Patent Attorneys, AI, and the Skills Gap: Insights from AIPLA Spring Meeting

1. What’s Causing the AI Skills Gap Among Patent Attorneys?

Patent attorneys, especially those in private practice, often face a skills gap when it comes to effectively adopting and deploying AI tools. While many are experimenting with ChatGPT, Claude, and similar generic GenAI platforms, few have hands-on experience with domain-specific AI solutions purpose-built for patent prosecution.

It is common for patent legal professionals to overlook the need for developing key skills such as prompt engineering and tool assessment for effective implementation into workflows. Additionally, most attorneys are pressured to juggle additional training on AI with day-to-day client work. As a result, many firms are underutilising the capabilities of AI, or worse, relying on general-purpose platforms for tasks that require legal and technical precision, such as aspects of drafting and prosecution – tasks that are not well-suited to be carried out by generic AI tools.

There’s a skillset around understanding the language of these models. For example, prompting isn’t just about asking a question - it’s about understanding how to communicate with the model. For general-purpose AI tools, the difference between a mediocre and an insightful output often comes down to how well the tool is prompted.

‍

2. Why Aren’t General AI Tools Enough for Patent Work?

A show-of-hands during the session confirmed what many suspected: most attorneys already use AI in some form. But based on the back-and-forth discussion between panel and audience, it’s suspected that these are overwhelmingly general-use tools, not fine-tuned and specialised for patent workflows.

There are immediate benefits from GenAI by applying it to low-risk, high-frequency tasks, including:

●  Generating template documents and structuring plans for the specification

●  Translating/framing internal communications

●  Drafting non-substantive sections (e.g. background)

●  Formatting and proofreading

While tools like ChatGPT can help with tasks like summarisation and drafting outlines, they are ineffective when it comes to complex patent tasks such as claim drafting (nor is it secure from a data confidentiality standpoint), argumentation and office action responses. These are examples of functions that require linguistic consistency, legal accuracy, and stylistic fidelity (e.g., firm-specific preferences, jurisdictional requirements, phrasing, tone, etc.).

One panelist highlighted that, for current generic tools of GenAI platforms, when you increase the number of relationships between documents, as is the case for complex office actions, e.g., a specification, previously amended claims, multiple prior art references, and office action – GenAI’s ability to connect the dots can decline noticeably, if not fine-tuned effectively.

‍

3. Can AI Replace Junior Patent Attorneys?

It was suggested that many law firm partners and in-house stakeholders are operating under inflated expectations. There’s a pervasive belief that AI will replace junior attorneys and other legal staff. But the situation is more nuanced.

The most effective use of AI in patent law is not autonomous text generation; it’s assisted acceleration past the blank-page stage, allowing attorneys to pivot into focusing on aspects that require expert critical thinking, which is ultimately a more valuable use of the attorney’s time. Tools can streamline parts of the drafting and prosecution process, but legal quality and strategy must remain in expert (human) hands.

As Ian Clouse of Holland & Hart aptly noted: AI should be the engine that powers an attorney’s workflow–not the navigator. Attorneys still need to steer. Often, practitioners think they’ll input a prompt and get an immediately usable end-product, but for patent workflows, an interactive process of working with the AI in a piecemeal fashion ensures attorney approval and, most importantly, a high-quality patent application or response. 

‍

4. What Should You Look for in a Reliable Legal AI Tool?

Like many areas of law, patent drafting and prosecution are unforgiving regarding errors - even small inconsistencies in terminology can jeopardise validity and enforceability. That’s why generic tools fall short for complex tasks like office action responses or claim amendments.

A consistent complaint among panellists was the unreliability and opacity of many general-purpose AI tools. Unfortunately, general-purpose models often introduce inconsistent terminology or overlook nuanced distinctions, and ‘close enough’ isn’t good enough in patents. Models hallucinate, forget previous context, and behave inconsistently when dealing with complex questions or tasks, making them ill-suited for end-to-end patent workflows.

