Patent prosecution feels a lot like negotiating in real life.
Some days, you hold your ground because the idea is strong and the examiner is simply missing it. Other days, you choose a smaller step on purpose, because getting something allowed now is worth more than fighting for everything later.
Founders get stuck here all the time. You built real tech. You want real protection. But the patent office pushes back. And now you have to decide: do you push, or do you narrow?
This article is about making that choice with clear eyes.
Not with ego. Not with fear. Not with “let’s just accept it.” And not with “we must win every argument.”
Instead, you will learn how to pick the move that gives you the best business result: strong coverage, faster allowance when needed, and clean room to build more patents around the core.
If you are building in AI, robotics, or deep tech, this matters even more. These fields move fast. Words matter. And one wrong amendment can quietly kill the value of the whole case.
Tran.vc helps founders make these calls early, before the record gets messy. If you want your first patents to be assets that raise your leverage—not paperwork that sits in a drawer—you can apply any time here: https://www.tran.vc/apply-now-form/
Prosecution Strategy: When to Push vs When to Narrow
The core question you are really answering
Every office action forces one decision: are you trying to protect the full idea as you first described it, or are you trying to secure a smaller win that still blocks real copycats. That sounds simple, but it is not. Because “full idea” is often a mix of what you built, what you plan to build, and what you wish the market will value later.
A strong prosecution strategy starts by naming the real goal of the case. Are you trying to protect your most valuable technical edge, or are you trying to get a granted patent fast so you can show traction to partners, customers, or investors. The right move depends on which of those is true today.
Tran.vc works with technical founders on this exact choice, early, before you burn time and money in the wrong direction. If you want help turning your tech into clean, defensible IP, you can apply any time: https://www.tran.vc/apply-now-form/
Why “push” and “narrow” are both valid moves
Many founders assume narrowing is failure, like you “lost” the argument. That is not how patents work in real business. Narrowing can be a smart trade when the narrowed claim still maps to what competitors must do to match your results.
Pushing is also not about pride. It is about keeping scope that you will never get back if you give it away too early. Once you add limits to a claim, you may lock yourself into those limits forever, especially if the file history becomes a set of promises.
So this is not a moral choice. It is a timing and leverage choice. The best teams learn how to push in the right moments, and narrow in the right moments, without harming future filings.
What “push” actually means in prosecution
Pushing means you are defending your original scope, or close to it. You argue that the examiner’s cited prior art does not teach your key parts. You explain why the combination the examiner suggests is not reasonable. You may provide technical detail to show the examiner is reading your claim too broadly or reading the reference too generously.
Pushing often includes adjusting the claim language without truly shrinking the scope. That can mean clarifying a term, fixing a vague phrase, or pulling an element from a dependent claim into an independent claim in a way that still covers the core idea.
The big point is this: pushing is not “no changes.” It is “changes that do not give away the business value.”
What “narrow” actually means in prosecution
Narrowing means you accept that the broadest version is not going to pass, at least not in this case and at this moment. You choose a more specific claim that avoids the prior art. You might add a technical limit that is clearly present in your product and clearly absent from the cited references.
Done well, narrowing is not random. It is planned. You narrow toward the thing your competitors cannot avoid. You narrow toward your advantage, not toward a detail that is easy to swap out.
A narrow claim that lands directly on your secret sauce can be far more valuable than a broad claim that never gets allowed.
Why this matters more for AI and robotics founders
In AI, “novelty” arguments can get muddy fast. Many references describe similar model flows at a high level. If your claim is too abstract, it can be easy for an examiner to say, “this is all known.” That is when smart pushing focuses on the real technical improvement, not the marketing story.
In robotics, the prior art is often deep and spread across old papers, patents, and industrial systems. Examiners may cite combinations that look plausible on paper but do not work in practice. That is where pushing can win, if you can explain why the combination fails or why your constraints are different.
In both areas, narrowing too early can trap you in a corner. But pushing too long can waste cycles. The skill is knowing which path builds a real moat.
The hidden risk founders miss: the file history becomes a weapon
What you say during prosecution can later be used against you. If you argue “my invention requires X,” you may have just limited your patent to only systems with X, even if the claim does not clearly say that.
