An AI agent's usefulness depends partly on the information it can access and the actions it can take. In this Mile High Makers video, Jeff Schell describes his approach to building custom workflows around patent and legal work.
Separate public reference material from confidential records
Jeff discusses making public legal resources available for research while applying different controls to client and invention information. That distinction is important: a tool that can retrieve a public manual should not automatically have permission to transmit confidential records or change a client system.
His description of particular tools and infrastructure reflects the recording. Local processing can be part of a data-control approach, but location alone does not establish security or compliance.
Define the agent's permissions explicitly
- Which sources may it read, and which records are outside its scope?
- Can it write to a system, or only prepare material for review?
- What data may be sent to an external provider under the applicable terms?
- Which actions require human confirmation?
- How will errors, source references, and completed actions be recorded?
Build around the professional task
The conversation also connects Jeff's entrepreneurial experience to patent strategy. A useful workflow should help evaluate the actual business problem, rather than standardizing away the judgment needed to address it.
Start with one bounded use case, such as organizing public references for an attorney's review. Expand only after the team understands its accuracy, data flows, and permission boundaries. Efficiency is easier to assess when the expected output and the responsible reviewer are clear.
Related support from Schell IP
This article provides general information. It does not replace legal advice about a particular invention, company, or transaction.
