Architecture without artifacts.
The previous page told you what the four engines do. This one shows you what they're made of — the layers underneath, the components inside, the connections outward. Without showing you the workflow files themselves, because those belong to the clients we built them for.
Andar dekhna chahte hain? Theek hai. Magar puri kitab nahi mile gi.
Three layers. One reason most automation fails to feel like anything at all.
Most marketing automation you've seen is one layer thick — it executes. Posts go out, ads run, emails send. That's it. Underneath, nothing watches what worked. Above, nothing thinks about what to try next. The system you're paying for is doing exactly the thing you could have hired a junior for, just faster.
The brainlitt system has three layers. Each one watches and acts on the layer below it. Together they make the difference between automation that runs and automation that runs your business.
THE HANDS
The layer that does the work.
Sixty posts a month, six platforms, every campaign you have running, every WhatsApp conversation, every approval workflow, every report. The execution layer is what most agencies sell as 'automation.' It's necessary but it's not the point. By itself, it's a faster intern.
THE EYES
The layer that watches what the hands did, and intervenes.
Every two hours, the optimization layer reads what every campaign just produced and decides what to do about it. It pauses ad sets that just hit fatigue. It moves budget from underperformers to winners while you're asleep. It triggers fresh creative briefs when frequency climbs above three. The eyes don't watch — they reach in and adjust the hands. By the time you read your morning briefing, twenty-eight decisions were already made on your behalf.
THE BRAIN
The layer that thinks weekly about what to do next.
Once a week, the intelligence layer reads the entire seven-day record — every campaign, every conversation, every conversion, every drop-off — and writes you a hundred-and-fifty-word summary of what's working, what isn't, and the single most promising thing to try next. Not a dashboard. A recommendation. By Monday morning, you know exactly what the system thinks you should change.
Most automation stops at the first layer. We think a system isn't really running your business until all three are.
Six capabilities. Each one running on its own clock.
The three layers don't act on themselves — they act through six discrete capabilities, each running on a different cadence. Some run every five minutes. Some run once a week. Together they cover the entire surface of your marketing operation.
01 / Always-On Monitoring
Always-On Monitoring
every two hours, around the clock.
Every active campaign, every channel, every conversation thread, every metric that matters. The system is reading what your business is doing right now while you read this sentence.
02 / Adaptive Targeting
Adaptive Targeting
continuously, learning per audience.
Audience selection tightens with every conversion. Lookalikes refresh as your customer base shifts. Cold audiences generated from public competitor and intent signals.
03 / Continuous Campaign Generation
Continuous Campaign Generation
daily, across platforms.
Sixty posts, twelve reels, multi-platform — produced and scheduled at the times your audience is actually online. Brand voice held consistent without anyone manually checking.
04 / Automated Creative Production
Automated Creative Production
on-trigger, before fatigue sets in.
Image and video creative generated through the pipeline when frequency climbs or performance drops. A new ad in the system within an hour of the old one being killed.
05 / Live Budget Optimization
Live Budget Optimization
on-loop, autonomous.
Spend flows toward what's working, away from what isn't, without anyone clicking a button. The decisions get logged. You'll see them in Sunday's report.
06 / Always-On Reporting
Always-On Reporting
daily briefings, weekly intelligence.
A morning WhatsApp telling you what happened overnight. A Sunday email telling you what happened this week and what to try next. No dashboards to log into unless you want to.
The system, as a graph.
This is the system, drawn as a connection map rather than a feature list. Lines are real — every arrow represents a component talking to another component. Boxes are abstract — each one stands for a function, not a specific workflow file.
The system, drawn flat. Lines are real. Boxes are abstract.
— Solid lines: scheduled triggers (cron-driven)
— Dashed lines: event triggers (webhook or message-driven)
— Hollow boxes: external services we connect to
— Filled boxes: brainlitt-built components
What we plug into. And why.
The architecture above is what we built. Below is what we built it to talk to — six categories of external services, APIs, and platforms the system connects with on your behalf. Most are ones you already use. None require you to learn a new tool.
ORCHESTRATION
n8n
WhySelf-hostable workflow orchestration with a real visual editor and proper execution history. We needed code-grade reliability without writing every connector from scratch — n8n is the rare tool that delivers both.
LANGUAGE & REASONING
Anthropic Claude
WhyWe use Claude Sonnet across the system for content generation, decision logic, and conversation handling. We chose it for tone consistency, instruction adherence, and the safety profile we need when AI is replying to customers in your name.
