If you file patents in more than one country, you will get office actions. Not “maybe.” You will. The good news is: most office actions are normal, predictable, and beatable if you plan ahead. The bad news is: they can drain time, budget, and founder focus if you treat them like random bad luck.
At Tran.vc, we help technical founders turn inventions into real assets early—before the seed round pressure hits. If you want help building an IP plan that is strong across borders, you can apply anytime here: https://www.tran.vc/apply-now-form/
The simplest way to think about an office action
An office action is a letter from a patent office that says, “We reviewed your patent application. Here is what we do not like yet.”
It can be about:
- what your invention is (clarity)
- whether it is new (novelty)
- whether it is too close to what already exists (obviousness / inventive step)
- whether your claims are too broad
- whether your drawings or wording have issues
- whether you used the right format, fees, or steps
No matter the country, the pattern is the same:
- You file.
- An examiner reviews it.
- The examiner raises issues.
- You reply with arguments and claim changes.
- You either move toward allowance or you fight more.
What changes across countries is how strict the examiner is, what they care about most, how they write the office action, and how much room you have to fix things without losing value.
Here is the core founder lesson: office actions are part of patent creation, not a mistake. The goal is not “avoid office actions.” The goal is “expect them and control them.”
And yes—this matters even more for AI, robotics, and deep tech, because those fields trigger extra scrutiny in many places.
Why cross-country office actions feel harder than they should
Most founders assume patents are like code: one build, then ship everywhere. Patents are more like language. You can express the same idea in many ways, but each place has its own grammar and tone.
That is why a claim that looks fine in one country can get hit fast in another. You might feel like you are answering the same questions again and again. But you are not. You are adapting to different examiner habits and local rules.
The fastest way to reduce pain is to plan your “global claim story” early. That means:
- Decide what the invention really is at the core (the one sentence truth).
- Decide which parts must be protected no matter what.
- Decide which parts are optional add-ons you can trade if you need to narrow claims.
- Build a spec (your written description) that supports multiple claim shapes.
Founders who do this upfront usually spend less later, and they end up with claims that are easier to enforce.
If you want Tran.vc to help you design that kind of plan, apply here: https://www.tran.vc/apply-now-form/
What you should expect in almost every country
Let’s talk about the “repeat offenders”—the same themes that show up in office actions almost everywhere.
Clarity attacks: “Your claim is not clear.”
This one is common and often underestimated. In deep tech, founders love compressed language: “agent,” “model,” “optimizer,” “policy,” “node,” “controller,” “digital twin,” “planner,” “inference,” “feature,” “latent.” These words mean a lot to you, but not always to an examiner reading fast.
If your claims rely on terms that can be read in two ways, expect a clarity rejection.
What works in practice is boring but powerful: define your key terms in simple words inside the spec. Not pages of theory. Just clean definitions, grounded in what the system does.
Example idea (not legal advice, just a pattern):
If you say “planner,” add one sentence in the description that says what it outputs, what it takes as input, and where it sits in the system.
Clarity fights are usually winnable, but they can still burn cycles. So set yourself up to win early.
Prior art attacks: “This already exists.”
In the office action, the examiner will cite earlier patents or papers (or products) and say your claim is not new or is too close.
Founders often respond emotionally to this: “But ours is different.”
The examiner does not care about “different.” They care about what is claimed.
Here is the tactical move that helps most: stop arguing at the “system” level and argue at the “mechanism” level.
Instead of “we do robotics control better,” focus on:
- what exact step is performed
- when it is performed
- how it changes a signal, decision, plan, or output
- what constraint or condition triggers it
- what result it guarantees or improves
Then shape claims so that the key mechanism is not just described in the spec, but also actually claimed.
A lot of AI and robotics patents fail early because the claim is written like a product page. Examiners treat product-page claims as broad and “obvious.”
Eligibility / excluded subject matter: “This is not patentable here.”
This is where cross-country differences get real.
Some places are more friendly to software-type claims. Some are stricter. Some require “technical effect.” Some focus on whether it is “just math.” Some focus on whether you are claiming a business idea with a computer.
If you work in AI, you should assume you will face some version of this issue somewhere in your global plan.
The way out is usually not to argue philosophy. It is to ground the claim in a technical system:
- tie the method to a machine or device context
- claim data that comes from sensors or physical processes
- claim constraints linked to real-world limits (battery, latency, torque, safety)
- claim improved control, stability, reduced error, reduced compute, reduced memory, or better real-time performance
- show that the model is not “an abstract model,” but a tool that changes how a machine behaves
The important point is not to “add hardware words” as a trick. The point is to show the examiner you are solving a technical problem in a technical way.
A quick map: the main “styles” of office actions you will meet
Instead of listing every country right away, it helps to understand examiner styles. Most patent offices fall into a few broad styles:
Style A: prior-art heavy, logic-heavy.
The examiner will cite documents and map them to your claim line-by-line. Replies need careful structure and tight claim edits.
Style B: clarity and form heavy.
The examiner cares a lot about claim wording, support in the description, and clean structure. Replies need careful drafting and compliance.
Style C: eligibility and “technical” heavy.
