consulting experience often appears through nuance, uncertainty, and lessons learned from complex customer environments.
- Good consultants use generative AI as a tool to accelerate learning and validate thinking, not to replace judgment or accountability.
- Technical hiring processes that reward polished recall over applied experience are increasingly vulnerable to AI-assisted interview preparation.
- The future of consulting interviews will depend less on memorised answers and more on the ability to navigate messy, real-world problems.
If your organisation is navigating complex technology decisions and wants consultants who can operate beyond the textbook, we’d love to help.
At Arinco, we focus on practical, real-world outcomes grounded in experience, judgement, and modern capability.
