Patent Basics in the AI Age -A Practical Five-Bucket Framework for Building an Investor-Ready Patent Position

By Daniel J. Holmander and Michel “Mike” Morency, Ph.D. on August 24, 2026

Quick answer for investors and growth companies: A provisional patent application is not investor-ready simply because it exists.  It becomes useful when the company can show five things:  (i) what was disclosed and to whom (and under what protection); (ii) what the inventors actually invented and who conceived it; (iii) what prior art was searched and how the invention was assessed; (iv) what the filing was drafted to support; and (v) how the 12-month provisional year will be used to build a stronger position.  Those are the five buckets below.

A company can now use AI to organize an invention disclosure, search for prior art, summarize the science, draft figures, prepare investor materials and produce something that looks like a provisional patent application in a matter of hours.

That speed is useful.  It is also dangerous.

Used well, AI helps a company arrive at its first meeting with patent counsel better prepared. Used carelessly, it produces a polished document that gives founders, executives and investors a false sense of protection.

In the AI age, then, the first patent question is not whether a company can file something quickly.  It is whether the company can build a patent position that investors can trust.

The Five Buckets

#	Bucket	The question it answers
1	Controlled disclosure	Who knows what, and under what protection?
2	AI-assisted invention disclosure	What was invented, by whom, and what is missing?  Use AI to organize the facts for counsel..
3	Patent assessment	Is it new, non-obvious, eligible, and worth patenting?  Search broadly first.
4	Professional provisional drafting	Does the filing support the claims the company will need later, including after a pivot?
5	Patent roadmap	How will the provisional year and the years after it be managed?

The legal backdrop matters.  Provisional applications are governed by 35 U.S.C. 111(b) and MPEP 201.04.  Everyone substantively involved in a patent application owes the USPTO a duty of candor under 37 C.F.R. 1.56 and MPEP 2001.  And AI, software, diagnostic and other life sciences inventions regularly face subject-matter eligibility questions under MPEP 2106, the USPTO’s 2024 AI eligibility guidance and a series of 2025-2026 USPTO updates discussed in Bucket 3.

This is not theory.  It is the workflow that matters when a company wants to raise capital, work with collaborators, protect AI or life sciences technology, preserve foreign rights and survive diligence.

Bucket 1:  Controlled Disclosure

Patent work should begin before drafting.  It should begin with control.

The first question is simple: who already knows about the invention?  The second question is more important: under what duty of confidentiality?

A disclosure to patent counsel is different from a casual disclosure to a potential partner, developer, investor, customer, contractor, university, hospital, CRO, accelerator or public AI tool.  The first may occur within a professional framework of confidentiality and privilege.  The others may depend on a contract, a course of dealing, a later factual dispute or nothing at all.

Before the company sends a deck, shares a data room, posts code, gives a demo, circulates a white paper, submits a conference abstract, talks with a developer or describes the invention to an AI platform, the company should know whether the disclosure is protected.

Why it matters legally.  The United States gives an inventor a one-year grace period for the inventor’s own public disclosures (35 U.S.C. 102(b)(1)).  Most of the rest of the world does not.  In Europe, China, Japan and many other jurisdictions, a single public disclosure before the first filing (a conference poster, a preprint, a podium talk, an unprotected pitch) can destroy novelty outright.  For a life sciences company that publishes and presents, the practical rule is to file before you speak.

Public AI tools are disclosures too.  Typing an invention into a consumer AI platform may place it on a third party’s servers under terms that permit retention or training.  The USPTO’s April 2024 guidance on AI tools warns that this can compromise confidentiality and, where servers sit outside the United States, can implicate foreign filing license and export control rules.  Enterprise AI tools with no-retention, no-training terms are a different matter, but someone should have read those terms.

Ownership is part of control.  Every inventor, founder, employee and consultant should have signed a present assignment of inventions (“hereby assigns,” not “agrees to assign”) before anything is filed; the Supreme Court’s decision in Stanford v. Roche is the standard cautionary tale.  Work done at or with a university or hospital is usually subject to that institution’s IP policy.  Work funded by NIH, NSF or SBIR grants carries Bayh-Dole reporting and election obligations that investors check in diligence.

