The short answer is yes, you can, but not all of it, and not without reading it as if someone else had written it, because in a sense that’s exactly what happened.
The longer answer starts with the moment I first got a finished draft of a sponsor update from a machine. It was correct, well organized and better worded than what I would have written myself on a Friday at six in the evening, and that’s exactly what worried me.
Your sponsor reads the report, and reads you along the way
A sponsor update has two layers. The first is information: what happened, what slipped, what you need. The second, less visible, is a signal: is the PM in control of the project, do they know what’s going on, can they be trusted.
A sponsor who has been reading the same PM’s reports for six months knows how that PM writes. They know when the PM is worried, because the sentences get shorter. They know when something is being skirted, because more generalities appear. They can’t always put a name to it, but they feel it.
When a report suddenly arrives smoother, more even and without those signals, the sponsor may notice nothing, or may notice that something is different and start wondering what changed.
Where the machine genuinely helps
In an earlier article I wrote about the structure I’ve used in sponsor reports for years: status, consequence, options, recommendation. When I held it up against working with AI, the line came out fairly clearly.
The machine does the status well. From notes, email threads and changes on the project board, it can put together two sentences about what happened. It does this faster than I do and often with fewer omissions, because it has no tendency to forget things that are inconvenient for it. I have one condition, though: I check every date and every number against the source, because the machine can build one false sentence out of two true facts.
It can propose options too. Sometimes it suggests a path I hadn’t thought of, because I was too close to the problem. But I always check whether each of them has a real price, because the machine happily lists options that sound reasonable and are simply not doable in this particular project.
Where it doesn’t help, and even does damage
The consequence is the first place where the machine gets it wrong, and in a way that’s hard to catch. In a sponsor update, the consequence isn’t about the schedule. It’s about the sponsor’s world: a promise made to the board, the relationship with the client, their own position in the company. The machine doesn’t know these things, because nobody told it. So it writes a consequence that is generic, correct and useless, something like “the delay may affect the project completion date.” The sponsor reads that sentence and knows nothing more than a moment before.
The recommendation is the second place, and this is where the risk is greatest. The recommendation is your opinion, signed with your name. Sponsors often accept it with a single word, because they trust that someone who knows the project stands behind it. If the recommendation was written by a machine and you simply passed it on, the sponsor is making a decision based on a sentence nobody actually thought through.
Picture a draft where the machine recommends moving the deadline, because it’s the safest of the options. On paper it sounds reasonable. Except you know that last month the sponsor promised that date to their own boss, and moving it is the most expensive option of all for them. The machine doesn’t know that, and you do.
The line I’ve drawn for myself is about exactly these sentences, the ones that lead to a decision. Behind each of them there has to be a person who thought it through and can defend it. Whether a machine helped with the rest matters much less.
Tone is information too
There’s one more thing the machine smooths out by reflex: worry. When I write a report about a project that worries me, it shows. The sentences are shorter, “I need to flag” appears, and the recommendation is sharper than usual. A sponsor who knows how I write picks that up in the first paragraph and often calls before reading to the end.
A draft from the machine always sounds equally calm. It describes a healthy project and a project on the edge in the same measured tone. That’s why, after every draft, I check whether the report sounds the way I actually feel about the project right now. If I’m worried and the text doesn’t show it, I rewrite the first sentences in my own words, because they set how the sponsor will read the rest.
Should you tell your sponsor?
I don’t have one answer for everyone here. There are companies where everyone writes with the help of AI and nobody mentions it, and there are companies where the sponsor would feel deceived if they found out later.
I prefer to say it once, at the start of working together, how I work: that a machine helps me prepare part of the report, and that I write the consequences and recommendations myself. It takes one sentence, and it removes the uncertainty on both sides. If the sponsor asks directly, I answer directly. Few things undermine trust as effectively as discovering later that the PM kept something quiet, even if that something was completely innocent.
A test before you hit send
Before you send a report the machine helped you with, read the consequences and the recommendation one more time. For each sentence, ask yourself one question: could I defend this out loud if the sponsor called in five minutes and asked how I know? If you hesitate on any of them, rewrite it from scratch, yourself.
The rest of the report, once you’ve checked the dates and numbers, can stay the way you got it.
All situations described in this article are based on real events, but contain no company names, no individual names, and no data from any specific project.
