Three decisions, applying functions Orisdale already publishes
The Energy page is a solution blueprint. It documents three designs: drilling performance and contracts, land and lease obligations, and operational health. Each design names a user, a decision, source categories, a grain, the calculations the business would own, the output, the limit on interpretation, missing-data behavior, and human review. None of them is a demonstrated implementation. The patterns are synthetic. They do not report measured savings, and they are not customer deployments.
Energy is an industry: a place where a decision is situated, not a function beside revenue, finance, tax, operations, and executive intelligence. Each of the three designs applies one or more of those published functions to a decision that is specific to drilling, land, or asset operations — a contract rate that makes a time record a finance review, or a priority list that an executive review reads. That is how the pattern stays bounded: a decision earns a place on this page only when the sector specificity is real, not by broadening the page to any operational question.
This note stays inside those three decisions. It does not add production optimization, predictive maintenance, or autonomous control. Those are different problems, with different safety and authority questions, and they are not what the blueprint describes.
- Drilling performance and contracts
- Land and lease obligations
- Operational health
A synthetic drilling period, including a missing day
Label: Energy design example. Synthetic. Solution blueprint. Not a customer deployment. Not a demonstrated runtime.
A drilling-performance analyst is preparing a contract conversation. The decision is which wells, rigs, or contract periods need a person to look before that conversation. The grain is a well or rig, by day or by a declared contract period. For this example, take one synthetic well and one declared contract period. No operator's records are used. No threshold is stated, because the blueprint does not publish one.
Separate four kinds of statement. They are easy to blend into a single "insight," and the blend is the failure.
Observations. Drilling time records are present, a contract rate is present, and cost categories are named only as categories. One day inside the period has no time code. That absence is an input fact. It is not yet a calculation, a breach, or a cause.
Calculations. The blueprint's calculations are time classification, an approved contract rate, and an exception flag the business has defined. Classification uses the codes that exist. The rate applies only where classified time and the approved rate are both present. The missing day is not given a code to complete a total. If the rule cannot classify the period, the output is a missing-data label, not an exception and not a silent pass.
Contractual rules. The review list contains the periods the organization's own rule selects, with the governing fields attached. This article does not invent that rule's numbers. A drafted sentence does not put a period on the list.
Proposed explanations. The agent may say which fields are present, which day has no time code, and whether the approved rule listed the period or labeled it incomplete. It may not name an operational event as the cause, and it may not recommend a field action. The explanation stays proposed until a person confirms the exception. Confirmation is what makes a review item an operational finding or a contract claim.
A fluent summary is tempted to complete the missing day so the paragraph reads cleanly. The blueprint's design response is to forbid that: a missing time code, rate, or period stays missing and is labeled, rather than filled to make the output look complete.
How the three patterns compare
The drilling example is one pattern. Land and lease and operational health are documented at the same maturity. They share a review-list shape and they do not share a grain.
| Pattern | Decision | Grain | Ruled output | Missing data | Human review |
|---|---|---|---|---|---|
| Drilling performance and contracts | Which wells, rigs, or contract periods need review | A well or rig, by day or contract period | Review list from the approved exception rule, with governing fields | A missing time code, rate, or period stays labeled missing | A person confirms the exception before it is a finding or a contract claim |
| Land and lease | Which obligations need attention before a date passes | A lease or tract, with obligation date and as-of date | Queue with governed status, date, and source fields | A missing date or interest is not estimated | Land staff confirm a status change before it is the position of record |
| Operational health | Which assets or work orders the existing priority rule lists | An asset or work order over a declared period | Prioritized list with the governed priority and its fields | An incomplete work order or downtime record stays incomplete | Reliability staff confirm the item before a maintenance action is recommended |
A lease status without an as-of date is not a current status, and the agent may not infer that an obligation was satisfied. Operational health stops at downtime classification, a priority rule, and an overdue status. It does not predict a failure, recommend a part, or close a work order.
The same industry crosses three functions
A link is published only where the documented decision already involves that function.
Drilling performance and contracts spans Operations Intelligence and Finance Intelligence. The exception is an operations review of well and time records, and a contract rate makes it a financial review. Land and lease spans the same two functions: obligations are operational commitments, and an interest or a status change carries financial review. Operational health spans operations and Executive Intelligence. The priority list can be read in a management review. That reading does not create a new calculation or a field instruction. Tax and revenue are not linked, because the documented decisions do not support them.
The reference pattern, and what it does not assume
The architecture on the Energy page is a proposed reference pattern, not a demonstrated runtime. Operational sources feed Alteryx One for governed preparation and business logic. BigQuery is where governed analytical outputs would be read. Semantic context says what a time code, a contract period, a lease status, or a downtime category means. Where a technical team builds an agent, the pattern names Gemini Enterprise Agent Platform. Google Cloud describes that platform as the place for technical teams to build, scale, govern, and optimize agents, and the product page identifies it as formerly Vertex AI. The same page says to use the Gemini Enterprise app to securely register, manage, and govern custom-built agents. A person confirms a finding after that.
Orisdale's work is decision design, semantic context, implementation, and validation. The pattern is not an endorsement by Alteryx or Google Cloud. Snowflake and Databricks remain peer choices when the data already lives there. See Google Cloud, the architecture, and Governed Business Agents on Trusted Platforms.
In that selected path, Alteryx documents Live Query for Google BigQuery as Private Preview. Functionality and availability might change before general release. Supported operations can be delegated as queries that BigQuery runs. The limitations page does not establish that a result is stored as a physical table. Confirm persistence for the output path you select. Do not assume a persisted table. See Live Query for Google BigQuery.
Decision support is not a safety system
Each pattern outputs a review list or a queue. A person confirms an exception, a status change, or a maintenance priority before it is a finding. The blueprint does not close a valve, dispatch a crew, amend a lease, or file a contract claim.
An unlabeled gap can look like a clean period. A labeled gap keeps silence from being read as compliance. It does not make the operation safe. Safety systems and equipment interlocks have other owners. This blueprint is not those systems. "The approved rule listed this period" is a statement about a flag. "The well should be shut in" is an instruction the blueprint does not produce. "The pump will fail next week" is a prediction the health pattern does not make.
The packet for a consequential flag still needs the grain, the rule version if one exists, the governed fields, and the missing inputs. Human-in-the-Loop AI is the general note.
Where to take one of the three decisions
Request a strategy session about one decision, not about energy in the abstract, on the contact page. The session does not promise a result, a threshold, or a production date.
Read the Energy blueprint for the three designs in full, including the questions the page already answers about scope, maturity, data, review, and the next step. Use Operations Intelligence, Finance Intelligence, and Executive Intelligence for the functions a given decision actually spans.