Now What

You built it.

You built it.

Now What grows it.

The autonomous growth system for the products you’ve built. It researches, decides, spends and proves every dollar, inside the limits you set.

Building got cheap.Growing didn’t.

AI made shipping nearly free. Now there are a thousand new products a day, and the hard part is everything after launch: who to reach, what to say, where the next dollar goes, and whether it worked. That’s the job Now What does, continuously, for every product you own.

How it works

Research. Decide. Execute. Prove.

One loop, running all the time for every product you own. Each turn ends in evidence, and the evidence sets up the next turn.

A branching amber filament inside a titanium ring, lit from its root outward

The Growth Brain

It reads your product, your market and your numbers, then builds a model of how you grow. Where it has no first-party data, it says so instead of borrowing an industry average.

Research

Reads everything. Guesses nothing.

  • Who buysInferred · high confidence

    Solo founders shipping B2B tools

    Your site + 4 competitor pages

  • PriceObserved

    $29/mo, 20% off annual

    Your pricing page

  • Contribution marginNot measured

    Not measured yet

    Connect Stripe to measure

Illustration

Decide

Every action, judged nine ways.

Proposed action

Move $40/day from Meta prospecting to Google brand search

Reversible
Instantly
Of monthly limit
12%
Audience
Unchanged

Within your limits. Runs on Autopilot.

Cost is one of nine checks, never the only one. A free headline change can still wait for you.

IllustrationSee all nine

Execute

Done, then checked.

  1. Proposed

    Shift $40/day to brand search

  2. Governed

    Inside your limits on all nine checks

  3. Executed

    Budget changed through the Google Ads API

  4. Verified

    Read back from the account: $140/day

  5. Logged

    In your ledger, one click to undo

Only actions on an approved list, through official APIs. If the read-back doesn’t match, it rolls back and tells you.

Illustration

Prove

Lift, not attribution.

Saw the campaign4.6% converted
Held out3.9% converted

+18%

Incremental lift, trailing 28 days. 95% interval: +9% to +27%.

A randomly held-out group shows what would have happened anyway. Platforms grade their own homework; holdouts don’t.

Example · trailing 28 days

The number we answer to

Revenue per Owner Minute.

Not impressions. Not clicks. How much new revenue each minute of your attention produced. It rises when Now What does more and asks less.

$4,290

Incremental revenue, this month

14 min

Your minutes on growth, this month

$306

Per minute of your time, this month

Example · one product, one month

Autonomy

Autopilot where you want it. Nowhere else.

Autonomy is set per capability, not per account. Let it run your ads while it only recommends pricing and never touches your code. Try it.

  • Paid acquisition

    Budgets, bids and creative within your capital limit

  • Search & AI discovery

    Crawler access, structured data, content gaps

  • Site & content

    Landing pages, copy, metadata

  • Lifecycle

    Onboarding and win-back email

  • Pricing

    Plans, trials, discounts

  • Product & source code

    Pull requests to your repository

2 of 6 acts on its own. 3 proposes, you approve. 1 stays out.

Illustration · nothing here is saved

Autopilot still works inside your capital limit and your guardrails, and every action lands in a ledger you can undo from. Measurement and research run regardless; they only read.

Capital

Three numbers. Never blurred.

What you spend today, what you allow, and what should actually run are different numbers. Now What keeps them apart, and every one carries its period.

Observed

Existing spend

$6,200/mo

What your connected ad accounts actually spent. Read, never typed in.

Set by you

Authorized capital

$10,000/mo

The most Now What may manage. A ceiling, not a target.

Now What’s call

Recommended deployment

$7,400/mo

What it believes should run right now, with the marginal return that justifies it.

Budgets are sized to your revenue, margin and growth rate. When the economics aren’t connected yet, you see “not measured” and what to connect, never a projection built on industry averages.

Example · monthly

Portfolio

Which product deserves the next dollar?

Build five products. Grow the one that’s working. Now What compares marginal return across your whole portfolio and moves capital to where it earns most, inside the limits you set.

Four titanium rings in a row, their filaments lit to different degrees, left to right: moderate, brightest, faint, dark

Fieldnote

Scale

$41,000/mo

Revenue · +3% vs last month

Constraint · Paid search near saturation

Maintain

The next dollar here returns 0.9×. Hold spend, keep testing creative.

Tidewell

Growth

$18,400/mo

Revenue · +22% vs last month

Constraint · Budget-limited, return still rising

Invest more

Marginal return 2.1× on the last $2,000/mo. It gets the next dollar.