The examiner may push back if your invention feels like software “in the abstract.” Replies need technical framing and strong spec support.
In real life, a country can be a mix. But thinking in these styles helps you prepare your filings in a way that travels well.
The founder’s playbook before the first office action shows up
This section matters because the best office-action response often happens before the office action exists.
If you do just one thing, do this: write your first filing as if you will need to narrow later.
That means your spec should include multiple fallbacks. Not a long list. Just smart depth.
For a robotics or AI startup, the fallbacks usually come from:
- different sensor sets (camera-only vs camera + IMU vs camera + LiDAR)
- different training or update modes (online vs offline, centralized vs on-device)
- different timing models (real-time constraints, latency thresholds)
- different safety or constraint logic (fail-safe steps, boundaries, guardrails)
- different representations (graph vs voxel vs mesh vs keypoints)
- different control schemes (MPC vs learned policy vs hybrid)
- different compute placements (cloud vs edge vs onboard)
Why this matters: in many countries, you can only claim what you properly described at filing. If you did not write it down early, you may not be able to add it later. That is where founders lose global claim strength.
So the ideal mindset is: your first filing is not a single shot. It is a claim support warehouse.
This is exactly the kind of work Tran.vc funds as in-kind IP services—so you can build the right base before you start spending real cash on global filing and prosecution. Apply anytime: https://www.tran.vc/apply-now-form/
What “good” looks like when you respond to an office action
A strong response is not a long response. It is a response that makes the examiner’s job easy.
The best replies usually do three things:
- They clearly state what the examiner is missing.
- They show the missing piece in the claim language, not only in explanation.
- They narrow only where it protects value, not where it kills value.
That last point is the hard one. Many teams narrow too fast. They accept language that gives them a patent that is technically allowed but commercially weak.
A practical way to avoid that trap is to ask a simple question when you consider edits:
If a competitor copied our core product, could they avoid this claim easily?
If the answer is yes, that claim change may be too costly.
This is why experienced counsel matters. Not because of fancy writing. Because office actions are negotiation. You are trading scope for allowance. The goal is to trade in places that do not touch your moat.
Country differences start with one big concept: “inventive step” is judged differently
In the US, a lot of the fight is around obviousness and how references combine. In other places, the “inventive step” analysis can feel more structured and sometimes stricter about what counts as a true technical leap.
This affects AI and robotics heavily because examiners often see “model + data + output” and think “routine.”
So what makes an AI or robotics invention look inventive on paper?
Usually one of these:
- a specific pipeline that produces a measurable technical gain
- a control loop improvement that changes real machine behavior
- a training or adaptation method that solves a real deployment pain (drift, latency, memory, safety)
- a new way to fuse sensors or represent the world that improves robustness
- a method that reduces compute or improves real-time response under constraints
In many countries, the examiner is not impressed by “we used a neural network.” You need to show why your approach is not the default move.
This is not about exaggerating. It is about being precise.
The hidden cost: office actions are not only legal, they are operational
Founders often think office actions are “attorney work.” But they pull on the team:
- You may need an engineer to explain the difference from prior art.
- You may need experiments or logs to support a technical effect.
- You may need to decide whether to narrow claims in a way that matches the roadmap.
- You may need to align with fundraising timing.
That is why early planning is so valuable. Office actions come months or years after filing. Your team may have changed, your product may have shifted, and you may have less context.
If your original filing is clear and deep, your future self wins.
A note about timing: office actions will not arrive at the same pace
Even for the same invention family, countries move at different speeds. Some are slow. Some are fast. Some are fast if you request it. Some become fast if you use work-sharing programs.
This matters because you may want at least one allowed case early for investor proof. Or you may want to delay some costs until the company is ready.
A smart global plan is not “file everywhere.” It is “file in the places that match your market, your risk, your budget, and your timeline.”
Tran.vc works with founders on this exact question: how to build a strong IP path without burning runway. If you want that help, apply here: https://www.tran.vc/apply-now-form/
Office Actions in the United States (USPTO)
What a US office action usually feels like
A US office action is often direct and practical. The examiner will cite past patents or publications and explain why your claims are not new or are too close to what already exists. You will usually see a focus on obviousness, meaning the examiner thinks someone could combine two or more old ideas and get your invention.
Many founders get frustrated here because the examiner may sound confident, even when the fit is weak. In the US, that tone is normal. Your job is to reply with calm structure and show why the examiner’s mix-and-match story does not hold when you look at the exact claim steps.
The most common US rejections you will see
In the US, you should expect novelty and obviousness rejections early in the process. You will also see clarity-type issues, but the big time sink is often obviousness. The examiner may say your claim is “an obvious variation” and cite two or three references.
If you work in AI or robotics, another common theme is “this is just software” or “this is an abstract idea,” sometimes tied to eligibility rules. That does not mean you cannot patent AI in the US. It means your claims must show a real technical process that does real work in a real system.
How to respond in the US without weakening your moat
A strong US reply starts with one simple move: pick the one or two claim parts that create real separation from the prior art. Then build the response around those parts like a spine. When you try to defend every word, you often lose the reader and you waste your best points.
