The Menu Changes, Not the Price
The Menu Changes, Not the Price
A Few Thoughts on Patent Prosecution in 5, 10, and 15 Years, for In-House and Outside Counsel
At a recent conference I attended, a panel was asked what patent prosecution might look like in five, ten, and fifteen years, for both in-house and outside counsel. I have some thoughts. Predictions in this area should be made with some humility. Roy Amara gave us the right caution: we tend to overestimate the effect of a technology in the short run and underestimate it in the long run. With AI I would add a wrinkle. The short run may be underestimated too. The deep changes are likely to arrive faster than much of the profession expects, even if not as fast as the loudest voices promise. The honest answer is that we are early, and that the shape of the next decade is still being decided.
One prediction is already common. As AI models improve, the price of patent preparation and prosecution (prosecution being the back-and-forth process of moving an application through the patent office to an issued patent) may fall by double-digit percentages over the coming years. The efficiency gains behind that prediction are real, and we are beginning to see them, particularly in the work of putting words on the page. We are also only at the start. Much of the surrounding work still takes the time it always has, and a good deal of it takes longer than people assume. The gains will grow, and I expect them to grow substantially. We are not there yet.
Even when those gains arrive in full, I think the prediction misreads what is happening. It assumes the menu stays the same. If we keep offering exactly the same items, then yes, the price of those items should come down. That is not what I see happening. The menu itself is going to change, and what it changes toward is higher-quality patent applications. We will not keep selling the same application for less. We will sell a better one. The first thing AI changes is not the price. It is the pressure on the people doing the work, and after that, the menu itself.
Back to the future
Begin with the people, because that is where the change lands first. For twenty to thirty years the economics of patent preparation have been quietly unsustainable. Prices stayed essentially flat. The demands did not. The amount of information that has to go into a competent application kept climbing, pushed by courts and patent offices around the world that expect ever more detail and ever stronger support, and client expectations rose alongside it. The time available to meet those demands kept shrinking. Something had to give, and what gave was the practitioner. The result has been relentless pressure on attorneys and agents, real and rising burnout, and fewer people choosing to enter the profession at all.
AI changes that equation by giving time back. When it absorbs the routine, it relieves a pressure that had become genuinely hard to carry, and it lets attorneys put real thought back into the claims and into the quality of the work. In that sense the near future is less a leap forward than a return. I remember practicing in the early-to-mid 1990s, when it was still possible to spend real time on a deep analysis of the prior art before drafting a single claim, to sit with the references and understand the field before committing to a strategy. That kind of care grew harder and harder as budgets stagnated and volumes climbed, until for many matters it simply was not possible. If AI lets us get back to it, this is a back-to-the-future moment for the profession: the craft of the 1990s, practiced with the tools of the 2030s.
The menu changes, not the price
With that time returned, the work product itself changes. Attorneys can think harder about the claims and use AI as a thought partner to test directions, pressure-check scope, and sharpen language. Let me be concrete about what a stronger claim set means, because it is not a single thing. It means claims that are better aligned with the commercial and strategic value of the invention, so the protection tracks what the business actually cares about. It means claims that are better differentiated over the art, because we have had the time to read and digest that art. The same tools let us analyze a broader and more relevant body of prior art for novelty than was ever practical to review by hand, and to harden the application against both Section 103 (non-obviousness, whether the invention is a real step beyond what already exists) and Section 101 (patent eligibility, whether the invention is even the kind of thing a patent can cover) while it is still being written, rather than discovering the weaknesses years later in an office action (the formal written objections a patent examiner issues).
Quality of this kind begins well before the drafting. It starts with a careful review of the disclosure materials and real conversations with the inventors, the parts of the job that have been squeezed hardest and that matter most. It runs through fewer or no Section 101 problems and a strong written narrative for subject-matter eligibility in the technology areas where that fight is real, and into claims positioned for the continuation applications that will matter later. It begins even before the decision to file. A technology can be eminently patentable and still not worth a patent. One of the more valuable things AI frees us to do is to assess, at the disclosure stage, whether a filing carries enough strategic and commercial value to justify itself. If an invention offers little injunctive leverage, limited damages potential, and thin licensing or cross-licensing value, the right call may be to forgo the filing even though the legal case for patentability is clean. Deciding well what not to file, and saying so clearly, is part of the better menu.
Higher-quality input tends to produce higher-quality output, and over time that carries a cost dimension worth stating honestly. A patent that is stronger as filed, built on a wider novelty analysis and hardened against the predictable rejections, may need fewer rounds of prosecution and less back-and-forth with the USPTO (the U.S. Patent and Trademark Office). That can lower the total cost of reaching an issued patent even when the value of the work itself holds steady. It is a long-run, quality-driven effect, not a discount available today. It is also the opposite of commoditization.
