How to draft solar investor reports with AI

The monthly investor report is one of the most expensive documents an asset manager produces. It rarely holds anything new: generation, performance ratio, availability, savings, and a page of commentary on what went wrong. Most of the effort goes into pulling numbers from several tools, pasting them into a template, and checking them twice. Very little goes into the commentary, which is the part investors actually read.

That split makes reporting a good fit for AI, as long as you're clear about which parts you hand over.

What to automate

The data pull. Generation, PR, availability, savings and billing figures per site, from one source of truth. If your monitoring data is already standardised across brands, this step is mostly a query.

The comparisons. Actuals against budget, against last month and against the same month last year, with every variance above your threshold flagged. This is where manual reporting goes wrong most often, because one wrong cell reference spreads through a whole table.

The first draft of the commentary. For each flagged variance, the AI drafts a short explanation from the data: which sites, how much energy, which events (downtime, curtailment, soiling, grid outages). Every sentence links to the numbers behind it.

The formatting. Your Word, PowerPoint or PDF template, filled the same way every month, with charts in your branding.

What people should keep

Root causes that aren't in the data. If a site underperformed because a customer disputed an invoice, or a landlord blocked roof access, the data shows the drop but not the reason. Someone who knows the portfolio has to add it.

Anything forward-looking. Forecasts, guidance, remediation plans and their cost. These are commitments, and a person should make them.

Sign-off. A named person approves every report before it goes to investors.

Two rules that make it safe

First, the AI only reports numbers it can trace to source data. If a figure is missing or estimated, the draft says so; it never fills a gap with a plausible guess.

Second, every draft has a reviewer. The point is to turn two days of compiling into an hour of checking, with the reviewer's time spent on the commentary instead of the copy-paste.

Where to start

Pick one report, one month and one portfolio. Run the AI draft alongside your usual process, compare the two line by line, and fix what differs. After two or three months you'll know which parts you trust and which still need a person. That's a far better basis for rolling it out than any demo.

At AMMP, this is part of what Ray, our AI coworker, does: it pulls the portfolio data from AMMP OS or your existing monitoring stack, drafts the report with every number linked to its source, and delivers it as Word, PowerPoint or PDF for your team to review. More at www.ammp.io/ray.

Questions about your own portfolio? Write to sales@ammp.io.