Procedural Fairness When One Party Has AI (And the Other Doesn't)
- Shiv Martin
- 2 days ago
- 5 min read
As AI-generated complaints, submissions and legal arguments become increasingly common, complaints handlers, conciliators, tribunal members and Ombudsman staff face a new challenge: how do you maintain procedural fairness when one party arrives with AI support and the other does not?
The new fairness question
A question is starting to come up in conversations with complaints handlers, conciliators, tribunal members, and ombudsman staff that I haven't heard before:
"What do I do when one party has AI and the other doesn't?"
It's a fair question. And it doesn't have an easy answer.
A complainant arrives with a thirty-page AI-drafted submission, legal framing, case citations, and clear procedural asks. The respondent arrives with a one-page email written in their own words, asking what's going on. Or the reverse: a corporate respondent with AI-assisted legal support sits opposite an unrepresented complainant who is doing their best.
Both situations raise the same underlying question. When access to AI changes the apparent capability gap between parties, what does procedural fairness actually require?
The temptation to overcorrect
The first instinct, often, is to compensate. To work harder on behalf of the less-resourced party. To soften the questions for one side and sharpen them for the other. To level the playing field by adjusting how the decision-maker engages.
This is well-intentioned. It can also undermine the fairness it's trying to protect.
Procedural fairness isn't about producing equal outcomes. It's about ensuring each party has a fair opportunity to be heard, to understand the case against them, and to respond. The presence or absence of AI doesn't change that obligation. But it does change how teams need to think about it.
What's actually different
A few things are genuinely new:
Apparent capability can mislead. A polished submission doesn't mean the person understands their case. It means they have access to a tool. Staff who assume capability based on the document risk missing where the person actually needs support.
Verification work falls unevenly. When one party submits AI-generated content with fabricated citations or confident-but-wrong legal framing, the burden of correcting that falls on the decision-maker and on the other party. That's an asymmetric cost that didn't exist when both parties were limited to what they could write themselves.
Expectations are shaped before contact. People arrive with AI-generated understandings of your process, your role, and likely outcomes. Some of those will be accurate. Some won't. Either way, the conversation starts further along than it used to.
Self-representation looks different. Previously, an unrepresented party might have been recognisable by their plain language and uncertainty. Now they may sound legally fluent. That changes how staff read the room.
If you're interested in the broader impact of AI on complaints systems, you may also enjoy my article AI in Complaints and Disputes: The New Reality.
What hasn't changed
Almost everything else.
Parties still need to understand what's happening and why. They still need a fair opportunity to respond. Decision-makers still need to ask the questions that matter, hear from both sides, and explain their reasoning. The principles of procedural fairness are intact. The tools each party brings to the conversation are different.
The risk isn't that AI breaks procedural fairness. The risk is that staff feel pressured to adjust their approach in ways that do break it. To soften questioning because the person seems vulnerable. To accept AI-generated assertions because they're presented confidently. To treat unrepresented parties as more capable than they are because their submission sounds polished.
If your team is grappling with AI-generated complaints, mismanaged expectations, or growing pressure on fairness and trust, I support complaints bodies, regulators, tribunals and Ombudsman offices through training, facilitation and system design advice. You can learn more about how I can support here: Ai in Complaints
Practical guidance for decision-makers
Five things help:
1. Read substance, not source. Assess what's actually in the material. Are the facts verifiable? Are citations real? Is the issue in jurisdiction? Is the outcome being sought one your process can deliver? These questions apply whether the submission was AI-drafted, lawyer-drafted, or hand-written.
2. Ask the same direct questions of everyone. A polished submission doesn't deserve gentler scrutiny. A handwritten letter doesn't deserve harsher scrutiny. The questions are the questions: what happened, when, who, what evidence, what outcome.
3. Use early human contact to test understanding. A short phone call surfaces quickly whether a party actually understands their case or whether they've been carried by an AI tool. This isn't about catching anyone out. It's about ensuring the person can genuinely engage with the process.
4. Be transparent about verification. Tell parties upfront that submissions may be checked against original documents, and that factual claims and citations may need to be supported. Set the standard openly so no one is surprised.
5. Protect dignity in both directions. Don't assume the AI-assisted party is acting in bad faith. Don't assume the unassisted party is less capable. Treat both with the same respect and the same procedural rigour.
The deeper point
Procedural fairness has always required us to look past presentation and into substance. AI just makes that requirement more visible.
The polished submission has been around for as long as lawyers have. So has the unrepresented party doing their best with limited resources. What AI has done is democratise the polished submission, which means decision-makers can no longer use polish as a shortcut for understanding capability.
That's not a problem. It's a return to first principles. Look at what the person is actually saying. Check whether it's true. Ask the questions that matter. Give both sides a fair opportunity to respond. Explain your reasoning.
This is what dispute resolution has always required. AI just makes it harder to skip steps.
Free guide: Responding to AI-Generated Complaints
A practical intake tool for complaint handlers, conciliators and intake officers working through AI-assisted correspondence. Focus on substance, ask better questions, and progress matters fairly. Download the guide here.
A note on supporting staff
This work is harder than it used to be. Verifying citations takes time. Untangling AI-generated framing takes time. Working out whether a party truly understands their case takes time.
If your team is feeling the pressure of this, that's a signal, not a personal failing. The work has genuinely shifted. The training, resources, and procedural support that worked five years ago may not be enough now.
Investing in staff capability, in clear public information, in good auto-responses, and in process design that builds in early human contact isn't a luxury. It's how organisations protect fairness, manage workload, and keep good people in the work.
Curious about my training, speaking and policy supports in this area?
👉 Book a confidential conversation to discuss what an AI-ready, human-centred response could look like for your organisation.
My next live course in this area is this June! Sign up before June 30 to use this financial years budget! Find out more here.
Hi, I’m Shiv Martin. I’m a nationally accredited mediator, lawyer, conciliator, and conflict management specialist with over a decade of experience working across government, business, and community settings. I support teams to navigate complex and emotionally charged situations through mediation and conciliation, conflict skills training, facilitation, and practical advice on policies and processes. My approach is grounded in law, psychology, and real-world dispute resolution, with a strong focus on clarity, fairness, and workable outcomes.
Shiv Martin Consulting offers a structured three-level professional development pathway for dispute resolution and regulatory teams.
• Level 1 – Core Dispute Resolution Skills: For new starters or professionals working in complaints, case management or early resolution roles.
• Level 2 – Managing Challenging & High-Risk Interactions: For experienced conciliators, mediators, complaints managers and regulatory officers.
• Level 3 – Community of Practice for regulatory professionals: For experienced staff committed to reflective practice and continuous improvement.
All 3 levels can be offered in-house, simply email us to discuss your specific needs and we can tailor training to suit your team.
Comments