Judge Maddox is Coming for Your Award, Should You Be Worried?
- 1 day ago
- 10 min read
Updated: 4 hours ago
Written by Annette Densham, Chief Storyteller
Annette Densham is a multi-award-winning Public Relations (PR) specialist and is considered the go-to for business storytelling, award submission writing, and helping business leaders establish themselves as authorities in their field.
In a near-future Los Angeles, Judge Maddox sits in judgement of those accused. In a court run by an Artificial Intelligence (AI) judge, a detective accused of murdering his wife has 90 minutes to prove his innocence or be executed. With a high probability of guilt before he opens his mouth, the race is on. Judge Maddox has zero mercy. This AI judge rules on the numbers it has access to and what it is fed. That’s it. It doesn’t fall for sob stories or read between the lines. It never questions what’s missing, distorted or hyperbole.
That’s the problem with AI. We treat what it spits out as the final word when it’s the first step. Handing high-stakes calls to a machine designed to process what it’s given, not what’s missing, is how you get a confident answer to the wrong question.
The awards world is facing this conundrum. What is the role of AI in awards? Should it be used to streamline what can be the onerous judging process? If it is used, can it tell the difference in an entry between something true and something well written?
AI strategist and educator Inbal Rodnay, who helps accounting, legal and advisory teams understand the tools, mindset and support they need to make AI work for them, said, “One of the toughest challenges with AI is there’s no textbook or clear guidelines for its use. Everything changes so quickly.”
“If awards are using AI for judging, they assume AI removes bias. It doesn’t. It just standardises it. If you start with a system that’s already 97% certain, the outcome isn’t really being judged anymore. It’s being confirmed. That’s the risk when we overtrust AI in any decision that matters.”
“What AI gives is speed and structure. What it can’t give you is judgement and nuance. AI can tell you what fits the criteria.”
Current state of awards judging
Before the awards industry gets precious about protecting the integrity of human judging, it might want to look at what that means in practice. I've been writing award submissions for almost 15 years. I've written thousands of entries across the A to Z of industries.
Most programs don’t brief their judges properly, if at all. There's almost zero transparency about the judging process. Criteria are so vague and generic that two judges scoring the same entry can come to two very different decisions. That’s not because one is wrong, but because what right looks like isn’t made clear.
Judges fill the gaps with personal opinions and biases. If you don’t win, you have no explanation why. Ninety-nine percent of awards programs give zero feedback.
When judging an award, I pointed out to the other judge how Caucasian her top five picks for finalists were. The interesting thing was that two-thirds of the entries were from other cultures. I suggested she had an unconscious bias. We received no training. Here are the entries. Off you go.
When there’s no training, guidance, rules or explanation of the criteria, the process becomes subjective rather than objective, like a Rorschach test. Every judge looks at the same entry but sees something different, shaped by their own experiences, assumptions and preferences. Not exactly fair.
Entrants want and need to know that the judging process has rigour. They need to trust that every entry is assessed against the same standards by judges who understand the criteria and recognise the biases they may bring into the room. So maybe the idea of introducing AI is a solution.
The algorithm has the gavel
Human judges aren’t as objective as we’d like to believe or assume. Human brains have this little room where their preconceived view of the world is locked away. Logically and consciously, we believe we’re being fair, but our brains are constantly making shortcuts, filling gaps and favouring what feels familiar without asking our permission. Like the ink in Mrs Marsh’s chalk test, “Ooohh, it does get in.”
That doesn’t make judges bad people. It makes them human. AI tools can help judges process large amounts of information more quickly, especially in complex entries. With hundreds of entries across multiple categories, that isn’t a trivial advantage. Speed, consistency and the absence of biased decision-making are appealing.
AI doesn’t care whether the submitter is well known, well connected or has their fair share of challenges. It can’t be charmed or swayed by a good sob story. It doesn’t score one entry more generously because the Chief Executive Officer (CEO) once sat on the same panel. If the criterion says “evidence of measurable impact” and the entry doesn’t have it, that’s that.
Awards bodies don’t need AI to be better than human judges. They need it to be good enough to clear a backlog of entries without chasing volunteers, managing conflicting scores or delaying the finalist announcement.
The argument for a hybrid option
The argument against AI in awards judging sounds reasonable until you sit inside the judging process. What do you do if two judges pull out at the last minute? You’re left with a stack of submissions needing hours of careful reading, scoring and written feedback. Judging is often a full marathon that many don’t have the time or energy for.
