AI sales agent + multi-tenant CRM
LeadFoundryYour sales team gets a 7 AM shift report.
A multi-tenant SaaS platform where an AI agent runs a manufacturer’s entire lead pipeline every day - capturing enquiries from IndiaMART, WhatsApp, email, the website and referrals, replying instantly, chasing quotations to a checkpoint, and delivering a shift report every morning that says exactly whom to call. Configured once per company, then it runs itself.
4 channels in · one pipeline out · runs itself daily
Shift report · 07:00 ✓✓
0 calls today · 0 quotes due · 0 escalated
- Live at
- leads.theairbox.in ↗
- Region
- B2B manufacturing
- Timeline
- 2026
- Platforms
- Web · Mobile web · WhatsApp · Email
Mobile
The same product, built for a phone - not shrunk onto one.
Built mobile-first, with the same design system and data layer as the web product.
- Web
- Mobile web
The problem
What made it hard
Small manufacturers lose deals to silence, not to competitors. Enquiries arrive across five channels, nobody owns the follow-up, and quotations go cold at day seven. The product had to work for owners who will not learn a CRM - which meant the software had to do the chasing itself and report in a format someone reads on a phone before reaching the factory.
Our role
- Product architecture
- Agent design
- Multi-tenant SaaS
- Integrations
- Billing and onboarding
The approach
How we solved it
- 01
A daily agent run on a fixed schedule: scan the pipeline, acknowledge overnight leads, surface follow-ups due, escalate anything untouched for 48 hours, and send the digest over WhatsApp and email before the shift starts.
- 02
Inbound capture through webhook integrations with IndiaMART, the WhatsApp Meta Cloud API and email, deduplicated and classified on arrival so one enquiry does not become three leads.
- 03
Outbound lead discovery via SerpAPI, so the same pipeline holds both the enquiries that came in and the prospects the company went looking for.
- 04
Instant acknowledgement replies drafted in the company’s own voice, with a human able to edit before send - speed without surrendering tone.
- 05
Quotation chasing on a checkpoint model rather than a reminder list: the agent knows which quotes are at day seven and raises them, along with spares due on installed machines.
- 06
True multi-tenancy - self-serve tenant signup, per-tenant CRM with its own catalogue, quotations and installed base, a superadmin dashboard across tenants, and Razorpay billing.
Architecture
What it runs on
- Application
- Next.js 14, server actions and route handlers
- Data
- PostgreSQL via Drizzle ORM, per-tenant isolation
- Agent
- Anthropic API, scheduled daily runs
- Inbound
- IndiaMART, WhatsApp Cloud API, email webhooks
- Outbound
- SerpAPI lead discovery
- Billing
- Razorpay subscriptions, AWS S3 for documents
- Next.js 14
- Drizzle ORM
- PostgreSQL
- Anthropic API
- Razorpay
- AWS S3
- SerpAPI
- WhatsApp Cloud API
Outcome
What changed
7 AM
digest before the shift starts
5 channels
captured into one pipeline
Multi-tenant
self-serve signup and superadmin
48h
escalation on any untouched lead
Want this level of detail for your own product? The architecture document comes out of discovery - and it is yours whether or not you build with us.
Book a consultationNext case study
GAMERGRAM