For a first patent conversation, we want to know:

• What has been disclosed, and what remains confidential?
• Who received the information, and when, when, and in what form (oral, written, demo, data room, code, abstract)?
• Was the disclosure oral, written, visual, digital, public or private?
• Was there a confidentiality agreement, joint-development agreement or other written obligation?
• Was AI used to process, summarize or describe the invention, and on what platform and terms?
• Who contributed to the invention, and has each contributor signed an assignment?
• Are universities, hospitals, government grants, CROs, customers or strategic partners involved?

This is not paperwork for its own sake. It is how a company avoids giving away valuable information before deciding what to patent, what to keep secret and what to share.

Practical rule Control the disclosure before relying on the filing, and file before you speak.

Bucket 2:  AI-Assisted Invention Disclosure

AI can be excellent at helping prepare an invention disclosure. That does not mean it should replace patent counsel.  It means AI can help inventors and business teams organize what they know before the legal work begins.

A great disclosure is not a marketing summary. It is a technical and business record that helps counsel understand the invention, the alternatives, the commercial embodiment and the evidence behind it.

A strong invention disclosure answers these questions:

• What problem does the invention solve, and what is the technical improvement over the prior approach?
• What did the inventors actually build, test, model, formulate or implement?
• What are the key components, steps, compounds, sequences, assays, models, devices or methods?
• What alternatives, variations and fallback embodiments are possible?
• What data, experiments, drawings, flowcharts, screenshots, protocols, architecture diagrams or results support the invention?
• What failed approaches were tried, and what did the team learn?
• Who conceived each feature, and what role did AI tools play?
• What is the current commercial embodiment, and what is likely to change in the next 6, 12 or 24 months?

For AI inventions, the disclosure may need to address model architecture, training approach, data inputs and outputs, system integration and the specific technical problem solved.  For life sciences inventions, it may need to address compositions, formulations, sequences, assays, protocols, ranges, controls, patient populations, methods of treatment, manufacturing steps or diagnostic workflows.  AI-enabled life sciences inventions usually need both.

Who is the inventor when AI helped? Only natural persons can be named as inventors; the Federal Circuit settled that in Thaler v. Vidal (2022). In November 2025 the USPTO replaced its earlier AI-inventorship guidance with Revised Inventorship Guidance for AI-Assisted Inventions, which applies the ordinary test: a person is an inventor if he or she conceived the claimed invention, and an AI system is treated as a tool, like laboratory equipment or modeling software. Using AI does not disqualify the humans who directed it, but the record should show what they contributed. Inventorship errors can be corrected, but investors and litigants look for them.

The disclosure interview.  Counsel should then pressure-test the document. A good interview is not a recitation of the invention disclosure. It is where counsel asks what is missing, what changed, what could be claimed, what should remain secret, who conceived each feature, who owns the work and what investors are likely to ask.

Candor.  If AI-assisted searching turns up prior art, publications, patents or competitor materials that may be material to patentability, share them with patent counsel.  There is no duty to search, but everyone substantively involved in the application, inventors included, has a continuing duty to disclose material information they know about (MPEP 2001).  The USPTO’s April 2024 AI-tools guidance confirms that using AI does not dilute that responsibility.

The goal is not to make the invention look better than it is. The goal is to make the record more complete, more accurate and more useful for legal judgment.

Practical rule Use AI to prepare, organize and find gaps.  Do not use it to bury inconvenient facts or to substitute for patent judgment.

Bucket 3:  Patent Assessment

Once the invention is captured, the next question is not “Can we file?”  Anyone can file something.  The better questions are whether the invention is worth patenting, whether it is mature enough to support a filing and whether a patent is the right business tool.

