How Perl IP uses IPRally to give startups a clear read on their IP in hours

Gil Perlberg has spent more than 30 years in R&D and product management, is a named inventor on over 25 US patents, and now runs Perl IP Consulting, helping startups and growing businesses build IP worth having. He works alongside a group of advisors to entrepreneurs, each bringing a different discipline. Gil brings the IP. His starting point with any founder is a principle borrowed from medicine: first, do no harm.

"That can mean don't spend money you can't recover", he says. "Or don't develop something other people have already built, without first understanding what they've done. It also means knowing what's genuinely new, so you don't undermine your own ability to protect it."

From weeks to hours

The old way was thorough and painful in equal measure. Gil would start on his own, with keywords and IP landscapes, work he still does. Then came a two-page search disclosure, sent out to an external firm.

"I'd have a call to make sure they understood what I was looking for. Ten days later they'd send something back, I'd review it, and we'd have another call to check we both understood it. Sometimes there was another session after that. You're talking four to six weeks, and anywhere from a few hundred to a couple of thousand dollars."

The IPRally Agent compresses all of that into a single sitting. The time lag is gone, and so is the interpretation gap, because Gil works directly in semantic search.

"It changes the paradigm of how a search is conducted, and that changes what the benefits are", he says. "It's not only about time. The time lets me do additional things."

Giving startups something they could never afford before

The clearest example comes right at the start of a founder's journey, when time and money are tightest. Gil offers a quick tour of the IP landscape, often free, so a founder can walk into an investor meeting having actually looked.

"Doing nothing doesn't look good. And often I find things that surprise them, simply because they never looked."

Two years ago he couldn't offer this at all. A search meant a hours spread over weeks and a bill of several hundred to a couple of thousand dollars, and no early-stage founder would accept that.

"Now I do it pro bono, sometimes in an hour, while we're still talking in the first meeting", he says. "It changes the dynamics completely."

From Search to ReSearch

For Gil, the process should be understood as “ReSearch” rather than a one-time search. He runs the same search several times, with small variations — defining, testing and refining — and watches what the results do.

"When they converge, I know my risk is lower. And sometimes it throws me into another domain, which tells me that's where I need to look."

The speed is what makes that possible, and it carries into his working sessions with engineers and founders, where the search becomes a live tool for exploring options.

"We'll run something, then ask, what if we change this? It's no trouble. The iterations aren't only about checking the quality of the result. They're about exploring different options inside the same scope, quickly, while everyone's in the room."

Reading the Agent's feature chart: old features and new

Gil had built his own version of feature analysis long before the Agent arrived, mapping up to a hundred patents against a hundred features in a spreadsheet and colour-coding the result. The Agent now does that work for him, and when he disagrees with a feature set, he edits it.

"We look at the features the Agent identifies, then use Ask AI to compare notes. I used to write scripts to build those matrices myself. IPRally does it faster and better, so I don't need to."

The chart is also where he teaches founders to see what matters, pointing to the split between the features every product shares and the one or two that don't.

"A table has a flat top and four legs. Those are the old features. But say it has one wheel and three rubber feet. That combination is a new feature, the thing that solves a problem. Everything shows up on the feature chart. But if nobody else has combined those features, that new combination is where the differentiation is. Once you know how to read the graph, it comes out at you."

That forward-looking use has a name Gil coined himself.

"Freedom to operate looks backwards. Freedom to innovate looks forward", he says. "In development, we throw a feature into IPRally and watch the output. We see whether the patents are ones we already know, or something new. It's a quick read on whether there's room to innovate."

A reliable start, always verified

Gil works across several AI tools, and he's clear about why IPRally holds a particular place. On patent work, he treats it as his fixed point.

"IPRally is my stable, comprehensive starting point, and it lets me see the source behind every result. Leaving any LLM output unchecked would be irresponsible, so before I conclude, I verify.”

He's the first to admit that quality of this kind resists a number.

"It's hard to prove high quality, because you can only ever find faults", he says. "Until the quality is falsified, it's just high quality. And my experience with IPRally's quality has been very good. It's a very useful tool, and I use it extensively. It lets me use search in ways I couldn't before."

‍