I built a 37-feature SaaS in 6 weeks as a non-developer. £4/month infrastructure. 0 customers. Here's everything.
Hey IH 👋
I'm Alok. 20+ years in enterprise IT (Oracle ERP, Workday, PMP certified). Never built a product before. 6 weeks ago I started building CLOBSOW — an AI-powered agency management platform.
Here's the raw, unfiltered story.
What it does:
CLOBSOW connects the workflow that every agency does manually: SOW → Project → Timesheets → Invoice → Revenue Forecast.
The killer feature: AI writes a complete Statement of Work from a one-line brief in 15 seconds. Scope, milestones, pricing breakdown, acceptance criteria — all generated.
The stack:
- Frontend: React (Vite)
- Backend: Firebase (Auth, Firestore, Hosting)
- AI: Anthropic Claude API via Firebase Cloud Functions
- DNS/SSL: Cloudflare
- Email: EmailJS
- Domain: clobsowagency.co.uk
- Total cost: £4/month
What I shipped (37 features):
- 4 AI agents (SOW Writer, Data Chat, Auto-Invoice, SOW PDF Parser)
- Client management (49 countries)
- Project management (T&M + Fixed Price)
- Weekly timesheets with approval workflow
- Smart invoicing (3 types) with global tax engine
- Revenue forecasting (3-12 months) — no competitor offers this
- Multi-currency (40+), multi-language (EN/FR/DE/ES)
- 5 colour themes, mobile responsive
- Rate card that AI uses automatically
- Change request automation
- Team management (5 roles)
The numbers (honest):
- Revenue: £0
- Customers: 0
- LinkedIn impressions: 5,700+
- Members reached: 1,539
- Demo video views: ~50
- Calls with prospects: 1 (she wanted to sell to me, not buy)
- API key stolen twice (cost me ~$15)
- Time to build: 6 weeks evenings/weekends
- Lines of code: ~8,000
- Coffee consumed: too much
What worked:
1. Build in public on LinkedIn — Google Gemini now knows my product from my posts alone. Zero SEO.
2. Commenting on agency owners' posts — 40% conversion to connections. Better than cold DMs.
3. Claude AI as my engineering partner — I'm a PM, not a developer. AI made this possible.
4. Flat pricing (£75/month unlimited users) vs competitors charging per-user (£22/user = £440 for 20 people)
What didn't work:
1. Cold DMs without research — got told off by a prospect. Lesson learned.
2. LinkedIn group posts — 60 impressions vs 500+ on personal feed.
3. Trying to pitch everyone — solo freelancers, marketing agencies, crypto founders. None were my customer.
4. Building features instead of talking to people — the product was ready at week 2. Weeks 3-6 were learning that building ≠ selling.
What I learned the hard way:
1. Nobody signs up from a LinkedIn post. They sign up after a conversation.
2. Your API key WILL get stolen if it's in browser code. Move it server-side immediately.
3. The name doesn't matter at 0 customers. The product does.
4. A partnership call where they try to sell you their service is still valuable practice.
5. 1,539 people knowing your product exists ≠ 1 customer.
ICP (took 4 weeks to figure out):
5-50 person IT consultancies and digital agencies in the UK who bill hourly, use Word for SOWs, and forecast in spreadsheets.
Competition:
- Scoro: £22/user, no AI, no forecasting
- Monday.com: No SOW management, no invoicing
- Productive: Closest competitor but no AI, per-user pricing
- Kantata: Enterprise, $1000+/mo — too big
- Harvest: Time tracking only
Month 2 goals:
- Get first paying customer
- 2 demo calls booked (1 confirmed)
- Post MP4 demos weekly on LinkedIn
- 5 comments/day on target audience posts
- Stop building features. Start selling.
Ask for the IH community:
1. Anyone here run an agency and manage SOWs? Would love your feedback.
2. What's your best channel for first B2B SaaS customers?
3. How long did it take you to get from 0 to 1?
📺 Demo: youtube.com/watch?v=JfJhNcz2A7s
Happy to answer any questions about the build, the stack, or the struggle.
— Alok
I run a small B2B services company (staff augmentation, so SOWs and MSAs are the whole business), so answering your three questions from that side of the table.
On SOWs: the painful part for us was never writing the SOW. It is the drift you and Aryan already named. The rate card changes, a scope tweak gets agreed on a call, and three months later the invoice and the SOW disagree. If your tool makes the change request the ONLY way a number can change, that is the feature I would lead with, not the 15 second AI draft. Agency owners have seen "AI writes your proposal" a hundred times this year. "Your invoice can never disagree with what the client signed" is rarer and it speaks to a real scar.
On first B2B customers: what finally worked for us was narrow, signal based outbound, not volume. Pick a trigger that means the pain is live right now: an agency that just posted a job for a PM or ops person, one that just announced a new contract, a founder publicly complaining about invoicing. Ten of those a week with a genuinely specific first line beat 200 generic DMs, which is the same lesson you already paid for. Also, your ICP sentence (5 to 50 person UK consultancies billing hourly, Word SOWs, spreadsheet forecasts) is good. Put that exact sentence in the first line of every message. People reply when they feel seen.
On zero to one: it took us about four months and the first customer came from a warm intro, not from a channel. The channel work is what earned the intro. Ask the one prospect you did talk to, even the one who tried to sell to you, who else she knows running this kind of shop. That question is usually the fastest path there is.
One more thing on your month 2 plan. Five comments a day on target posts is the right instinct. Make half of them questions instead of insights. Questions start the conversation, and as you said yourself, the conversation is what converts.
This is exactly the kind of feedback I needed — from someone who lives inside SOWs and MSAs daily."Your invoice can never disagree with what the client signed" — that reframe hits harder than anything I've been leading with. You're right that "AI writes your SOW" has become noise this year. The change control lock is the real differentiator and I've been burying it behind the AI headline. Changing that.
The signal-based outbound is smart. I've been doing volume comments when I should be finding the 10 agencies where the pain is live this week.
Job posts for PMs and ops roles is a great trigger
I hadn't thought of.
And the questions point — noted. I've been
commenting insights when I should be starting
conversations. Flipping half to questions
starting this week.
Thank you Roberto for taking the time to write this.
This is worth more than 1,000 impressions.
The biggest shift I noticed wasn't the AI—it was treating the SOW as the source of truth instead of just the first document in the process.
A lot of operational problems aren't caused by missing features. They happen because every downstream system gradually drifts away from the original commercial agreement.
That's exactly the insight that led to the entire architecture. Most tools treat the SOW as a PDF you file away. In CLOBSOW, the SOW is the live source of truth — rates flow into timesheets, milestones flow into invoicing, and any change request auto-updates the SOW value, project budget, and revenue forecast simultaneously. No drift. No re-keying. No 'that's not what we agreed.' The moment I connected the SOW to everything downstream, every other feature became obvious. Appreciate you articulating it so clearly — 'gradual drift from the original commercial agreement' is exactly the problem.
Glad it resonated.
Your reply made me think there's one strategic decision sitting underneath that architecture which becomes much more significant as CLOBSOW grows, but I don't think I can explain the reasoning properly in a thread without oversimplifying it.
If you're interested, what's the best email to reach you on?
Thanks! I’ve just sent it over.
Looking forward to hearing your thoughts whenever you have a chance.