Minerva’s agent now runs on a schedule

Every marketing team asks the same questions on Monday morning. How did last week's campaigns perform? Which segments moved? Where did spend stop paying off? Someone pulls the data, builds the comparison, writes it up, and does it again the following Monday.

Minerva's agent could already answer those questions, and starting today, it can own them. The agent now runs in the cloud, works through your data in Python, remembers how your business runs, and repeats the work on a schedule you set.

It keeps working after you close the tab

The agent now runs on Minerva's servers instead of in your browser. You can start a long analysis, close your laptop, and come back to the finished answer. If you lose your connection mid-run, the agent keeps going and picks the conversation back up when you return.

It sees the whole picture

Minerva loads your connected sources into a data warehouse set up for your business. The agent queries that warehouse directly. It can join ad spend, orders, and customer records in one analysis instead of requesting each piece from each platform’s API. A question that spans sources, like whether a Meta campaign brought in new Shopify customers, takes one query.

It does the math

The agent now has its own sandbox, where it writes and runs Python against the data in your warehouse. It can cover statistical tests, forecasts, cohort and retention breakdowns, and the custom calculations your team would otherwise build in a notebook. It brings back the results as charts, tables, and recommendations.

It remembers your business

The agent keeps two kinds of memory. Your personal memory holds how you work: the metrics you watch and how you like results laid out. Your organization’s memory holds all the shared context: your products, your goals, your definitions. Tell it once that an abandoned cart means someone started checkout, didn’t buy within seven days, and isn’t a subscriber, and every analysis and audience after that uses your definition, for everyone on your team.

It repeats the work

Turn any question into a schedule, once or on repeat. The agent runs it on time, saves each run, and keeps a history you can scroll back through. Pause, edit, resume, or run a schedule on demand. For work you still kick off by hand, save the prompt as a template or a skill and reuse it in one click.

Your analysis is ready for you

Set it once: "Every Monday, compare last week's Meta and Google Ads results by campaign and audience, explain the biggest changes, and recommend where to shift spend."

On Monday morning, the analysis is ready before your team meeting. Meta prospecting CAC rose 22% while Google Ads held steady, which points to a problem on Meta rather than on your site. The agent explains that frequency on your two top ads doubled while spend held flat, a sign of creative fatigue. It checks Meta’s reported conversions against new customers in Shopify, so you know the jump is real rather than a tracking gap. Because it knows your team optimizes for margin, it won’t just chase the highest-ROAS campaign. Sometimes it will tell you to leave spend where it is.

When you ask a follow-up, the agent answers with your goals and definitions already in mind. If the fix is a fresh audience, the agent can build it and send it to Meta that afternoon.

From answer to audience

Minerva’s agent knows which Minerva audiences each campaign targets, so it can trace a change in results to the audience behind it. Where scoring is enabled, the agent can also read Minerva’s propensity scores, which show how closely each person resembles your customers, and compare audiences on them before choosing who to target next.

Then it acts on what it found: it builds the next audience from Minerva’s data on more than 265 million people and sends it to Meta, Google Ads, Klaviyo, HubSpot, or a Lob direct-mail drop.

When you put the audience on a schedule, the agent rebuilds and resyncs it, so it never goes stale.

Try it

Customers: the new agent is live in your workspace today. Ask it: “I run lifecycle and retention. What 15 questions should I ask this week to find problems and opportunities? Run them and bring back the most useful findings.” Then schedule it for every Monday.

New to Minerva? Book a demo and we'll set up your first schedule with you.