Search broadly.  A serious assessment starts with a meaningful search, and not one limited to the first few U.S. keywords that come to mind.  A practical search covers U.S., European, Chinese, Japanese and Korean patent filings and PCT publications; WIPO’s 2025 figures show China, the United States and Japan as the three largest sources of patent filings worldwide.  For life sciences, the closest prior art is often not a patent at all: PubMed, preprint servers, ClinicalTrials.gov, FDA labels and review documents, meeting abstracts and published grant abstracts all count.

The assessment then applies the legal and business filters:

• Novelty:  Does any single prior-art reference disclose everything the company wants to claim?
• Non-obviousness:  Would the invention be considered an obvious variation or combination in view of what was known?
• Subject-matter eligibility:  Is the claim directed to patent-eligible subject matter, a recurring issue for AI, software, diagnostics, and natural-product inventions?
• Written description:  Does the disclosure show the inventors possessed the full scope of what will be claimed?
• Enablement:  Does the disclosure teach how to make and use that full scope without undue experimentation?
• Business value:  Does the patent protect something central to valuation, differentiation, licensing, partnership, reimbursement, regulatory strategy or exit?
• Trade secret fit:  Should some material stay confidential instead of being published in a patent?

Subject-matter eligibility deserves special attention.  The Supreme Court’s decisions in Mayo (diagnostic correlations), Myriad (isolated natural DNA) and Alice (abstract ideas on a computer) remain the framework.  On the life sciences side, diagnostic method claims have fared poorly since Mayo and Ariosa v. Sequenom, which is why a diagnostic invention is often best protected through claims to the assay, the reagents, the treatment step or the system rather than the correlation alone.  On the AI side, cases such as Enfish and McRO show that claims to a specific technical improvement stand on firmer ground than claims that apply a generic computer to an abstract result.

The USPTO’s current approach is set out in MPEP 2106 and its 2024 AI guidance, and has been refined by an August 2025 examiner memorandum, the precedential Ex parte Desjardins decision (recognizing specific machine-learning improvements as technical improvements) and 2025-2026 guidance encouraging applicants to support eligibility with technical evidence.  See the USPTO’s subject-matter eligibility page for the current set.  The direction of travel at the Office has been toward eligibility for well-described technical improvements, but this remains an area of active change, and should be checked at the time of filing.

The business payoff.  A reasoned, objective assessment lets a company speak credibly with investors, partners and board members.  It surfaces strengths, risks and likely prosecution issues before the company relies on a filing in diligence.  No lawyer can guarantee what will happen at the USPTO, but experienced counsel can usually anticipate the kinds of issues that will arise: prior-art and obviousness rejections, eligibility challenges, written-description and enablement concerns, and pressure on claim scope.

Investor-ready answer Investors do not just want to hear that a company filed. They want to know whether the company understands what it filed and what is likely to happen next.

Bucket 4:  Professional Provisional Drafting

This is where cost-cutting can become expensive.

Writing your own invention disclosure with AI may be useful. Writing your own provisional patent application with AI may be dangerous.

A provisional application is not examined. That misleads people. Because no examiner reviews it on day one, a weak provisional can sit quietly in the file while everyone assumes the company is protected.

The problem surfaces later:  when the company converts to a nonprovisional (the full application that is examined), when counsel tries to draft claims broader than the original disclosure, when an investor asks whether the provisional supports the current product, or when a competitor challenges priority.  The rule is unforgiving.  A later claim gets the provisional’s filing date only if the provisional describes that claimed subject matter in the manner Section 112 requires.  In New Railhead v. Vermeer (Fed. Cir. 2002), the provisional omitted one claimed feature, the claims lost their early date, and the patent was invalidated by the company’s own earlier sales.  The Federal Circuit’s decision in Dynamic Drinkware is a useful reminder that priority depends on whether the earlier filing provides the necessary support.

A strong provisional is therefore drafted with the future in mind.  It must provide written-description support (MPEP 2163; Ariad) and enable the invention (MPEP 2164; Amgen v. Sanofi) across the full breadth of the claims the company will want.