For in-house counsel: the Jevons surprise
The forward-looking picture for in-house counsel runs against the common assumption. The efficiency gains are real, but the idea that they will substantially reduce the in-house workload is, I think, mistaken. The workload is going to grow, and the reason is worth sitting with.
In 1865 the economist William Stanley Jevons observed that using coal more efficiently increased total coal consumption rather than reducing it, because efficiency made coal worth using in more places. Jevons’ Paradox is coming for patent work. The velocity of everything that drives patent filings is increasing. Product development cycles are compressing hard, which means more inventions and more technologies arriving to be assessed, and arriving faster than before. Efficiency per matter will rise. The number of matters will rise faster. Net, the volume of work in front of in-house counsel goes up, not down.
This is why I am cautious about the in-house instinct to bring drafting in-house. It tends to underestimate the job. Putting words on the page is only a fraction of what it takes to prepare and file a strong application. The rest is inventor meetings and the scheduling of them, the collaborative work of drafting with the inventors, the rounds of revision and review, the communications, and a long tail of administrative work that has to happen before a filing can exist. Those tasks are being automated and will keep getting more efficient, but they do not disappear, and their volume is climbing with everything else. My expectation is more outsourcing of patent preparation and prosecution over the next decade, not less. Any tool vendor that models this work as text generation is measuring the smallest part of it, and underestimating, often badly, how much of the real effort lives everywhere else.
There is a basic arithmetic point underneath all of this that is easy to miss. AI speeds up the drafting, the generation of text. It leaves much of the surrounding work untouched or at least very much less impacted. Opening the file, carefully reviewing what the AI produced, updating the document in response to comments from the inventors and from in-house counsel, and the mechanics of filing the application all still take an underestimated amount of time. Reviewing AI-generated content with the necessary care is itself real work, and it is work the AI creates rather than removes. The consequence is that a saving of, say, thirty percent on the drafting does not translate into a thirty percent reduction in the total time a matter takes. Drafting is one slice of the work, not the whole pie, and the headline efficiency number quietly assumes otherwise.
For outside counsel: a new deliverable
On our side of the table, a new deliverable is already emerging, and it is a direct consequence of the same trend. As in-house teams use AI to review the applications we send, the review itself produces more comments. AI-generated comments that are not contained or particularly well-directed can be voluminous, and the work of responding to them, and of explaining the strategic reasons behind particular choices, can quickly become overwhelming for outside counsel.
Our response has been to produce, alongside the specification, a complementary strategic document that sets out the reasoning behind the key decisions, particularly on the claims. It pre-empts a great deal of the AI-generated review traffic by answering the questions before they are asked. It also does something more durable. It preserves the strategic thinking for later, for the humans and increasingly for the AI agents that assist once prosecution begins, which can be a considerable time after the initial filing. Capturing that context at the point of filing keeps the strategic memory from being lost, so that the reasoning behind a claim is still available when it is finally needed.
Five, ten, and fifteen years out
Set these threads next to Amara’s rule and a shape emerges. At five years, expect applications that are materially stronger as filed, prosecution that is shorter on average, and attorney time that has moved decisively toward strategy and the claims. At ten and fifteen years, expect the strategic, advisory, and inventor-facing parts of the job to dominate, with AI carrying far more of the production on both sides of the table. The part of the work that grows is judgment. That is the opposite of a profession in decline.
For anyone deciding whether this is a career with a future, that is the answer, and it is a hopeful one. This is the most engaging period I have known in patent law. The challenge of the work and the quality of what we can now produce are deeply satisfying, and the intellectual sparring with an AI over claim direction is a genuinely new dimension to the job. It is something like having an expert at your shoulder for every matter, one that can bring thinking and analysis that previously would have been impossible to obtain and bring to the task.
The human still holds the advantage that counts most. We live in the real world, we carry the client and inventor relationships, and we pick up the nuances that a model cannot. Pair that judgment with AI and you amplify it rather than replace it. There is also far more room now for the collaboration with inventors that makes the work rewarding in the first place. The work product we are putting out today would, in many cases, have been impossible to produce without these tools, and anyone entering the profession should be genuinely excited about what they will be able to do, and about how they will work with inventors and clients to deliver real value to clients and satisfaction to practitioners.
As the profession moves from billing for time toward billing for value, I do not expect the price reductions the prediction promises. The assets we are building are becoming more valuable, and the work it takes to build them well is becoming more interesting. The future of the job, in other words, may turn out to be the best part of the job.