Everyone else builds your campaigns. Darwin writes the strategy they come from.
Give Darwin a goal, an offer and a timeline. It drafts the whole
strategy — KPI targets, segments, budget split, channel roles and a phased plan
— you revise it by prompt, 22 agents run it, and the report tracks results
against the plan you approved.
Then it writes the win-back email, finds the budget quietly bleeding out,
and puts it in front of you finished, waiting for one click.
Paste your website. Darwin drafts a
first 90-day plan: three priorities, who does what, and a sample email and ad written in the
business’s own voice — each one waiting for approval, like everything it makes.
Darwin reads your public home page and drafts a first plan in about 20 seconds.
Nothing is posted, sent or saved to an account.
Or see a sample:
Sample plan · fictional home services business · drafted by Darwin
Harbor Line Plumbing
Family-run licensed plumbing company offering emergency repairs, drain cleaning, water heater work, and bathroom remodels. Sells to: Homeowners in a mid-size US city and its suburbs who need reliable, upfront-priced plumbing help.
Three priorities
1
Local search visibility
Plumbing searches are highly local and urgent; being found in the right suburbs matters most for emergency jobs.
MeasureMap pack impressions and website visits from local search
2
Emergency call capture
24/7 emergency claim only pays off if calls after hours actually convert into booked jobs.
MeasurePercentage of after-hours calls that result in a booking
3
Seasonal service bookings
Winter freeze checks and spring sump pump service are named offers that can fill slower periods.
MeasureBookings tied to seasonal campaign links or promo codes
The first 90 days
Weeks 1-4Foundation: local presence and tracking
Draft service-area pages for the city and each suburb for owner review
Set up call and booking tracking so results by channel can be measured
Draft review-request process to start building visible customer feedback
Weeks 5-8Launch targeted campaigns
Launch search ads for emergency repairs and water heater services in served suburbs
Draft and test after-hours call script proposal to capture more emergency bookings
Start seasonal email reminders tied to the current time of year
Weeks 9-12Review and expand
Review which services and areas are generating the most calls and bookings
Propose expanding location pages or ad budget toward best-performing services
Draft next season's reminder campaign ahead of the transition
Who does what
SEO SpecialistDraft local SEO fixes: city/suburb service pages, map listing cleanup, review request flow.
GEO ManagerPropose location pages for each suburb served, targeting emergency and remodel searches.
Ad CreativeDraft search ad variants for emergency repairs, water heaters, and bathroom remodels.
Ad ManagerPropose search campaign structure split by service type and by season.
Email MarketingDraft seasonal reminder emails for pipe freeze checks and sump pump service.
AI ReceptionistPropose call-handling script for after-hours emergency calls and booking capture.
Drafted for your approval
Email · awaiting your approval
Is your home ready for freezing temperatures?
Winter is coming, and frozen pipes can turn into a costly repair fast. At Harbor Line Plumbing, we offer winter pipe freeze checks so small issues get caught before they become emergencies. We're family-run, licensed and insured, with upfront flat-rate pricing, so you know the cost before we start. Book your freeze check online or give us a call. Our vans are on the road serving the city and surrounding suburbs, ready when you need us.
ApproveEditReject
Ad · awaiting your approval
Flat-Rate Plumbing, No Surprises
Family-run, licensed and insured plumbers serving the city and suburbs. Emergency repairs, drain cleaning, water heaters, and remodels. Upfront pricing every time.
Book Online Now
Assumptions Darwin made: Assume mid-size city means service area of roughly 20-30 mile radius. · Assume most customers are residential homeowners, not commercial or property managers. · Assume current online booking and review presence is minimal since not described on site.
An answer from your own numbers, with the plan attached.
Ask Darwin a question and it answers from your CRM and campaign data, says why, and drafts what it would run. It ends where every answer ends: nothing is sent until you approve it.
Ask Darwin, answering which audience to target next. Demo account.The same chat on a phone. Demo account.
The gap
Every AI tool writes campaigns. Almost none writes the plan.
01 · Assets, not direction
A generator
makes an email or an ad set. Who to target, what to spend where and what success means is
still your problem.
02 · Strategy arrives as a document
Then
someone turns it into campaigns by hand, and it is stale within a month.
03 · Results measured against nothing
Dashboards
show what happened, not whether it is what you planned — so nobody knows when to change
course.
The loop
Create, execute, measure — and back again
Darwin executes across every channel, so the loop runs end to end. The
steps that are human by design stay that way: assets wait in approvals before anything goes
out, and SDR outreach SMS and calls run behind consent and do-not-call checks.