A good provisional should describe alternatives, include useful drawings, preserve fallback positions, avoid unnecessary admissions, describe the business-relevant embodiment and leave room for reasonable pivots.

What that means for a life sciences provisional.  Amgen v. Sanofi (2023) held that a broad genus of antibodies defined by function was not enabled by a few dozen examples plus a screening method.  Juno v. Kite (Fed. Cir. 2021) reached a similar result on written description for broadly claimed CAR constructs.  The lesson is concrete: a provisional that will later support genus claims needs working examples, a representative spread of species, structural or mechanistic explanation of why the genus works, and the ranges, formulations, and controls that matter.  Prophetic examples are permitted but must be written in the present or future tense, never as if performed.

Beyond that, a good provisional describes alternatives, includes useful drawings, preserves fallback positions, avoids unnecessary admissions about the prior art, describes the business-relevant embodiment and leaves room for reasonable pivots.  Investors know companies pivot.  Products change, data improves, models evolve, assays are refined.  No provisional can predict every turn, but a well-drafted one preserves far more room than a thin AI-generated filing prepared from a pitch deck.

Why choosing the right patent attorney matters

Choosing patent counsel is not simply a vendor decision.  The application filed today may become the priority document for continuations, foreign filings, investor diligence, licensing, pivots, enforcement and maintenance decisions for the life of the patent family, potentially two decades.

The size of the firm is not the issue.  What matters is whether the attorney understands the technology, the business objective, the investor audience, the likely prosecution path, and the diligence environment the company will face.  Reasonable questions to ask a prospective attorney include:  how many applications have you prosecuted in this technical area, what do you expect the main rejections to be, how will you draft for a pivot, and who on your team will actually do the work.

Confirm that counsel is registered to practice before the USPTO.  Only registered patent attorneys and agents (and a narrow category of individuals granted limited recognition) may represent applicants in patent matters; the USPTO maintains a page on finding a registered practitioner.

A low-cost filing may be appropriate for a low-value idea.  When the company’s valuation depends on the technology, the provisional should not be treated as a commodity.

Practical rule The cheapest provisional becomes the most expensive one if it publishes the roadmap without supporting the claims the company later needs.

Bucket 5: The Patent Roadmap

The provisional filing is not the finish line.  It is the start of a process that, managed well, can protect the company for roughly twenty years from its first nonprovisional filing.  (The provisional year does not count against that term, which is one of the provisional’s advantages; for FDA-regulated products, patent term adjustment and patent term extension can add more.)

When	Patent step	Why it matters and what the company should do
Now	Disclosure review and confidentiality check	Identify who knows what, whether each disclosure was protected, whether any public disclosure has started the U.S. one-year clock or foreclosed foreign rights, and what should stay confidential.
Weeks 1-2	Disclosure interview and patent assessment	Capture the invention, confirm inventorship and ownership, review prior art, evaluate novelty, non-obviousness, eligibility and business value.
Weeks 3-6	Provisional drafting and filing	Draft for future claims: alternatives, written-description and enablement support, drawings, data and the embodiments investors will ask about.
Months 1-12	Provisional build period	Capture improvements, new data, product changes, and model updates in follow-on provisionals.  Log investor and partner disclosures.  Decide what stays a trade secret.
Around month 9	Conversion planning	Decide on nonprovisional and/or PCT strategy, and foreign jurisdictions early, so the 12-month deadline does not force rushed decisions.
By month 12	Non-provisional and/or PCT filing	Preserve the priority date for disclosed subject matter.  The same 12-month window governs foreign filings claiming the provisional.
2+ years often	USPTO examination	Respond to office actions (102, 103, 101 and 112 rejections) with claims aligned to the business.  If investor timing requires a faster first action, Track One prioritized examination is available for an added fee.
After allowance	Issue, continuations and maintenance	Pay the issue fee, file any continuation or divisional application before the parent issues, and docket maintenance fees at 3.5, 7.5 and 11.5 years.

