Written by Maria Alejandra Alvarado and Anita Shaw
AI investment continues to rise, but the conversation around value is becoming more disciplined.
Companies are not retreating from AI. They are asking harder questions about where to invest, what outcomes to expect, and how quickly those investments should begin producing measurable business value.
In the United States alone, the largest hyperscale technology companies are expected to spend approximately $800 billion in capital expenditures in 2026, according to consensus analyst estimates cited by Goldman Sachs Research.
While A 2025 Deloitte survey of executives across Europe and the Middle East found that 85% of organizations had increased AI investment in the previous year and 91% planned to increase it again. Yet only 6% reported payback in less than a year, with most respondents saying a typical AI use case generated satisfactory returns over a two-to-four-year period.
Leadership teams are navigating increasingly complex questions about where their companies are creating enterprise value, what qualifies as innovation, what is simply an improvement, and which advances are most likely to support long-term growth and revenue.
These questions become even more difficult in AI.
Models change. Architecture evolve. Vendors come and go. Open-source and Standards technologies accelerate development. Products increasingly combine proprietary technology with external datasets, APIs, research partnerships, foundation models, and third-party software.
The traditional approach to intellectual property: identify an invention, protect it, and move on, is becoming challenging to apply in an environment where innovation is continuous, collaborative, and increasingly interconnected.
To explore how companies should approach these challenges, I sat down with Anita Shaw, Of Counsel and UK & European Patent Attorney in Caldwell’s London office. Anita brings more than two decades of experience working at the intersection of technology, intellectual property, and business strategy, advising engineering, legal, and commercial teams on how to identify, protect, and commercialize innovation across fast-moving areas such as AI, cybersecurity, and data.
From Anita’s perspective, three issues are becoming increasingly interconnected: how innovation is created, how it is protected, and how it is governed.
Together, these three areas provide a practical framework for building an IP strategy that can support innovation, competitive advantage, and long-term enterprise value in the AI era.
1. Establish Who Created the Innovation
AI complicates one of the most fundamental questions in intellectual property:
Where did the protected innovation come from?
That question matters because current intellectual property regimes continue to place considerable importance on human authorship and inventorship.
The precise legal treatment differs by jurisdiction and by form of intellectual property. Patent law, copyright law, trade secret protection, database rights, contracts, and licensing arrangements each raise different questions.
But one principle is becoming increasingly important for companies using generative AI and other AI tools in the development process:
Human contribution needs to be identifiable.
Anita describes this as the first practical law of AI creation.
If AI is being used during R&D, software development, product design, engineering, or content generation, businesses should be able to explain the role humans played in producing the relevant innovation.
That includes questions such as:
- Who identified the technical problem?
- Who conceived the relevant solution?
- What decisions were made by engineers or researchers?
- What role did an AI system play?
- Which inputs were used?
- Where did those inputs originate?
- Which individuals contributed to the final inventive concept?
Provence is becoming an IP issue
For years, many companies have treated records relating to datasets, software components, development decisions, and AI tools primarily as operational information.
That is changing.
In AI-enabled development, provenance can become central to determining ownership, inventorship, licensing obligations, confidentiality, and the scope of rights a company may ultimately be able to assert.
A useful audit trail can help a company establish:
What entered the development process, who contributed what, what was created internally, what originated externally, and which rights apply to the resulting technology.
That documentation may also become important downstream in licensing negotiations, investment diligence, strategic partnerships, disputes, and M&A.
It is therefore a mistake to think of documentation simply as administrative overhead.
In an AI-driven company, provenance increasingly forms part of the infrastructure supporting the value of the IP itself.
2. Identify the Invention Behind the AI
The second challenge is determining what is actually worth protecting.
Taking patents as an example, there is still a tendency for companies to describe innovation at the category level:
- “We use AI.”
- “We built a machine-learning platform.”
- “We developed a generative AI solution.”
Those descriptions may explain what technology a company uses. They do not necessarily explain what the company invented.
As AI becomes increasingly accessible, the more important question is:
What technical or commercial problems did the organization solve that others could not solve in the same way before?
This distinction matters legally.
Software and AI inventions are sometimes discussed as though they are categorically outside patent protection. That is oversimplification.
Patentability depends on the particular invention, how it is claimed, and the relevant jurisdiction. In areas involving AI, software, cybersecurity, automotive technology, hardware, and other technical systems, the legal analysis often turns on the specific technical contribution made by the invention.
Move from invention disclosure to invention harvesting
Anita advocates periodic IP harvesting: bringing technical, product, commercial, and IP teams together to identify developments that may have strategic value.
Engineering teams are typically focused on solving problems and shipping products.
They may not recognize that a workaround, architecture, security mechanism, data-processing technique, user-experience improvement, hardware integration, or optimization represents valuable intellectual property.
Regular harvesting allows the organization to ask:
- What did we solve?
- How did we solve it?
- Is the solution strategically important?
- Could competitors reproduce it?
- And what is the most appropriate way to protect it?
Patent protection may be appropriate in some cases.
In others, trade secret protection may offer greater strategic value.
Copyright, contractual restrictions, licensing terms, data rights, defensive publication, or combinations of protections may be more suitable elsewhere.
The objective is not portfolio volume.
The objective is to deliberate protection of innovation that matters to the business.
3. Treat AI Compliance as Part of the Product Architecture
The third component of Anita’s framework concerns AI compliance.
This is where IP strategy increasingly intersects with data governance, cybersecurity, product regulation, and risk management.
The EU AI Act is one prominent example of a broader global movement toward more formal AI governance.
For companies developing or deploying AI systems across jurisdictions, the practical lesson is straightforward:
Compliance cannot be treated exclusively as a pre-launch legal review.
Many regulatory questions originate much earlier in the product lifecycle.
These decisions are often made while engineers are designing the product.
That means companies increasingly need legal and regulatory thinking to enter the development cycle earlier – ideally, compliance should be “baked in” to product / service roadmaps.
When governance requirements are understood early, technical teams have more options for incorporating them efficiently.
When they are discovered immediately before launch, companies may face expensive redesigns, delayed deployments, contractual complications, or unnecessary commercial risk.
Governance can strengthen the asset
There is another reason legal and technical teams should think about compliance together.
Many of the disciplines required for responsible AI governance can also strengthen the company’s ability to understand and defend its intellectual property.
- Documenting development decisions.
- Tracking data provenance.
- Recording human contribution.
- Maintaining technical documentation.
- Understanding third-party dependencies.
- Implementing cybersecurity controls.
- Maintain audit records.
These activities serve different legal purposes, but they can reinforce one another.
For an investor, commercial customer, strategic partner, or acquirer, a company that can clearly explain how an important AI system was created, what it owns, where external technology enters the system, and how key risks are governed may present a more credible technology asset.
This is why AI governance should not automatically be viewed as separate from innovation strategy.
For General Counsel, IP leaders, CTOs, R&D teams, and technical founders, that may be the most important shift AI demands:
Build an IP strategy around the ideas, capabilities, and advantages that can endure as technology changes.