AWS says AI agents built with Duke Energy cut grid-study data prep from weeks to hours

September 2026 — Amazon Web Services launched Agentic Grid Planning on Sept. 17, 2026, a program giving utilities AWS-managed AI agents for interconnection studies, and said Duke Energy had cut some data-preparation tasks from two weeks to hours. Here's what happened, what it cost, and what went wrong.

Cost and outcome (as of Sept. 17, 2026)

Claimed outcome: data-preparation tasks went from two weeks of manual work to hours at Duke Energy, per AWS. (AWS)
Cost: pricing not disclosed.
Who decides: engineers direct studies and make final decisions, AWS says. (AWS)
Verification: none independent; vendor release only.

The problem it targets

AWS cites Lawrence Berkeley National Laboratory: more than 2,000 GW of generation and storage were seeking U.S. transmission interconnection at the end of 2025, and projects that entered operation that year had spent a median of more than five years in the process. (AWS)

It adds that requests to connect large loads in ERCOT alone reached about 410 GW in March 2026, citing an ERCOT update to the Texas Senate Committee on Business & Commerce on April 1, 2026, and that demand for studies is outpacing the supply of qualified engineers. (AWS)

FERC Order No. 2023 requires first-ready, first-served cluster studies and study deadlines, while ERCOT, NYISO and SPP are developing new large-load processes, AWS notes. (AWS)

How AWS says it works

The agents run study workflows using the physics-based simulation software, grid models, scripts and engineering standards utilities already use. The simulation software does the deterministic modeling; the agents run and coordinate the workflow and keep a versioned, reviewable record. (AWS)

AWS lists three uses: preparing study-ready cases, running steady-state transmission studies, and codifying engineering practice for reuse. Agents propose a point of interconnection and topology updates, run power flow and contingency analysis, categorize violations, propose mitigations and draft reports; engineers confirm or override at each step. (AWS program page)

AWS says much of each study, such as preparing data, moving it between tools and assembling reports, needs engineering oversight rather than continuous engineering judgment. (AWS program page)

Access and audit trail

The program is available to qualified utilities and grid operators, with AWS Professional Services offering integration support. (AWS) AWS says the agents version study files and record workflow steps so simulation output, AI reasoning and engineer decisions stay distinguishable and traceable. For reuse, engineers direct a first study interactively, then the agents draft a workflow guide from the session for engineers to review. (AWS program page)

Duke's role

Duke engineers supplied transmission-planning expertise and feedback to build the agents and now use them to streamline data preparation, analysis and workflow coordination, AWS says. The two are extending the work to large-load planning. (AWS)

"Our engineers lead the work and make all final engineering decisions, while AI agents provide them with more time for engineering judgment and review," said John Pressley, Duke's managing director of digital strategy and engineering. (AWS)

"Duke Energy maintains one of the largest utility grids in the United States, and our engineers must complete a higher volume of increasingly complex studies as demand for generation and load grows," Pressley said. (AWS)

"Utilities are facing a challenge that scaling their manual processes alone cannot solve," said Joseph Santamaria, AWS's general manager for energy and utilities. (AWS)

What went wrong

Vendor-only. Every result comes from AWS's own release and product page; neither Duke nor a regulator has published figures.
Data prep isn't the whole study. The weeks-to-hours claim covers data-preparation tasks, not full interconnection studies. (AWS)
Unverified speed claims left out. Faster-study and engineer-time percentages on the AWS product page couldn't be independently checked, so we don't use them.
What we don't know yet: whether any study finished sooner end to end, what the program costs, and which other utilities have signed on.

Our follow-up question for Duke Energy

Has using the agents shortened any full interconnection study, and how is accuracy checked?

Sources

Amazon Web Services, “AWS Launches Agentic Grid Planning Program to Accelerate Interconnection Studies,” Sept. 17, 2026. press.aboutamazon.com
Amazon Web Services, “Agentic Grid Planning on AWS” program page. aws.amazon.com

Disclosure: [not in copy] Editor placeholder