Office actions are normal, not a sign of failure. The USPTO explains that each requires a signed written response addressing every ground of rejection and objection; see responding to office actions.  Maintenance fees on U.S. utility patents are payable without surcharge between 3 and 3.5, 7 and 7.5, and 11 and 11.5 years after issuance, with a six-month surcharge window after each; see USPTO maintenance-fee guidance.

International decisions also need to be made early. A PCT application does not itself create a worldwide patent, but it preserves the option to enter national or regional phases later, typically at 30 or 31 months from the priority date; see the USPTO’s Patent Cooperation Treaty overview.

Timing and cost vary with technology, strategy, claim scope, jurisdictions and quality of representation.  The larger point is constant: a patent position is managed over time, not filed once.

The provisional year is a build year.  Companies file and keep building.  They improve the product, update the model, generate better data, refine the assay, identify new uses, meet investors and enter partnerships.  If those improvements are not captured in follow-on filings, the original provisional may not protect the business the company is actually becoming.  Each quarter of the provisional year, someone should ask:  Is a follow-on provisional needed?  Should new data be added?  Should a new embodiment be filed before it is disclosed?  Should some material stay a trade secret?  Should the strategy be narrowed, broadened or redirected?

Practical rule Treat the 12-month provisional year as a build period with a calendar, not a waiting period.

Plain-English FAQ for Investors and Founders

What is a provisional patent application?

An early, unexamined U.S. filing that establishes a priority date for the subject matter it adequately describes.  It expires after 12 months and never becomes a patent by itself; a nonprovisional or PCT application claiming its benefit must be filed within that year.  (A missed deadline can sometimes be cured within two further months if the delay was unintentional, but no one should plan around that.)

Is an AI-written provisional patent application a good idea?

AI can help organize a disclosure.  A provisional, however, must be drafted with future claims, written-description and enablement support (Section 112), priority and likely business pivots in mind.  That is legal and technical judgment, not formatting.

Can AI be an inventor, and does using AI affect who is named?

No, only people can be inventors.  Under current USPTO guidance, a person who conceived the invention is the inventor whether or not AI tools helped, and the AI is treated as a tool.  The record should show what the humans contributed.

What makes a great invention disclosure?

It identifies the problem, the technical improvement, what was actually built, alternatives, supporting data, drawings, failed approaches, who conceived what, ownership facts, known prior art and expected product changes.

Why assess patentability before drafting?

Because a filing should rest on novelty, non-obviousness, eligibility, adequate support, business value and a trade-secret decision.  Filing first and asking later creates false confidence.

What should investors ask before trusting “patent pending”?

What was filed, what it covers, what prior art was considered, whether any public disclosure preceded it, whether the provisional supports the current product, whether improvements have been captured, whether every inventor has assigned, and whether qualified patent counsel has mapped the next steps.

Why does the choice of patent attorney matter?

Because the first filing shapes prosecution, continuation strategy, foreign filings, licensing, diligence, enforcement and maintenance decisions for the life of the patent family, up to twenty years from the first nonprovisional filing.

What to Expect from the First Patent Conversation

A good first patent conversation does not begin and end with “tell us your idea.”  It covers disclosure history, confidentiality, inventorship, ownership, AI-tool use, prior art, business relevance, investor timing, commercial roadmap, technical maturity, trade-secret alternatives, budget and next steps.

For the company, the goal is not to file quickly.  It is to file intelligently.  For investors, the question is not whether a company is patent pending.  It is whether the company’s patent process shows discipline, judgment and alignment with the business.

A Note on Patent Scams and Filing Confusion

Be careful about invention-promotion firms, non-USPTO solicitations, and official-looking notices.  The USPTO maintains a scam prevention page and a page on non-USPTO solicitations.  If a notice looks like an invoice, a renewal demand or a required publication fee, confirm with counsel that it actually came from the USPTO before paying.  Early-stage companies without docketing systems or established IP counsel are the usual targets.