Maybe it’s not one or the other. Perhaps AI isn’t the threat to awards integrity. It might be a chance to enforce it.
Karen Perks, CEO of Mikare Health and Meeka Technologies, built an AI awards judging tool after living this scenario. She found AI could process and review a submission in two minutes, freeing human judges to spend 20 focused minutes digging deeper into the answers rather than decoding them.
“The knowledge embedded in the tool mirrors the same rubric matrix methodology used in tenders and grant assessments, where evidence against criteria is the basis of the decision. This AI is doing the reading so humans can do the thinking,” – Karen said.
Karen’s AI judging model takes the first pass, reads every submission against the criteria and produces a structured assessment. That initial pass flags where entries meet the evidence threshold, scores them against the rubric and gives written feedback to every submitter, regardless of whether they progress. Human judges then receive a shortlist, not a mountain of entries.
“The attention goes to the entries where the scores are high, where two submissions are separated by a fraction, or where something in the AI assessment warrants a closer look. The human brings contextual understanding, industry experience and the ability to see original thinking.”
“The AI handles volume and consistency. The human handles interpretation and final verdict.”
It’s not AI versus humans. It’s a division of labour where AI filters, structures and stress tests, and humans decide the winner.
The case against
Before handing entries over to Maddox, award platforms need to answer this: what’s the judging process there to do?
Is it there to clear 500 entries before the finalist announcement or identify the entrant whose work genuinely deserves to win? Those two goals can coexist, but when speed is driving the decision, Houston, we have a problem.
AI can process evidence in seconds, assess against criteria and pump out a verdict. What it can’t do is wonder whether it’s been told the whole story.
Awards judging has the same problem. Business awards aren’t marking a multiple-choice exam. Judges are tasked with deciding what excellence looks like across businesses of different sizes, sectors, locations, resources and stages of growth. A 20% increase in revenue can mean one thing in a startup and something different in a business that’s been around for 10 years. Saving five jobs in a regional town may have more impact than creating 50 roles in a capital city. A smaller result achieved with the odds stacked against you can say far more about leadership than a bigger one backed by a hefty budget.
The numbers matter, but there has to be meaning attached. A good judge looks at what happened, how it happened, and the challenges faced and overcome.
“We introduced monthly team meetings to improve communication” may sound like a leadership initiative, but unless those meetings solved a specific problem and produced a measurable change, it’s routine business as usual, dressed in a fancy frock to sound impressive. A judge can tell when an entrant has buried something remarkable under three paragraphs of clunky writing.
AI judges the version of excellence it’s been instructed to look for. Tell it to reward measurable impact and it may favour the entry with the most percentages. Tell it to prioritise innovation and it may reward the biggest idea. Tell it to assess storytelling and the best written submission has an advantage.
At that point, who is really judging the award? The AI, the person who wrote the prompt or the company that built the model?
CommsHero Awards’ 2025 experiment is a glimpse into this problem. A human and an AI judged four anonymous entries. The judging brief and criteria were fed into ChatGPT, Claude and Gemini. While all three came up with the same winners, how they got there exposed noticeably different lenses.
Gemini was generous towards ambition, scale and cultural change, even when detailed results were thin. Claude placed more weight on structure, purpose and ethical framing. ChatGPT was stricter and favoured hard metrics, replicable processes and technical logic. That’s a consistency trap.
An AI model can be completely consistent with itself while consistently applying an interpretation the awards body didn’t anticipate. We give AI the keys to the mansion and trust what it spits out. AI can get the same thing wrong in every entry, and because it does it consistently, no one may realise the scoring is flawed. AI doesn’t remove bias. It can standardise it.
Research into Large Language Models (LLMs) used as judges has found that their decisions can be influenced by factors such as where information appears, how the scoring prompt is worded and the fluency or apparent quality of the response. In other words, the machine can be nudged while appearing to make a purely evidence-based decision.
A polished entry may be easier for AI to process than one written by someone whose first language isn’t English. A long answer may appear more substantial than a concise one. A business using familiar corporate language may fit the model’s idea of leadership better than an entrant describing the same achievement in plain English. An entrant who understands how to write for AI may score better than one who did the better work.
We could end up rewarding the submission most compatible with the machine rather than the achievement most worthy of recognition.
