It's Monday, 7am. The analysis is already done.

Fountlabs agents work on schedules you set, as an extension of your growth team. They run overnight on your data, so by the time you log on the briefs are written, last week's campaigns are analysed, the forecasts are updated and anything that went wrong has been flagged, diagnosed and given a suggested fix.

While you were out
Scheduled work, Demo Mutual, Sunday night to Monday 06:00
1 needs your OK
Sun 22:00
Pipelines refreshed, new rows onlyEvery night
Done
Mon 02:10
Campaign analysis, last weekEvery Monday
38 campaigns
Mon 03:30
Quarterly forecast re-runEvery Monday
3% under plan
Mon 04:05
Budget pacingEvery day
Meta $22k over
Mon 04:40
Florida home leads down 23%When something moves
Diagnosed, fix ready
Mon 05:15
Test resultsEvery Monday
2 ready, 1 stopped
Mon 06:00
Briefs deliveredEvery Monday
6 people
Ready before your first meeting

The briefs are written

A short brief for each channel owner and one for you: what moved last week, what needs attention and what we couldn't check yet.

Last week's campaigns are analysed

Every campaign, ad set and keyword read on cost per bound policy and expected value, with the ones to cut and the ones to feed.

Forecasts are updated

Monthly and quarterly forecasts for policies, premium and cost per policy, re-run on last week's binds and set against plan.

Issues are diagnosed

Anything that moved more than it should have is flagged, traced to a cause and handed to the right person with a suggested fix.

Pacing is checked

Where every channel will land against budget by month end, and which ones need a nudge today rather than on the 28th.

Partners are audited

Lead quality by partner, quote to bind and early cancellations, so a partner going bad is caught in a week instead of a quarter.

Test results are written up

Experiments that reached enough policies are written up, and any that crossed a limit have already been stopped.

The record is kept

Last week's decisions go in the ledger with what they did, so the quarterly review is mostly written before anyone starts it.

Signals

They watch the numbers your business runs on.

A signal is one number you care about, like cost per bound policy, quote to bind, lead volume by partner or projected loss ratio, with the band it should sit in and the data it depends on. Each one is checked on its own schedule and marked on track, slipping, off track or waiting on data. Overnight an agent goes through them, starting with what's broken, and writes the brief you read in the morning. Every check is kept, so you can see exactly when a number first slipped.

All day, each on its own timerSignals are checked
Cost per bound policy, autoOn track
Quote to bind, home, webSlipping
Lead volume, Meta, FloridaOff track
Leads from comparison sitesWaiting on data
Projected loss ratio, new businessOn track
If a file is late, the number waits for it, so nobody chases a problem that isn't real.
OvernightAn agent goes through them
Starts with what's off track, then what's slipping
Checks the data, the definitions and last week's ledger entries
Notes which numbers are waiting on a late file
Writes it up, with the numbers behind each point
Every step it took is recorded, so you can see exactly what it looked at.
Before your first meetingThe brief is in your inbox

Florida leads on Meta are off track, down 23% since Thursday's audience change.

Home quote to bind is slipping. Two more weeks like this and it's off track.

Comparison site leads are waiting on data. Today's file hasn't arrived yet.

If not much happened, the brief is short.

Flagged, diagnosed, and a fix waiting for your OK.

When a number moves more than it should, an agent doesn't just raise an alert and leave you to it. It checks the usual causes the way your best analyst would: spend, tracking, competitors, seasonality and every change your team made that week. Then it writes up the cause and suggests a fix.

The fix waits for a person to approve it. Most of the time, that's the only part of the investigation you have to do.

Florida home leads down 23% week on week
Flagged Monday 04:40, diagnosed by 04:52
Waiting for Maria
Spend is flat, so the budget didn't move
Tracking and lead events are firing normally
Competitor bids in Florida are stable
Ledger: Thursday's audience change narrowed the Florida lookalike seed
CauseThursday's audience change cut reach in Florida by 41%.
Suggested fixRestore the previous seed in Florida only and keep the change everywhere else.Expected: about 180 more leads a week at the same cost per bound policy.
Approve the fixEditDismiss

Forecasts that update themselves.

Every week the agents re-run your monthly and quarterly forecasts for policies, premium and cost per policy on the latest binds, and set them against plan. When the quarter starts slipping you hear about it in week five, with the channels that would close the gap, instead of in week twelve.

Q4 bound policies, forecast against plan
Actual to week 5, then forecast. Re-run Monday 03:30
3% under plan
Forecast
11,840
Plan
12,200
Gap closes if
Search keeps its pace

You set the schedule. They do the work.

Tell Fountlabs what you need and when, in plain language: the Monday channel brief, a daily pacing check, the board pack numbers on the first of the month, a diagnosis whenever cost per policy moves 15%. The agents do it on time and send it to the right person.