This is why “push” must be careful. When you push, you should not over-promise. You should not describe the invention as smaller than it is just to win an argument.
Narrowing has the same risk. If you narrow with an amendment that makes your claim look like a single narrow embodiment, you may have made it easier for competitors to design around you.
A clean prosecution strategy keeps the record tight, technical, and consistent.
The first step before you decide: map claims to business outcomes
Before you answer an office action, you should be able to say which claim matters most and why. Not in legal terms. In business terms. Which claim, if granted, makes a competitor’s life hard. Which claim helps you close a deal. Which claim supports your next fundraise.
If you cannot answer that, you will be tempted to argue everything, or accept anything. Both are costly.
A simple way to think about it is: you want at least one claim that reads on your product today, and at least one claim that reads on the market direction you expect. Whether you push or narrow should protect those two goals.
If you want a partner who can do this mapping with you, Tran.vc helps founders turn their tech into an IP plan that matches real company milestones. Apply here: https://www.tran.vc/apply-now-form/
The decision framework: when to push and when to narrow
Push when the examiner is stretching the prior art
Sometimes the cited reference does not really teach what the examiner says it teaches. The examiner may point to a generic statement and treat it like a full disclosure. Or they may treat an optional feature as if it is required.
This is a strong time to push. Your job is to calmly show the gap. You do that by quoting the exact parts of the reference that matter, then explaining the missing link in plain technical language.
The key is to make it easy for the examiner to change their mind without losing face. A clean, respectful argument often works better than a long lecture.
Push when you can win with claim interpretation
Examiners often apply broadest reasonable interpretation. That does not mean any interpretation. If a term has a clear meaning in your field, you can use that. If your spec defines a term, you can anchor to that.
A smart push is sometimes just tightening the wording so that your claim points to the correct technical meaning. You are not shrinking the invention. You are removing ambiguity that the examiner used to connect your claim to the reference.
This is one of the safest forms of pushing because it reduces confusion and still keeps scope.
Push when you have dependent claims with strong value
If your independent claim is broad and gets hit, you may still have dependent claims that capture the key improvement. Pushing does not always mean fighting only for the independent claim as-is.
It can mean negotiating scope by moving one powerful dependent feature up into the independent claim. That is a narrowing move in form, but a pushing move in strategy, because you are still defending the part that matters most.
This is why drafting matters. Good dependent claims give you options. Poor dependent claims force you into bad amendments.
Narrow when the prior art is clearly on point
Sometimes the reference really does disclose your broad concept. Or a combination truly covers it. In that situation, pushing can turn into expensive delay.
A clean narrow move is to accept the reality, then choose a claim path that still blocks competitive copying. This is where founders need discipline. You do not narrow toward a feature that feels “technical” but is easy to remove. You narrow toward the step that creates the advantage.
If you are unsure whether the art is truly on point, that is a signal you need a stronger technical read and possibly a quick prior art search to confirm.
Narrow when time matters more than breadth
There are moments where speed is worth more than broad scope. For example, you may be entering a partnership where “patent pending” is not enough, and you need an issued claim to reduce procurement friction. Or you may be in a space where investors ask for granted claims as proof you can defend your edge.
In those cases, a planned narrowing can be a smart business move. You take a narrower allowance now, and you keep broader ideas alive through continuations or new filings that cover adjacent scope.
The mistake is narrowing without a second plan. If you narrow and end the case, you may have traded away the future for a short-term win.
Narrow when your spec does not support the broader version
Founders sometimes learn, during prosecution, that the broad claim they want is not fully supported by what they wrote in the original filing. If the spec does not describe key variants, pushing can fail because you cannot reasonably stretch the disclosure.
A careful narrowing then protects you from a worse outcome: making arguments that contradict your own document, or trying to add new matter, which you cannot.
This is why Tran.vc focuses on strong early drafting. A strong first filing gives you room to push later without breaking support. Apply any time if you want that kind of foundation: https://www.tran.vc/apply-now-form/
The “do not regret this later” test
A good decision framework includes a regret test. If you narrow today, will you be upset in two years when a competitor ships a variant that you could have covered if you had pushed. If yes, you should slow down and consider a different move.
On the other hand, if you push today, will you burn months and legal budget and still end up with the same narrow claim, just later. If yes, you should consider taking the narrow allowance sooner and saving your energy for a continuation.