To be viable in legal practice, AI tools must offer:

●  Transparent, context-specific outputs that are editable by the attorney in the driver’s seat

●  Robust data security and confidentiality safeguards so sensitive data is protected

●  Customisable formatting and instructions to the AI to adapt to individual styles and needs

●  Consistent or enhanced performance that aligns with evolving AI models

The best legal AI tools don’t try to replace attorneys. They can–and should–be considered reliable collaborators where the attorney is always in control. One panellist analogised that treating general LLMs like Wikipedia from its early years would set the right expectation when integrating AI into workflows.

‍

5. Why Don’t Most Firms Build Their Own AI for Patents?

Developing bespoke patent AI tools remains out of reach for most small and mid-sized firms. Even for larger firms, although attorneys wish that AI could learn from decade’s worth of their office actions or applications filed, this database still would fall short of the volumes typically required for meaningful model training and fine-tuning.

Instead, many firms are turning to patent-specific platforms built on multiple models that are enhanced for patents, secure, and adaptable. Approaches such as retrieval-augmented generation (RAG) and patent domain-specific fine-tuning offer a practical path forward, balancing functionality, cost, and customisation. This is what we’re doing at Solve Intelligence.

‍

6. How Can Attorneys Stay Relevant in an AI-Assisted Practice?

As one panellist put it, the real threat isn’t AI itself - it’s being outpaced by peers who know how to use it. Like learning to use a computer decades ago, developing AI literacy is fast becoming a baseline skill in the legal profession. The panellists emphasised how attorneys using AI will prove more competitive, efficient, and therefore valuable within the profession.

Firms that take the approach of choosing the right tools, investing in AI literacy amongst its staff, and maintaining attorney-led supervision of AI-assisted outputs will lead the next phase in patent preparation and prosecution.

Platforms like Solve Intelligence demonstrate how AI tools are revolutionising patent workflows, keeping the attorney as the ultimate decision-maker, all the while ensuring compliance with legal standards.

‍

7. What’s the Best Way to Get Started with AI in Your Firm?

For law firms exploring AI adoption in earnest, the panellists offered actionable guidance rooted in real-world use cases and common pitfalls.

Start Now, and Start Small If Needed

AI literacy is no longer optional - it’s foundational. Just as attorneys decades ago had to learn to use a computer, today’s professionals must build comfort with interacting with AI tools. A good way to begin is by using AI to overcome the “blank page” problem. For example, here’s a tip for first-timers: try using it to generate background sections, structure outlines, or draft low-risk internal communications.

Prioritise Confidentiality

Never enter client data into a public GenAI platform without understanding its data policies. Tools built for legal workflows should have strict security protocols, enterprise-grade encryption, and clear data use and retention policies so sensitive, private information (like inventions) is fully protected.

The Learning Curve

A hurdle in AI adoption is the perceived complexity of learning new tools. However, with Solve Intelligence, attorneys can start integrating AI into their workflows immediately, without changing how they already work. Solve’s Patent Copilot is designed to complement existing drafting and prosecution habits, not replace them. It operates within familiar user interfaces, uses language that attorneys understand, and delivers output that can be reviewed (including track changes for AI-generated output) and edited like any traditional work product. Whether you’re reviewing invention disclosure material, generating first drafts, automating routine habits, or automating and validating citations in office action responses, the tool meets practitioners where they are, making AI adoption feel less like a learning curve and more like a natural extension of your toolkit.

Choose the Right Partner

Rather than attempting to build your own model from scratch or a general-purpose GenAI tool that isn’t fine-tuned to your needs, look for platforms that are purpose-built for patents and keep you, the legal professional, in the driving seat. Customisability, context-awareness, and transparency in how the AI makes decisions are key features to look for. Firms that begin adopting AI now - safely and strategically - will position themselves ahead of the curve. Solve Intelligence is designed, tested, and purpose-built for patent legal professionals and could be an invaluable tool for any patent attorney, whether you are a seasoned AI user or a first-timer.