Ledgerline

Maintenance

$6,800/mo

Revenue · −4% vs last month

Constraint · Retention flat, acquisition costs climbing

Reduce

Cut paid spend 40% and harvest. Lifecycle email keeps running.

Quarry

Validation

$300/mo

Revenue · Flat for 6 weeks

Constraint · No retention signal yet

Pause

Stop spend until a pricing test shows people stay.

Every product has a stage, and the stage sets the strategy.

  1. 01Validation
  2. 02Launch
  3. 03Growth
  4. 04Scale
  5. 05Maintenance
  6. 06Sunset

Validation gets cheap, fast tests. Scale gets efficiency. Sunset gets harvested. Evidence stays with the product that earned it; what generalizes carries over as a prior, and is labeled as one.

Example portfolio · fictional products · monthly

Search & AI discovery

Be the answer, not a link.

Buyers ask an assistant before they ever search. Now What makes sure the engines can read you, checks whether they recommend you, and fixes what keeps them from it.

ChatGPT, Gemini, Google AI Overviews, Claude, Perplexity, Copilot, Grok

example.com/robots.txt

Illustration
  • User-agent: GPTBotchecking…Allowed
  • User-agent: OAI-SearchBotchecking…Allowed
  • User-agent: ClaudeBotchecking…Allowed
  • User-agent: PerplexityBotchecking…Disallow: /
  • User-agent: Google-Extendedchecking…Allowed
  • User-agent: Applebot-Extendedchecking…Allowed
  • User-agent: Bingbotchecking…Allowed

PerplexityBot is blocked. You’re absent from Perplexity answers for every query.

− Disallow: /+ Allow: /Fix proposed · needs your approval

Site changes are public, so this one waits for you, even on Autopilot for discovery.

  1. 01 · Eligibility

    Can the answer engines read you?

    Your robots.txt, sitemaps and structured data, checked against every AI crawler. One stray Disallow can make you invisible to a whole engine.

  2. 02 · Intent sampling

    Are you in the answer?

    The questions your buyers actually ask, sampled across engines on a schedule. Which ones name you, which name a competitor, and which name nobody.

  3. 03 · Citations

    Which pages earn the mention?

    Pages that get cited are kept fresh; gaps become briefs for new pages. Search Console shows whether the clicks follow.

Trust

Cheap isn’t the same as safe.

Every action is judged on nine dimensions before it runs. A $0 headline change can need your approval while a budget shift runs on its own, because cost is only one of the nine.

Action A

Shift $40/day to brand search

Action B

Rewrite the homepage headline

Financial exposure
$40/day, already in budget (Low risk)
$0 (Low risk)
Cumulative spend
12% of the monthly limit (Low risk)
None (Low risk)
Reversibility
Instant, one call (Low risk)
Reversible, but seen first (Medium risk)
Audience size
Unchanged (Low risk)
Every visitor (High risk)
Public visibility
Inside the ad account (Low risk)
Your homepage (High risk)
Legal & compliance
No new claims (Low risk)
New claims in copy (Medium risk)
Brand impact
None (Low risk)
Your first sentence (High risk)
Product & production
None (Low risk)
Deploys to production (Medium risk)
Data confidence
28 days, holdout-tested (Low risk)
Untested copy (Medium risk)

Runs on Autopilot

Waits for your approval

Illustration

AI may suggest a guardrail. Only you can make one binding.

Every rule says where it came from, so an inference never quietly turns into a restriction.

Inferred by Now What

Suggested

“Write in a plain, direct voice. No exclamation marks.”

Shapes drafts. Never blocks an action until you confirm it.

Binding

Confirmed by you

“Never discount below 20%.”

A hard limit. Actions that would cross it don’t run.

Platform or law

Externally required

“Meta advertising policy: no health claims.”

Sourced and cited. Applies whatever autonomy you grant.

The ledger

Everything it did, how it checked, and a way back.

Illustration · try Undo
  • 09:14

    Moved $40/day from Meta prospecting to Google brand search

    Autopilot · read back $140/day

  • 08:52

    Paused 3 ad sets with cost per purchase above $62

    Autopilot · read back paused

  • Yesterday

    Added FAQ structured data to /pricing

    Approved by you · validated by Google

You built it.
Now What grows it.

Connect your site and the accounts you already use. Now What reads them, tells you what it found, and asks only what it can’t work out.