  • Work to your definitions, your plan and your thresholds
  • Show their working on every number in every brief
  • Hand anything that needs a decision to a named person
  • Run on pipelines, so a Monday of analysis doesn't run up your warehouse bill
Schedules
Set once, in plain language
New schedule
Channel brief and last week's campaign analysisEvery Monday, 06:00Media team
Budget pacing, flag anything 10% off planEvery day, 07:00Maria
Monthly forecast and the board pack numbersFirst of the monthHead of growth
Quarterly review draft, with every decision from the ledgerEnd of quarterLeadership
Diagnose the cause and suggest a fixWhen cost per policy moves 15%Channel owner

Yesterday's numbers, in your inbox.

Each person gets the brief that matters to them, in plain language, before their first meeting. Every line opens up to the analysis behind it.

Your morning brief
Written overnight, ready at 06:12
New

Yesterday wrote 61 auto policies at $204 each, better than your target. Home was quiet.

Florida lead volume on Meta dropped 23% after the weekend. Spend is flat, so it looks like the audience. The detail is one click away.

One number is on hold. Home leads from comparison sites are waiting on a late partner feed, so we haven't scored them yet.

The knowledge graph

They understand your book because it's written down.

Underneath the agents is a graph of what insurance marketing cares about. It holds your products and how they roll up, what counts as a bound policy, how you group your channels, the vetted queries behind your numbers and the dated incidents that explain a strange week. It teaches agents how your book works before they touch the data.

It gets sharper the more your team uses it. Every night it looks at the questions people keep asking, the ones an agent couldn't answer and the answers that got a thumbs down, and suggests what to add. Suggestions stay out of your numbers until someone checks them. That keeps unproven AI guesses out of your numbers. Once someone confirms a suggestion, it becomes part of the graph.

ConceptBound policy
ConceptCost per bound policy
ConceptLoss ratio
SkillDiagnosing a lead drop
ChannelComparison sites
ProductAuto
ProductHome
SuggestedCount both aggregators as comparison sites
IncidentPartner feed down, 3 to 9 March
  1. Suggested

    Anything an agent writes starts here. It's only a suggestion, and it never goes into an answer.

  2. Checked

    Someone has checked it and it's safe to use. Answers only ever use checked or approved facts.

  3. Approved

    Someone on your team has signed it off. Only a person can approve something.

Why they get it right

A general AI agent plugged into your CRM waits to be asked, then guesses.

A general-purpose assistant sees tables. It doesn't know that a quote isn't a bind, that your aggregator leads arrive twice, or that last March looks odd because a lead partner's feed went quiet for a week. Fountlabs agents do, because they work from the customer data platform, your definitions, the profitability layer and the ledger.

A general agent on your CRMA Fountlabs agent
Waits for someone to askWorks to a schedule and on triggers you set
Tables and column namesYour products, your funnel and your signed-off definitions
Clicks and conversions per platformEvery touch joined to the policy it became
No idea what a customer is worthExpected value of every customer, from your own models
No memory of what your team triedThe full ledger of decisions and what they did
The whole warehouse, scanned again for every questionPipelines it built once and keeps fresh

And when you do have a question, ask.

The same agents answer questions between scheduled runs, from Fountlabs or from inside Claude. They read your definitions and the ledger first, and every answer comes with the query behind it.

MR
Maria, Growth09:41
Which of our auto campaigns wrote the most profitable policies in September, and what did each policy cost us?
Analytics agent09:41
Read your definitions of a bound policy and cost per policy0.4s
Matched September auto policies to the campaigns that brought them in2.1s
Checked policy and spend data are current to yesterday0.3s

Safe Driver Search wrote the most profitable business, 412 policies at $198 each with a projected loss ratio of 0.58. Meta prospecting in Texas is close behind on cost, but those policies are cancelling faster.

Want me to turn this into a report for Monday's meeting?

Every number comes with its receipt.

Open any line of any brief to see how the number was made: the definition it used, the policies it counted, the spend it matched and how fresh each source was. If a data file is late, the number waits for it, so you fix the file and don't go chasing a problem your business doesn't have.

How this number was made
$198 per bound policy, Safe Driver Search, September
Ran 09:41
MeasureCost per bound policyAd spend divided by bound policies, using your signed-off definition
Policies412 bound policiesIssued 1 to 30 September, first premium paid, cancellations inside 30 days removed
Spend$81,576 in Google Ads spendEvery campaign in the Safe Driver Search group
MatchingFirst touchEach policy credited to the campaign that first brought the customer in
FreshnessCurrent to yesterdayPolicy file and ad spend both arrived on time
Open the full query, or copy it for your analytics team.
What agents won't do

Spend your money

Agents diagnose, recommend and prepare. A person on your team approves anything that touches an ad account, a customer or a budget.

Make up a number

If the data can't answer the question, the agent says so in the brief and shows you what's missing.

Rewrite your definitions

Agents can suggest a definition when they spot a gap. It stays a suggestion until it is checked, and only a person can approve it.

Where it sits

Part of one system.

The agents work from everything underneath them. When a number moves, they check what changed in the data and in the ledger before they tell you why, and they put a dollar figure on the answer.

Marketing analytics agents

Agents that work on schedules you set, alongside your growth team. By Monday morning they've written the briefs, analysed the campaigns, updated the forecasts and diagnosed what went wrong. Each fix waits for your OK.

Start every week with the work done.

See Fountlabs