The money moved to the long, boring tasks
Sail Research raised $80M for agents that run for days, not seconds. The same shift says you are automating the wrong half of your raise.
On 25 June 2026, Sail Research announced $80 million across a seed and a Series A, at a $450 million valuation. Kleiner Perkins led the Series A. Sequoia led the seed.
The company builds infrastructure for what it calls long-horizon agents. Not a chat that answers in four seconds. An agent that works on one task for hours or days.
They also ship a sandbox that stays alive for days and only bills for the time the agent is doing something.
The shift hiding in that round
Read the pricing model, not the pitch. They bill for time the agent is actually working, across a session that can last days.
You only build that if you believe the valuable work is long and mostly idle. Waiting, checking, retrying, picking up where it left off.
That is the opposite of how founders use AI today. Most use it in short bursts, for writing.
You are automating the wrong half
Ask a founder mid-raise where they use a model. The answer is almost always the same. Drafting the email. Tightening the deck. Rewriting the update.
All of it is the fast half. All of it takes minutes and is genuinely the part you should be doing yourself, because your voice is the product.
The slow half never gets touched.
Two halves of a raise
The half everyone automates
Short, creative, visible
- Drafting the cold email
- Rewriting the one-liner again
- Polishing a slide nobody reads
- Total time saved per week: under an hour
The half nobody touches
Long, repetitive, invisible
- Checking which funds announced deals this week
- Chasing the eleven people who owe you a reply
- Keeping the tracker current after every call
- Total time lost per week: most of a day
The weekly loop
I run the slow half as a fixed sequence every Monday morning. It takes about twenty minutes and it replaces the two hours I used to lose to it.
Nothing here is clever. It is the same four questions, in the same order, every week.
Monday, 20 minutes
- What moved Which funds on my list announced anything in the last seven days. New signals change priority order.
- Who is overdue Every open commitment past its date, mine and theirs. Sorted by how long it has been sitting.
- What went quiet Conversations with no contact in 14 days and no scheduled next step. These die silently.
- What to do first Five actions, ranked. Not fifteen. A list you cannot finish is the same as no list.
The prompt
Feed it your tracker. It returns the week.
You are running a Monday review of a live fundraise. I will paste my
tracker. Today's date is [DATE].
[PASTE THE TRACKER ROWS: firm, partner, stage, last contact date, what I
owe them, due date, their stated reason, re-contact date]
Produce exactly four lists. No commentary before or after.
OVERDUE FROM ME
Anything I owe past its due date. Sort by days late, worst first. For each,
write the one-line message I should send today. Write it as I would say it,
not as a template.
OVERDUE FROM THEM
Anything they owe past its date. For each, say whether chasing is worth it,
based on their stage in the pipeline. Say NOT WORTH IT plainly when it is not.
GOING QUIET
Every conversation with no contact in 14 days and no scheduled next step.
For each, one line on whether this is a soft pass or a dropped ball. Use the
stated reason field. If there is no evidence either way, say UNCLEAR.
THE FIVE
Exactly five actions for this week, ranked by effect on closing the round.
Not six. For each, one sentence on why it is above the one below it.
Rules: never invent a fact that is not in the tracker. If a field is empty,
treat it as unknown and say so. Do not encourage me.The four-list shape matters more than the wording. Same questions, same order, every Monday. You are building a loop, not asking for advice.
Why this is the durable version
Model quality keeps changing. The loop does not.
I have swapped the model underneath this three times in a year. The prompt barely changed, because the questions did not change. That is what makes it worth setting up.
Sail Research raised $80 million on a version of the same bet. The long, unglamorous, repeated task is where the value sits.
Your raise has one of those, and it is running right now whether you look at it or not.
This prompt and the tracker layout it expects are on the resources page. All free, nothing gated.