This regret test forces you to think like a company builder, not like someone trying to win a debate.
A note on tone and relationship with the examiner
How you communicate matters. Examiners are human. If your responses are clear, technical, and focused, you are more likely to get a fair read. If your responses are hostile or messy, you may face more resistance.
Pushing does not mean being aggressive. Narrowing does not mean being passive. Both should be done with calm confidence.
The most effective prosecution teams treat the examiner like a technical peer, not like an enemy.
The tactical moves that make pushing work
Build one strong argument instead of five weak ones
A common mistake is trying to attack every sentence in the office action. That usually leads to long responses that are hard to follow and easy to dismiss.
A better approach is to find the one key gap in the examiner’s logic and build the response around it. If the examiner’s rejection requires one missing element, focus there. If the combination requires an assumption that breaks the system, focus there.
When you give the examiner a simple reason to allow the claim, you reduce the chance of a new rejection popping up in the next round.
Use your spec like a technical anchor
Your spec is not just background. It is your best tool for claim meaning. When pushing, you want to show that your claim terms have a specific technical role in your system.
That means pointing to how your spec uses the term, how it interacts with other parts, and what problem it solves. This helps push back against overly broad readings that pull your claim into prior art that only looks similar at a surface level.
If your spec is thin, this is harder. If your spec is rich, pushing becomes far more effective.
Add clarity without boxing yourself in
Some amendments are “clarifying” but still dangerous if they create accidental limits. The safest clarifying changes are those that align with your original description and do not introduce a single narrow implementation.
For example, if your invention is about an improved control loop, adding language that ties the loop to a specific sensor model may shrink your future options. But adding language that ties the loop to a measurable system state can clarify meaning without forcing one hardware choice.
This is where experienced patent strategy matters. Good wording can keep the door open while still distinguishing over the art.
Know when an interview can save a full cycle
Sometimes you can resolve a misunderstanding faster by talking. Examiner interviews can clarify claim meaning, narrow the real dispute, and reveal what the examiner would accept.
Interviews are not magic, and they do not replace strong writing. But they can prevent you from writing a long response that misses the real concern.
A good interview ends with a clear plan: either you will amend in a certain way, or you will argue a certain gap, and the examiner confirms that would address the rejection.
Avoid statements that shrink the invention
When founders get nervous, they sometimes make absolute statements like “the invention requires X” or “unlike all other systems, ours always does Y.” These lines can haunt you later.
Even if you are pushing, you can often argue without making the invention smaller. You can say, “in the cited reference, X is not disclosed,” instead of “our invention requires X.”
This keeps your patent flexible while still addressing the rejection. It is a small writing choice with a big legal impact.
Narrowing without losing the moat
Narrowing is a design decision, not a surrender
When narrowing is done well, it feels intentional. You are not shrinking because you are tired or unsure. You are shrinking because you see a cleaner line between what is known and what you uniquely built.
The goal is not to make the claim easy to allow. The goal is to make the claim hard to copy. That difference matters. Many patents get allowed and still fail to protect anything useful. Smart narrowing avoids that trap.
At Tran.vc, this is where we spend real time with founders. We ask what part of the system cannot be removed without breaking performance, cost, or reliability. That answer guides every narrowing move.
Narrow toward the constraint competitors cannot escape
Every real system has constraints. Some are chosen. Some are forced by physics, cost, or data limits. The best narrowed claims sit right on those forced constraints.
For example, if your system works only because of how data is synchronized across components, narrowing toward that synchronization logic may be powerful. A competitor can change UI, hardware, or even models, but they cannot escape the need to coordinate state in the same way.
Narrowing toward surface features, like a specific format or sequence that is easy to alter, usually leads to weak protection. Narrowing toward structural or operational constraints usually leads to strong protection.
Use narrowing to lock in performance advantages
One of the safest ways to narrow is to tie the claim to a performance outcome that flows directly from your design. This is not marketing language. It is technical cause and effect.
If your architecture reduces latency because of a specific processing order, narrowing to that order can be strong. If your robot improves stability because of how sensor fusion is timed, narrowing to that timing can matter.
Performance-linked narrowing makes it harder for competitors to argue they are doing something “different” when they are chasing the same result.