Pacific Power bill assistance: AI finds 37,000 eligible homes
A Pacific Power bill assistance campaign in Oregon, guided by Bidgely's AI analytics, identified about 37,000 additional households that may qualify for its Low-Income Discount program and contributed to about 9,000 new enrollments, the utility said on October 27, 2026. The work earned a 2026 AEIC Achievement Award. For any company running a discount or rebate program, it shows how customer data can direct outreach to the people most likely to qualify, while the eligibility decision stays with people.
Why eligible customers were missing out on discounts
Pacific Power had a discount program that many qualifying Oregon customers were not using. Its Oregon Low-Income Discount program cuts bills by between 20% and 80% for customers at or below 60% of state median income, adjusted for household size. But a discount only helps the households that know about it and apply, so the utility set out to find which of its customers might qualify and reach them directly.
Pacific Power serves more than 800,000 customers in Oregon, Washington and California and is part of PacifiCorp.
What Pacific Power used to find them
The utility used Bidgely's UtilityAI-powered Analytics Workbench. With it, Pacific Power reviewed census information, customer data, household energy use and communication preferences, and identified about 37,000 additional households that may be eligible for the discount. Bidgely, based in Los Altos, California, describes them as hidden but eligible households.
How the AI shaped outreach, not eligibility
The analysis decided who to contact, not who qualifies. Pacific Power says the AI-powered work was used to inform customer outreach only. Customers chose whether to apply, every application went through the program's established process, and customer information was handled under the utility's existing privacy policies.
The campaign data also showed where people dropped out. Some customers started an application and did not finish it, so Pacific Power changed its messages to remind applicants to have their account number and income information ready before they began.
"Technology is most valuable when it helps solve a real customer problem." — Barbara Modey, customer satisfaction market research manager at Pacific Power
What the outreach achieved
The effort contributed to about 9,000 additional enrollments in the Low-Income Discount program, the company says. Between 2025 and 2026, enrollment grew from about 69,000 to 78,000 customers. Awareness of Pacific Power's assistance programs rose 10% in the J.D. Power Electric Utility Residential Customer Satisfaction Study, while satisfaction among program participants held steady at about 90%.
The work earned Pacific Power a 2026 Achievement Award from the Association of Edison Illuminating Companies, announced on October 27, 2026.
What the Pacific Power bill assistance case means for utilities and service companies
For a utility, the value is in targeting. The company focused its outreach on the households the data flagged, and it used the results of the campaign to fix the step where applicants were getting stuck. James Owen, vice president of customer experience and innovation, called reaching the right customers with the right message a prudent use of outreach resources.
The same method fits any company that runs a benefit, discount or rebate program that customers must apply for: insurers, telecom providers, banks and public agencies. Usage and account data can point to who is likely eligible, while the formal eligibility check stays with the existing process. The safeguards Pacific Power describes, a human-run application process and existing privacy rules, are the part other companies would need to copy.
Questions and answers
How did Pacific Power use AI to find eligible customers?
It used Bidgely's UtilityAI-powered Analytics Workbench to review census information, customer data, household energy use and communication preferences. The analysis identified about 37,000 additional households that may qualify for its Oregon Low-Income Discount program.
Did the AI decide who gets the discount?
No. Pacific Power says the analysis was used only to inform outreach. Customers chose whether to apply, and applications went through the program's established process.
How many customers enrolled as a result?
The effort contributed to about 9,000 additional enrollments, and total enrollment grew from about 69,000 to 78,000 customers between 2025 and 2026.
MOZGII view
How we would build it
Finding people who qualify for a discount but never applied is a good first job for AI at a utility. The decision stays with people and the existing rules, and the model only ranks who to contact first.
For a program like this, we would start with the data the utility already holds: billing history, payment patterns, energy use by season, past assistance and how each customer prefers to be contacted. We would join that with public census data at the neighborhood level. The model would score each household on how likely it is to qualify, and the outreach team would pick how far down the list to go based on budget. A program manager would review samples of flagged households each month to check that the scores make sense and that no group of customers is being missed or singled out.
The weak spot is the step after the message. A well-targeted letter does nothing if the application is long, asks for documents people do not have at hand, or times out online. That is why the drop-off data in this case matters as much as the targeting.
The one metric we would watch is completed enrollments per household contacted, split by outreach channel. Next, we would build a short, prefilled application that uses what the utility already knows, so fewer people stop halfway.
Sources
Where the facts in this story come from.