A Special Note for Rhode Island Companies

Rhode Island Commerce has offered patent reimbursement through its Invention Incentive Program, which in its first round reimbursed up to $5,000 per applicant in USPTO fees and legal fees of Rhode Island-based registered practitioners for provisional or nonprovisional filings, for Rhode Island small businesses and resident inventors.  That round closed in May 2026.

Patent costs are real, and non-dilutive support can help a company take the process seriously earlier than it otherwise would.  Confirm current availability and eligibility before relying on any state program, but Rhode Island innovators should not leave the opportunity unexplored.

Conclusion

Patent basics have not disappeared in the AI age. They have become more important.

AI can help companies prepare better disclosures, run broader searches, organize technical information and make the first meeting with patent counsel more productive.  That is a good thing.

But AI can also make weak patent work look stronger than it is.  It can create polished documents that do not support the claims a company will need.  It can encourage premature disclosure.  It can make a company feel protected before it has done the work that a real patent position requires.

Our view is simple:  use AI to prepare, not to pretend.

The companies most likely to earn investor confidence will be the ones that control disclosure, capture inventions carefully, assess patentability and business value, draft provisionals professionally, choose counsel thoughtfully and manage the patent roadmap over time.

That is patent basics in the AI age.

Selected Source Links for Web Editor

Statutes, rules and USPTO guidance

• MPEP Section 201.04 – provisional applications (35 U.S.C. 111(b))
• MPEP Section 2001 – duty of disclosure, candor and good faith (37 C.F.R. 1.56)
• MPEP Section 2106 – subject matter eligibility
• USPTO 2024 AI subject-matter eligibility guidance – 2024 AI guidance update, August 2025 memorandum, Ex parte Desjardins, and 2025-2026 eligibility-declaration guidance
• USPTO AI tools guidance – use of AI-based tools in practice before the USPTO
• USPTO 2024 AI subject-matter eligibility guidance – AI-related inventions (effective July 17, 2024)
• USPTO guidance on AI-based tools – confidentiality, candor and export-control cautions (April 11, 2024)
• USPTO Revised Inventorship Guidance for AI-Assisted Inventions – November 28, 2025
• MPEP 2163 – written description
• MPEP 2164 – enablement
• USPTO responding to office actions – prosecution process
• USPTO maintenance fees – utility patent maintenance-fee windows
• USPTO PCT overview – international filing
• USPTO finding a registered practitioner – who may represent applicants
• USPTO scam prevention – invention-promotion scams and non-USPTO solicitations

Cases

• Mayo v. Prometheus, 566 U.S. 66 (2012) – diagnostic method eligibility
• Ass’n for Molecular Pathology v. Myriad, 569 U.S. 576 (2013) – natural products
• Alice v. CLS Bank, 573 U.S. 208 (2014) – abstract ideas
• Enfish v. Microsoft (Fed. Cir. 2016) – software technical improvement
• McRO v. Bandai Namco (Fed. Cir. 2016) – software technical improvement
• Ariosa v. Sequenom, 788 F.3d 1371 (Fed. Cir. 2015) – diagnostic claims after Mayo
• New Railhead v. Vermeer, 298 F.3d 1290 (Fed. Cir. 2002) – provisional must support later claims
• Dynamic Drinkware v. National Graphics (Fed. Cir. 2015) – provisional support on the prior-art side
• Ariad v. Eli Lilly (Fed. Cir. 2010) (en banc) – written description
• Juno v. Kite, 10 F.4th 1330 (Fed. Cir. 2021) – written description for biologic genus claims
• Amgen v. Sanofi, 598 U.S. 594 (2023) – enablement of functional genus claims
• Thaler v. Vidal, 43 F.4th 1207 (Fed. Cir. 2022) – AI cannot be an inventor
• Stanford v. Roche, 563 U.S. 776 (2011) – present assignment language

Others

• WIPO IP Facts and Figures 2025 – global filing statistics
• Rhode Island Commerce Invention Incentive Program –  patent reimbursement opportunity

Disclaimer

This article is for general informational purposes only and does not constitute legal advice. Companies should consult qualified counsel about their specific facts.