Human judges have blind spots too. The difference is, in a well-thought-out judging process, another judge can challenge the score. The panel can debate an interpretation, revisit the evidence and ask why one entrant has been marked more harshly than another.
You can say to a human judge, “You’ve given this entry 62 and everyone else has given it 85. Talk us through why you marked it this way.”
What do you ask AI? It’ll give you a beautifully structured explanation, a table and a score out of 100. That doesn’t mean the explanation reflects sound judgement. An 87.5 can still be a subjective opinion.
The ethics question
The ethical concerns relating to AI use in awards judging need to factor in fairness, transparency, accountability, reliability, privacy and trust.
Who trains the model? On what data? Whose definition of excellence gets encoded? If AI is trained on past winning entries, it’ll replicate what’s already been rewarded, which is a good way to lock in bias. The innovative, unconventional, and first-of-its-kind entry would most likely be scored lower by an algorithm that has no prior pattern to match it to.
Which version of AI is being used to judge? Was it ChatGPT, Claude, Gemini, or a purpose-built system? Which version? What prompt was used? Were the entries assessed individually or against one another? Did the order change? Was the process tested before live entries were uploaded? Could the same score be reproduced six months later after the model had been updated?
Unless an awards body can answer those questions, it doesn’t have a transparent judging system. It has outsourced judgement and is no better than judging being done by humans with no transparent process.
Awards submissions can contain revenue figures, profit margins, client outcomes, intellectual property, staff information and deeply personal stories. Entrants deserve to know whether their material is being uploaded to a third-party AI system, how it’s being stored and who has access to it. “We used AI to make judging easier” isn’t a rock-solid data policy.
There’s also the accountability gap. In a legal setting, decisions need to be explained. The Judicial Commission of New South Wales argues that giving reasons helps make a decision transparent, accountable and easier to trust.
Awards shouldn’t pretend they’re exempt from the same basic principle. If AI marks an entry down, who explains why? Can the entrant or judging panel question how the criteria were interpreted? Can anyone see which evidence the system ignored or misunderstood?
That’s the black box problem. A score comes out, but the reasoning behind it may be impossible to properly examine. If no one can explain or challenge the decision, can the decision be called fair? A bad call is still a bad call.
AI is still in test tube mode
In awards, experiments are happening, but they’re exactly that: experiments. The technology is moving fast, but fast isn’t the same as ready, and ready isn’t the same as right.
Inbal said the problem isn’t that AI is useless or wildly overhyped. “AI is still an emerging technology. A lot of what’s on the market right now is still being tested, tweaked and adapted. Yet, it’s being positioned as a shortcut.”
“AI is powerful in pockets and fragile in others. While it’s impressive, it’s far less integrated than many people expect. It's an emerging capability that still needs judgement, literacy, integration and clear ownership.”
Inbal said AI isn’t ready for us to outsource accountability to a tool. “There’s the AI literacy gap that has to be factored in. It's not just mastering prompts or using the latest model. It’s understanding what AI is good at, where it’s risky, and when human judgement still matters.”
The movie Mercy shows, amidst the action scenes and ticking clock, the danger of confusing speed with judgement. We humans have little patience, and using AI to judge gives judges a reason not to spend more time assessing, digging, reflecting and analysing. What could go wrong?
What’s the answer?
AI has a legitimate role in awards judging. It can help with shortlisting at volume, flagging entries that fail to meet minimum evidence requirements, identifying inconsistencies in scoring across a judging panel and checking criteria compliance. AI can make the process more rigorous, but it shouldn't make it less human.
If an awards program is going to hand over its credibility to an algorithm still in its infancy, it has to factor all this in. Otherwise, it's in danger of becoming a scoring system. Who wants a photo with a scoring system?
Annette Densham, Chief Storyteller Multi-award-winning PR specialist Annette Densham is considered the go-to for all things business storytelling, award submission writing, and helping business leaders establish themselves as authorities in their field. She has shared her insights into storytelling, media, and business across Australia, the United Kingdom (UK), and the United States (US), speaking for the Professional Speakers Association, Stevie Awards, Queensland Government, and many more. Three-time winner of the Grand Stevie Award for Women in Business, gold Stevie International Business Award, and a finalist in Australian Small Business Champion awards, Annette audaciously challenges anyone in small business to cast aside modesty, embrace their genius, and share their stories.
