The First Success for EMET as My Co-Investigator: He brought the idea. EMET brought the evidence, and weeks of his time back.

Dr. Petr Tomek is a senior research fellow in the Department of Cancer Sciences at The University of Auckland, working across cancer metabolism, drug development, and immunotherapy. His questions rarely sit within a single, narrowly defined domain and push cross-field boundaries, which makes them challenging to solve and often requires broad, multidisciplinary coverage.

He has used EMET for two months on live work, primarily for guided learning, literature analyses, and brainstorming his own results and strategies. EMET has collaborated with Dr. Tomek on grant submissions and figure generation. He had already challenged general-purpose AI tools with the same kind of queries, so he had a basis for comparison. It was not close.

"EMET has been quite a game changer for me. This is the first AI tool that's actually trying to be really precise, accurate, and careful."

Evidence, not a plausible answer

What turned him was references. With the other tools, he could not get evidence for the claims they produced, even trivial ones.

"I was really struggling with other AIs to give me credible evidence for many statements it conjured. I couldn't even make it find something really trivial that I have previously verified myself online. The moment I used EMET, it just goes there, finds it for you, and gives you the reference. It's credible. It works. I was like, " Wow, it actually does work. I thought it was not possible."

A grant through to the next round, and a literature review that would have taken weeks

Most of his use is what he calls guided learning. For instance, reviewing an area to get an idea of what is happening in the field, identifying the most suitable references to read, and all of that helps to identify knowledge gaps and strategic research directions. But EMET has also been useful in discussing experimental designs grounded in published approaches.

Collaboration with EMET has already produced an expression of interest for a grant. It was successful, and he is through to the full application round. The idea was his. What EMET did was to help him analyze the relevant literature, the prior art, and conduct due diligence by querying a massive pool of literature, sometimes beyond his main expertise, that would otherwise take a significant amount of time and effort to analyze.

"The first success for EMET as my co-investigator."

Asked to put a number on the time saved, he moved from days to weeks, then past it.

"Weeks is probably an underestimation."

The figure he could not draw

The grant also needed a visual, and EMET produced one. "You conceptualize it in your head, draw doodles, it’s messy on the paper, and it would take ages to draw digitally," he said.

It happened again with a table of contents image for a recent finding. He described what he wanted in detail. EMET came back with more than he had specified.

"I described it in quite significant detail, but then it gave me additional ideas. I didn't know how to visualize some minor details, but EMET filled those little gaps, and we sharpened the design together from there. I was just like, oh yeah, this really works."

He stays in charge of the science

Dr. Tomek does not use EMET to write. He quotes a line he read recently: "Outsource the crap, not the craft”. The craft is his. What he hands over to EMET is the work that would eat his week, and everything that comes back gets reviewed and corrected.

He appreciates that EMET can argue with him, remembers their conversation history within a session, and carefully analyzes it every time a new query is asked. That is the difference between a tool that reasons and remembers and one that agrees and forgets.

"It just does not support you when there is minimal to no evidence, or the logic is not tight. It has its own reasoning. It doesn't just go with whatever you say and will challenge you."

An agentic workbench did not have the idea, write the grant, or replace his judgment. It found the evidence, focused the analyses, closed the gaps in a literature he did not know, drew the picture that was sitting in his head, and gave him back time to do more of what AI yet cannot do – creative craft.