Claim edits can help, but you should not rush to narrow. A smart approach is to first try clean arguments, then narrow in a way that matches how your product truly works. When you narrow, narrow into your advantage. If your edge is safety, tie the claim to the safety constraint. If your edge is latency, tie it to the timing boundary.
What US examiners look for in AI and robotics claims
US examiners tend to respond well when you show the invention as a working pipeline. Instead of sounding like “we have a model,” you want to sound like “the system takes these inputs, runs this step, produces this control output, and improves this machine result.”
If your invention is about training, show how training connects to deployment, especially if it reduces drift, reduces compute, or improves real-time performance. If your invention is about control, show the control loop and the constraint logic. The more your claim reads like a technical recipe, the less it reads like a concept.
If you want Tran.vc to help shape US-first claims that still travel well globally, you can apply anytime: https://www.tran.vc/apply-now-form/
Office Actions in Europe (EPO)
Why EPO office actions feel different
The EPO is famous for being structured. The examiner often uses a problem-solution style, where they pick a “closest prior art” reference, describe the difference, and then ask whether that difference truly creates an inventive step.
For founders, this can feel more strict than the US. In reality, it is just a different game. The EPO wants a clear technical problem and a clear technical solution. If your filing is written like a product pitch, EPO prosecution can get painful fast.
“Technical effect” is not a slogan, it is your anchor
In Europe, the phrase you will hear a lot is “technical effect.” For AI and software, that means the examiner wants to see a real technical improvement, not just better business outcomes or better reporting.
If your invention improves control stability, reduces sensor noise impact, reduces compute load, improves battery use, improves network usage, improves safety, or improves real-time response, those are the kinds of effects that tend to land well. You do not need to use fancy words. You just need to connect the dots between the method steps and the technical result.
The EPO’s clarity and support expectations
EPO examiners can be strict about claim clarity and whether the description supports what you are claiming. If your claim uses a term like “score,” “state,” “confidence,” or “policy,” the examiner may ask what that means in the context of the system.
This is why early drafting matters so much for Europe. If the spec explains the components and signals clearly, you have room to adjust claims later. If the spec is thin, you may get trapped, because you cannot add new detail after filing.
How to reply to EPO office actions in a way that works
A good EPO response usually starts by accepting the structure the examiner is using, then guiding it to your advantage. You identify the closest prior art, then you state the exact difference, then you explain why that difference is not obvious because it produces a technical effect under a real constraint.
When you amend claims for the EPO, you often want to move from broad concepts to a tighter, technically grounded version that still covers your core product. A common founder mistake is narrowing into a random detail that is easy for competitors to avoid. Instead, narrow into the part that is hard to copy without rebuilding the system.
A practical tip for robotics and embodied AI in Europe
If your AI touches the physical world, Europe often becomes easier. If your method uses sensor data, affects motion planning, changes control outputs, or enforces safety limits, you can frame the invention as solving a technical problem in a machine environment.
The key is to write it that way from the start. If you file with a generic “computer-implemented method” tone, you may end up rewriting the story later. That is doable, but it can cost scope.
Tran.vc helps founders build that Europe-ready technical story early, so you do not have to patch it later under pressure. Apply anytime: https://www.tran.vc/apply-now-form/
Office Actions in the United Kingdom (UKIPO)
Why the UK can surprise founders
The UKIPO often feels like a close cousin of Europe, but it has its own personality. UK examiners can be very focused on whether the invention is excluded subject matter, especially for software-like claims. This shows up as questions about whether your invention makes a real technical contribution.
Founders sometimes assume that if Europe is fine, the UK will be fine in the same way. Often that is true, but not always. The UK’s framing can be slightly different, and the examiner may push you to explain the contribution in a very plain way.
The “technical contribution” test in simple terms
In the UK, you should be ready to answer one plain question: what technical thing changes in the real world because of your invention? Not “the dashboard is better,” not “the business process is better,” but what changes in the system.
If your invention improves a robot’s movement, reduces collisions, improves localization, reduces compute time on-device, reduces memory load, or improves network use for a deployed system, you can often shape a strong response. The UK tends to respect clear engineering stories.
How to draft and reply in the UK without playing defense
The best UK approach is often proactive. You frame the invention as an engineering solution from the first paragraph of the description, then your claims follow that same path. When the examiner objects, you reply by pointing to the technical parts, not by debating abstract definitions.
Amendments in the UK should be done with care, because you want to keep coverage for your product as it scales. If you narrow too quickly into one sensor, one robot type, or one dataset form, you may limit your future roadmap. A better move is to narrow into the control or compute mechanism that stays true across versions.
A founder-friendly strategy for US + EPO + UK together
If you are filing across these regions, you should aim for a shared core story that can be expressed three ways. The US likes a clean pipeline and strong distinction over prior art. The EPO likes problem-solution logic and technical effect. The UK likes technical contribution stated plainly.
You do not need three different inventions. You need one invention explained with three accents. That is why early IP strategy is not paperwork. It is product strategy, written down.
If you want Tran.vc to help you build a global-ready claim base from day one, apply here: https://www.tran.vc/apply-now-form/