Powerful AI Training for Researchers: RMIT’s Proven 195% ROI

Case Study · Higher Education & Research

From ‘Generation Tool’ to Colleague

RMIT School of Psychology  |  AI Training  |  March 2026

AI training for researchers at RMIT University's School of Psychology — Real Minds AI case study
AI training for researchers works best inside their own domain. Real Minds AI delivered 8 domain-specific, methodology-first workshops for RMIT University’s School of Psychology: 31 academics, 32 hours over 8 weeks (5 February – 26 March 2026), investment $12,800. Researchers shifted from using AI as a “generation tool” to working with it as an error-checking colleague. Documented outcomes: a data-preparation task cut from 8 months to 20 minutes, 200+ pages of reusable research briefs and teaching materials, 3–5 core adopters running 5+ sessions each, and a conservative first-year ROI of 195%. Russell Conduit, Associate Dean (Psychology), attended 7 of 8 workshops and is a named, verified referee. Delivered by Dr Dennis Wollersheim (PhD Computer Science, 53 publications) and Tracy Anthony of Real Minds AI, Melbourne.

At a Glance

ClientRMIT University — School of Psychology (31 academics)
Engagement8 weeks — 5 February to 26 March 2026
Format8 × 4-hour hands-on workshops (32 hours; ~180 person-hours delivered)
Investment$12,800
Core adopters3–5 researchers, 5+ sessions each
Legacy200+ pages of reusable research briefs, handouts, decks & transcripts

“I used to use AI to just generate stuff or search for stuff. But now I see it as almost like another colleague where you can bounce ideas off and evaluate and error-check. That’s been the biggest enlightenment for me — I never thought before these workshops of using it that way.”

— Russell Conduit, Associate Dean (Psychology), attended 7 of 8 workshops

The Engagement

The Challenge

  • No domain-specific training — university AI sessions covered prompting basics, not psychology research methods
  • No verification culture — researchers either trusted AI blindly or avoided it entirely
  • No methodology integration — AI used as a search engine, not embedded in research workflows
  • 31 academics with scattered, inconsistent AI relationships across the school

What We Built

  • 8 workshops covering the full arc from mindset shift to reproducible pipelines
  • 11 research briefs synthesising 50,000+ words of peer-reviewed evidence
  • 8 slide decks, 8 exercise handouts, 8 clean transcripts and prompt template packs
  • A unified ResearchOps Loop: Ingest → Structure → Transform → Verify → Ship

The Impact

  • 8 months to 20 minutes — data-extraction code that took 8 months now runs in 20 min
  • 500-paper screen with zero errors across 200 spot-checked papers
  • Teaching multiplier — core adopters now training their own students
  • Conservative Year 1 ROI of 195%+ against the RMAI fee
195%+
Conservative Year 1 Recovery
Against the RMAI engagement fee
200+
Pages of Reusable Materials
Research briefs, handouts, transcripts
32
Hours of Hands-On Training
~180 total person-hours delivered

In Their Words

Effective AI training for researchers lives or dies on trust. These are research methodologists — trained to be sceptical, and that scepticism is exactly why the shift stuck.

When I redid the coding with a phenomenological perspective, it produced ‘existential exhaustion’ instead of generic themes. The framework is the prompt.

Marcel Takac
Lecturer · Workshop 6

I spent nearly a year doing it a couple of years ago — and it took 20 minutes to write the code.

Research methodologist
On AI-assisted data extraction · Workshop 5

It’s scary because it fabricates data. If you don’t know enough, you end up with a fabricated outcome.

Leila Karimi
Professor · Workshop 1

What Made This AI Training for Researchers Different

Domain-specific, not generic

Every exercise used real psychology research tasks — real transcripts, real ARC criteria, real PICO frameworks. Researchers learned to integrate AI into methodologies they had already mastered.

Methodology-first, not tool-first

We started with “here’s what your research needs,” not “here’s what the tool can do.” Epistemological limits came first; practical application followed.

Honest about failures

We demonstrated 20% citation error rates live. We showed emotion sanitisation in qualitative data. We triggered automation bias on purpose — building trust through transparency.

Materials as institutional legacy

200+ pages of briefs, handouts, decks and transcripts available to all 31 academics — including those who didn’t attend. The long tail extends years beyond the program.

The Bigger Picture

Most AI training for researchers teaches people which buttons to press. The shift that matters — the one Russell named — is moving from treating AI as a generation tool to treating it as a colleague you can reason with, challenge, and error-check.

You could see it in the final workshop. Russell brought a real publication folder — years of figures, paper drafts and scoring files with no readme and no way to reproduce the analysis. In about 15 minutes it was reorganised into a clean project skeleton with a makefile and the R code extracted into standalone scripts. He brought live clinical-trial data from a bulldog study too — raw Excel exports, no data dictionary — and watched it work out the variable meanings from context alone and link records across files. “Just think of how many academic hard drives have got folders like this,” he said. “For the cost of 15 minutes, you could just clean them up.”

That shift doesn’t come from a tool demo. It comes from working through your own real tasks, with someone who knows both the technology and the limits of trusting it. For a research school, the payoff compounds: the people who went through this AI training for researchers are now training their own students, and the materials outlast the engagement.

“The time I’ve spent with you, Dennis, has been gold. It really has. I’ve really appreciated it every single second. It’s opened my mind — I really didn’t understand how powerful it actually is.”

— Russell Conduit, Associate Dean (Psychology), Workshop 8

Training a research team or faculty?

If your people are stuck using AI as a search engine — or avoiding it because they don’t trust it — we run domain-specific, methodology-first AI training for researchers that changes how they work. One conversation is all it takes to scope it.

Let’s Have a Conversation or call 0480 032 916

Case study published with client approval. Full documentation and evidence available on request.

How We Work Proof Talk to us
How We Work Proof Talk to us
Ask us anything