1 · Short intake
Objective and target,
brand, the product or offer from your catalog, budget, timeline. One screen, about five
things. No budget in mind? Darwin recommends one and labels it an estimate.
2 · Darwin drafts the strategic layer
KPI
targets, priority segments, budget allocation across channels, the role each channel plays,
and a phased plan with dependencies.
3 · You review and revise by prompt
Edit any
field directly, or just say what to change — “drop the paid social phase and put
the budget into reactivation”. Darwin rebalances the rest of the plan.
4 · Generate, schedule, measure
The agents
produce and schedule the work across every channel, and the report compares actuals to the
targets you approved — by week, by channel.
What Darwin writes
A strategy, not a content calendar
The four things that normally live in a strategist’s head, written down
where the report can track them.
KPI targets
What success means in numbers,
set before a single asset exists.
Segments
Primary, secondary and tertiary
audiences drawn from your own contacts and personas.
Budget allocation
Spend split across channels
with guardrails, so the plan adds up and nothing commits money you did not approve.
Channel roles and phases
Who builds awareness,
who nurtures, who converts — and in what order.
A strategy as a timeline: four phases over 26 weeks, each step marked approved
or waiting. Demo account.
Measured against the plan
Know when you are ahead, behind, or off course.
A dashboard shows what happened. It rarely shows whether what happened is what
you planned, which is why nobody can say when to change course. The strategy report compares
real results from your live campaigns against the targets you approved — projected
versus actual, by week and by channel.
When there is no data yet, it says so. Darwin does not fill a gap
with a plausible number, which is the same rule that governs every other answer it gives you
— and the reason a projection here can be put in front of someone who will check it.
Found money
It found what you didn’t have time to look for
The Monday briefing is ready before you are — Darwin has already
been through the budgets, the lapsed customers and the overdue invoices nobody gets round to.
01
It watches the budgets you stopped watching
Pacing, burn rate, and where the month actually lands — before it lands. Not a
dashboard you have to remember to open.
02
It finds the customers nobody chased
Lapsed buyers, abandoned carts, leads that went cold while everyone was busy. With the
win-back already written.
03
It chases the invoices
Who owes you, how long it has been, and a reminder drafted and ready to send.
Most of this is work nobody had time for. That is the
point — it is not replacing what you do, it is doing the part that kept slipping.
Every AI gives you an answer. Ask two of them something neither can know
— only one tells you. Pick a question.
A typical AI assistant
“Your landing pages are performing well — 1,240 visits last
week with a 3.2% conversion rate, up 0.4% from the prior period.”
Ask Darwin
“You have 6 landing pages — 4 live, 2 not published. One has
a compliance warning. Landing-page visits and conversions aren’t tracked anywhere
in the platform, so I can’t tell you how they performed — only what exists
and its state.”
The left-hand numbers are invented. That field was never
collected.
A typical AI assistant
“Done — I’ve sent the win-back campaign to 2,847
lapsed contacts.”
Ask Darwin
“I’ve drafted it and put it in front of you: 2,847
recipients, the subject line, and the full body. Sending email can’t be undone,
so I can’t send it — you can, from here. Nothing has gone out.”
It can write it. It cannot send it. That switch is hard-wired
off.
A typical AI assistant
“You have $12,400 remaining across your active
campaigns.”
Ask Darwin
“Two budgets are pacing under and one is fully spent. One
budget’s spend hasn’t been fetched yet, so I can’t tell you how
it’s pacing — I won’t report it as under-spending, because
that’s a zero nobody measured.”
A budget with no data isn’t a budget at zero. Most tools
report it as zero.
A typical AI assistant
“Competitor A leads with 34% share of voice, followed by B at
28%.”
Ask Darwin
“I can show you what your tracked competitors’ reviews say
and how your own visibility is trending. I don’t have their ad spend or share of
voice — that data isn’t in your account, and I’m not going to
estimate it.”
Share of voice is the most confidently invented metric in marketing
software.
This is why you can put its numbers in a board
deck. Whole categories of feature in this product are switched off at the model layer for
exactly this reason — they worked, and what they returned was invented. We found them in
our own product and took them away from the AI. See them →
The left-hand column
illustrates a common failure mode in AI assistants generally. It is not the output of any
named product.
Switched off on purpose
The features we refuse to ship
These are not among the 174 tools our 22 agents reach. They all
worked — that was the problem. Each one returned a confident, plausible answer that
nothing in your account supports, so it was taken away from the model rather than patched.
Below is the category, and what it used to invent.
Inbox search
We’re not connected to
your mail. Anything it listed would be invented, including who it was from.
Knowledge-base browsing
There’s no
knowledge base behind it. It used to return three articles that don’t exist.
