Dr. Victoria Gillmore is a postdoctoral researcher at the University of Toronto. Her work centres on a sugar and its handling inside cells, and at the start of her postdoc she picked up a new protein, one carrying more than a thousand papers and a first publication going back decades.
She came to EMET with a basis for comparison. She had already tried other AI tools on real research tasks, including experiment planning, and she used EMET on live projects rather than test examples. Her summary was direct.
"I feel like it is the smartest AI I've used for science. I haven't personally had any hallucinations."
A thousand papers, and a renamed protein
Starting on an unfamiliar protein, Dr. Gillmore wanted its full history: the first publication, what people believed it did in the 1980s, and how the thinking changed. With over a thousand articles, that is work most researchers quietly skip.
EMET returned a timeline: when the protein was named, that it was first studied in cattle, and how the field moved from there. It also caught something she had no way of knowing.
"They changed the name of the protein. I never would have realized that, and I would have missed all of the studies that they had done on it in B cells."
Her estimate of the manual equivalent was months of reading and summarizing. EMET took about five minutes.
Experiment planning with a reference for every dose
Dr. Gillmore had previously tried planning experiments with another AI tool. The comparison was not close.
"It's just not the same at all."
With EMET, she got doses, entire plate maps, and a citation attached to every dose it proposed. It also raised the design questions a colleague would raise, such as testing three doses first, or building in a time course.
"It always suggested things that a scientist would think of."
The tool she had used before behaved differently. In her words, it was "more of a chat box," and, plainly, "it's not a scientist."
An animal use protocol that knew what the CCAC needed
She also used EMET to draft an animal use protocol, a document with specific regulatory expectations attached to it.
"I just used it to develop an animal use protocol, and I feel like it already knew what that is and what was necessary to include for the CCAC."
She described the overall experience in terms of the people around her in the lab.
"It's almost like talking to a research associate."
Challenging EMET to defend the field's consensus
Dr. Gillmore then tested EMET against something her field broadly accepts, that the sugar she studies enters cells by endocytosis. She asked it directly what the evidence was.
EMET traced the claim through the review articles that repeat it and showed that the original support came down to a single graph. It then proposed what would actually settle the question, including inhibitor work combined with microscopy.
"When I asked it the question, it was able to think critically about the data and then suggest what it needed to be convinced of."
That is the pattern worth noticing. The scientist sets the question, keeps her scepticism, and gets evidence she can trace back to its source. What a lab gets from an agentic workbench is not a shortcut around judgement. It is months of groundwork cleared away so the judgement can be applied to something that matters.
Want to explore what’s possible?
Try EMET