ADVERTISING
Meta Graph API · TikTok Marketing API · Google Ads API
WhyThe Meta Graph API gives us campaign-level control with two-hour granularity — pause, scale, refresh creative, pull insights. The other ad platforms expose similar surfaces; we built integrations for them when client demand earned them.
CUSTOMER COMMUNICATION
WhatsApp Business API · Gmail · Slack
WhyWhatsApp Business is Pakistan's actual customer-service channel. Email and Slack handle internal communication; WhatsApp is where the customer relationship lives. Building the Sales engine on top of it was non-negotiable.
ANALYTICS & PRODUCTIVITY
Looker Studio · Google Workspace (Sheets, Drive, Calendar)
CREATIVE PRODUCTION
Kling · Nano Banana · Sora 2
Each of these is named because we use it in production. We don't list tools we have credentials for but don't run.
Three things land in your hands every week. Everything else stays in the engine room.
The system runs in the background. These are the three artifacts you'll actually interact with — the outputs that translate "things ran well this week" into something you can read, share, and make decisions from.
01 / The Morning Briefing— Mon-Sat, 9 AM, on WhatsApp.
Good morning.
Yesterday's numbers, the one decision we made on your behalf, and one thing worth your attention.
YESTERDAY
WE ACTED
Paused two ad sets in the Karachi audience that hit fatigue overnight (frequency 4.2). Redirected their PKR 1,800 daily budget to the Lahore lookalike, which converted at 1.6× the Karachi rate this week.
Generated three fresh creative briefs for the paused ad sets. Drafts will be in your review folder by 11 AM.
WORTH KNOWING
A regional dental network launched a discount campaign yesterday. We're seeing a 12% bid inflation in the dental keyword cluster as a result. Holding steady — no action needed unless this persists past Friday.
— brainlitt morning briefing
One paragraph. Three numbers. Sent before your first cup of chai.
02 / The Intelligence Dashboard— Live, always-on, available when you want it.
WEEK 12 · 16-22 MARCH 2026
How the week ran.
TOTAL SPEND
PKR 84,200
DAILY SPEND
LEADS BY SOURCE
COST PER LEAD
PKR 234
↓ 18% vs last week
CONVERSION RATE
3.2%
↑ 0.4 pp vs last week
ACTIVE CAMPAIGNS
8
2 paused, 1 launched this week
WORTH NOTING
The Karachi audience hit creative fatigue mid-week. We launched three new variants on Thursday — early signals are positive.
The same numbers your team sees. Updated every thirty minutes.
03 / The Sunday Report— Once a week, in your inbox.
BRAINLITT — WEEKLY INTELLIGENCE
What this week told us.
Week of 16 March 2026.
Spend up 4% week-over-week. Cost per lead down 18% — the largest single-week improvement of the quarter. The Karachi audience that plateaued in February has started converting again, driven by the creative refresh we launched on Tuesday. One concern worth flagging in the recommendation below.
The Tuesday refresh worked. Doing it monthly, not quarterly, is now standard practice for your account.
THREE OBSERVATIONS
The creative refresh outperformed the prior batch by a margin we haven't seen in eight weeks. The hooks that worked emphasized affordability over aspiration — worth replicating.
The regional dental network's discount campaign continues into next week. Their bid inflation is now affecting our dental keyword cluster. We've prepared a counter-positioning brief for your review.
Sunday spend was the week's highest at PKR 14,000 — but also produced the highest CPL. The audience is there on Sunday but the conversion behavior is different. Recommend testing a Sunday-specific creative variant next week.
ONE RECOMMENDATION
Move the next creative refresh from end-of-month to mid-month.
Based on this week's signals, fatigue is now setting in around day 10 of a creative's life rather than day 14-15. Refreshing mid-month gives us tighter cycles and keeps the audience seeing novel content before performance dips. We'll prepare the next batch for the 28th.
A hundred-and-fifty words. One recommendation. Read it in two minutes.
— A note on the choices above
Every tool on this page was chosen against an alternative we rejected. n8n over a custom orchestrator we briefly built and abandoned. Claude over the API we tried first. WhatsApp Business over the gateway services that promise easier setup and deliver brittle delivery. The system you'd be paying for is the system that survived our own arguments about it. That's the only credential that means much when you're choosing whose code is going to run your business.
You've seen the system. Now you're either a buyer or you're not.
If you're a buyer, the next conversation is short — what your business does, what you'd want this system to handle, where to start. If you're not, that's fine too; the page existed mostly to be honest with the people who were going to ask anyway.