Report export
No file is produced. It used
to return a 2.4 MB file size and a download link to nothing.
Webhook listing
There’s no webhook
store. It returned a hard-coded list.
Tasks board
The board doesn’t exist.
An empty result from it didn’t mean you had no tasks.
Creative performance prediction
Nothing
predicts it. It returned the same CTR every time.
Competitor keyword gaps
Competitor keywords
aren’t stored. It made them up.
Social listening / trending topics
No
listening pipeline. The “trends” were a fixed list.
Every AI product has these. Most ship them. We found
ours, counted them, and took them away from the model — because a confident wrong answer
is worse than no answer, and much harder to catch.
How the work lands
It proposes. You approve. Then it checks again.
1
It drafts
Reads your account, writes the thing, and shows you exactly what goes out —
recipients, subject, body, spend.
2
You approve
Nothing sends, publishes or spends without a person. Every tool that can send, publish
or spend sits behind the approvals queue, and the AI cannot reach past it.
3
It re-checks
If anything changed between your approval and the moment it runs — a budget moved,
a list grew — it stops and asks to be re-approved. If it can’t prove what you
approved, it refuses to run at all.
Step 3 is the one that matters. An approval is given
for a specific set of facts. If the facts move, the approval isn’t valid any more.
The approvals queue: the irreversible item first, with the exact email that
would go out. Demo account.
●
Every action is on the record
Every request, plan, agent execution, data change and approval is recorded with who and
when — retained and exportable. If procurement asks what happened and who signed
it off, the answer is a file, not a recollection.
■
One kill switch
Stop every agent across the organisation immediately. In-flight actions roll back.
One click, no ticket, nothing to uninstall or reconfigure afterwards.
▾
Autonomy starts at zero
The fear was never that AI writes a bad email — it’s that it sends one to
40,000 people. Every action is gated by default; you raise the dial per agent and per
action, only where you want to.
What it reads
22 specialists. 174 tools. All reading your account.
Ask about a campaign and the Campaign Manager answers. Ask about receivables
and Invoicing does. Ask something that spans everything and the Executive Assistant takes it,
with every capability available to it.
Not a chatbot with a search box bolted on — each question is routed to
the specialist that holds the data.
Autonomy is a setting, not a plan tier. Set it per campaign, per module or across the whole workspace, and change it whenever you like.
Manual
You do it
Every module keeps a full manual path. Build the campaign yourself and use the AI for nothing more than a second opinion.
Nothing runs unless you run it
Supervised
It drafts, you approve
The default. Agents research, write and schedule, then everything waits in your approvals queue until someone signs off.
Default on every plan
Autonomous
It runs inside your guardrails
Hand over end-to-end execution within the limits you set — budget caps, send windows, channels, spend. Stop it in one click at any time.
You set the limits first
The credibility close
We counted our own AI honestly
Most vendors quote an agent count. We built 153 agent
classes. We shipped 22 — the ones that could actually read your account.
The other 131 couldn’t, so putting them in a brochure would have
meant inventing a number, which is the one thing this product doesn’t do.
Ask any vendor quoting a big number how many of theirs reach your data. They
will not know. We went and counted.
How is this different from an AI campaign builder?
A builder makes one campaign from a prompt. Darwin decides targets, audiences, spend and channel mix first, then every campaign comes from that plan — and the report measures against it.
Which channels can it actually run?
All of them — email, sequences, social, ads, content, outreach and calls. Assets wait in approvals before anything goes out, and SDR outreach SMS and calls run behind consent and do-not-call checks.
So it can’t actually do anything?
It does the work — writes the campaign, finds the overspend, drafts the win-back. What it can’t do is send it without you. That’s deliberate: everything it touches costs money or reaches a customer.
How do I know it isn’t making things up?
Because it tells you when it doesn’t know. Every figure comes out of your account, and when a number isn’t there it says so. Whole categories of feature that returned plausible invented answers are switched off at the model layer rather than patched.
We already have AI in our stack.
Ask it something it can’t know, then ask this one. That’s the difference, and it takes ten seconds.
Isn’t this just a chatbot on our data?
A chatbot searches. This routes your question to the specialist that owns that data — 22 specialists reaching 174 tools. Ask about receivables and Invoicing answers, not a search box.
What can it not tell me?
Landing-page and web analytics aren’t tracked in the platform, so it won’t report them. It isn’t connected to your mail, so it can’t search your inbox. It says so rather than estimating.
Can it run on its own?
No. It proposes and a person approves — and if the facts change between approval and execution, it stops and asks again.
Get started
Ask Darwin what needs you today
It reads your account before you get in. Then it drafts the work and waits for your
click.
