apart program called an LLM + deterministic DOI calculator with a citation trail that is crucial to the project to handle Texas allocation math and burden layers. Evaluate all of this in the context of Moore’s Law. How soon will land administration analysts be replaced by a computer program? When this kind of project is finally undertaken (and it will be), best practices suggest that the project should start out narrow, with one operator, one style—prove accuracy and traceability, then scale templates and integrations toward the final product. Predictable Barriers Long title opinions, supplemental opinions, inconsistent formatting, and frequent amendments will increase the scope of work necessary to produce an enterprise-ready final AI product. The risk of manual error, time-to-pay delays, and audit exposure also add to the mix. AI can accelerate the review and math while preserving legal accountability, but extensive deep thinking requirements will balloon both time and cost. The onboarding of new staff to contribute to the development of the new AI program and its maintenance can’t be estimated realistically up front. What the AI System Must Actually Do End-to- End In a nutshell, the AI System must ingest PDFs and Word scans and process them through an OCR program capable of clarifying tables, exhibits, and metes-and-bound plats. Remember, Texas does not use the Jeffersonian System (Township- Range-Section). Once necessary documents are available to the AI System, it must understand ownerships, tracts, leases, burdens, lateral footage by tract, producing intervals, and dates, at a minimum.
(tract % x lease royalty x burdens = WI/NRI math). It must validate sum-to-one decimals, tract totalization, conflict detection, varying effective-dates logic . It must leave an audit trail—every decimal links to specific opinion text, exhibit, or table with a snippet and page reference. Finally, it must create and export to each accounting system needing to integrate the new data into its module. This must be done as an Excel spreadsheet (CSV and/or XLSX) using adapters. Human review in this entire process is required to ensure accuracy and completeness. Likely and Feasible Alternatives Ever heard the joke (or any number of others similar to it), How do you eat an elephant? One bite at a time. So it seems that incorporating AI to do the work of the division order analyst could begin with one bite at a time, at least for now. That means looking for specific smaller tasks we must do in the course of our work that could be quickly and easily done by an AI program then turned over to the analyst to complete. This would allow the analyst to concentrate on the more difficult deep thinking tasks. These smaller tasks might include time- consuming work such as re-formatting the template (or attorney’s DOI spreadsheet) to function properly for this particular well/ payout. Another small task might be fixing a DOI spreadsheet that doesn’t total to 1.00000000 and contains a number of burden groups to be internally calculated by the DOI database. There are many similar tasks depending on the workflow procedures and work product requirements of each employer that an analyst could identify if asked. Everything discussed thus far applies equally to lease analyst work as well as land analyst work and beyond, to virtually every discipline in upstream oil and gas. It becomes easy to see why we remain quite distant from an AI System that can replace fully anyone in Land Administration.
It must use a deterministic DOI calculator
28
N at i onal A ssociation of D i v i s i on O rder A nalys t s
Made with FlippingBook Ebook Creator