Looking for a needle in a stack of needles

  • In-house practitioners will expect high-quality data and analytics
  • Automation will replace many steps in today’s patent workflow, and will be embraced by clients
  • Improved analytics will lead to a greater focus on quality, rather than quantity

Compiling data is only half the story. The real value is in what you do with it – or, as one interviewee says: “The hardest thing with respect to intellectual property is the interpretation of data.” It’s here that tools using AI technologies such as deep learning and predictive analytics will play a greater role in the next few years. As the head of one major patent service provider points out: “Our fundamental strategy is that there are significant opportunities for automation.” Another adds: “People that provide good AI toolsets will make a lot of money,” while one head of intellectual property says: “We are crying out for it in-house.”

AI has the potential to speed up and reduce costs in many aspects of the patent process, such as prior art searching, docket management and renewals, as well as filing through the Patent Cooperation Treaty and even aspects of drafting applications. But it could go further too, wagers one service provider: “Large parts of the drafting process could be automated. You just need to find ways to take the value you bring of technical knowledge and translate it into words.” Another suggests: “AI will take off. It helps maintenance decisions and makes filing and prosecution strategy easier, as well as decisions about when to litigate and what the risk is in contracts. We will see that in the next five years.”

In the medium term, at least, AI is likely to supplement human brains rather than replace them in the patent field. It will enable professionals – whether in corporations, law firms or service providers – to focus more on sophisticated analysis and the nuances of each case, while the routine work is taken care of by a computer.

For example, one executive at an IP service provider notes: “We are increasingly using automation. There are algorithms to check the priority data and tools that can filter results before they go to specialists.” Another says: “When asserting a patent, you need eyes on a claim and to map it to a product. You can use tools to help with triage and teaching algorithms to learn – you can look at tens of thousands rather than thousands of patents, but it won’t replace humans.”

Take the example of docketing. “Full automation is years away – the risk of getting things wrong is too great,” argues one senior service provider. But he adds that the human-computer balance will change: today, a human gets data and documents, reviews them and updates their records; tomorrow, the software will recognise A, B and C and suggest a change to make. The human will then have to make the final judgement.

In-house clients will embrace such tools. One says: “I don’t see any boundaries as to what could get automated. We could see a development where automation goes quite far.” Another points to the big steps made in machine translations, which have gone from “nonsense” to “very good” in 10 years: “Maybe AI will be able to put a first draft of a patent for manual review – that would cut our costs and be welcomed.”

However, the investment required to train AI systems in the patent field is large. The question sets need to be highly specific and contextual, not just based on key words; patent searching is already moving this way but has much further to go. That is why these tools are likely to remain just that – tools to assist professionals. One service provider comments: “Patent information from various sources is easily programmable into computers. But you still need the human element to assess how it works for the company in a competitive sphere.”

One probable effect of AI is that it could drive a move from quantity to quality. “Analytics will make it much easier to show which patents are worthless,” foresees one executive. The volume and accessibility of prior art will increase dramatically; when we get to the point where every patent attorney can find prior art with a few taps on their smartphone, many patent applications are likely to be abandoned, but the ones that make it to grant will be higher quality. Some might fear this, but as one entrepreneur bluntly puts it, invalidation is the easiest way to find out the value a patent: “If a patent is invalid, its value is zero.”

Greater use of AI tools could therefore lead to abandonment rates increasing, and ultimately fewer patents being filed. As businesses are more likely to ask whether they really need 1,000 patents, tools that can sort through millions of documents to help make those decisions will be embraced.

“AI is an objective tool that stops you being the person responsible for making the wrong business decision,” says one provider – adding that often in-house counsel will choose to patent just because it is the safest option: “We have long conversations with clients about ‘should I patent it?’. Often we don’t think there’s any point in patenting it. Keep innovating and instead have a strategy if someone sues you.”

Another provider sums up how the landscape will change: “In the old days, you were looking for a needle in a haystack. In the future, you will be looking for a needle in a stack of needles.” And to those sceptics who say AI will never replace human skills, he retorts: “AI can sort through 20 million documents. It’s not 100% accurate but guess what – neither are you.”

Unlock unlimited access to all IAM content