How we built a multi-tenant AI proposal engine that takes agencies from draft to e-signed close
A B2B proposal and client-acquisition SaaS for agencies — AI-assisted proposal building, engagement analytics, and e-sign close — on a multi-tenant data model with region-aware pricing in a monorepo.
Visit the live productThe problem
Agencies win or lose business on the proposal, but most build them in static documents that go out and then go dark — no signal about whether the client opened it, lingered on pricing, or ghosted. The closing motion is disconnected from the artifact that does the selling.
Writing those proposals is also slow and repetitive. Agencies re-draft the same scope, deliverables and pricing structure for every prospect, which is exactly the kind of work an AI assistant should accelerate without flattening the agency’s voice.
And as a SaaS serving many agencies, the platform needs hard tenant isolation — one agency must never see another’s clients, proposals or analytics — while still serving a global customer base that pays in different currencies.
The architecture
- A Next.js monorepo (apps/web) on Firebase for app, auth and data, structured around an agency multi-tenant model so each agency’s clients, proposals and analytics are isolated.
- An AI-assisted proposal builder drafts and structures proposals, which then flow through an engagement-analytics layer that tracks client interaction.
- E-sign and close move the proposal from a document into a signed agreement inside the same flow, so the closing step lives where the selling happens.
- Razorpay handles both one-time orders and subscriptions, with region-aware INR/USD pricing selected per customer.
- The monorepo layout keeps shared logic and the web app together, simplifying the path from draft to engagement to signature.
The AI stack
Engineering challenges
Tenant isolation as a first principle
A proposal SaaS is only trustworthy if agencies are certain their pipeline is private. We modeled multi-tenancy into the data layer from the start so clients, proposals and analytics are scoped per agency — isolation is a structural property of the schema, not a filter bolted on at query time.
AI drafting that keeps the agency’s voice
The point of AI here is to remove the blank-page tax on repetitive scope and pricing sections, not to homogenize every agency into the same template. The builder assists drafting while leaving the agency in control of the final pitch.
Closing the loop with engagement signal
A static proposal gives zero feedback. By attaching engagement analytics, the agency learns whether the client engaged and where, turning follow-up from guesswork into a timed, informed nudge — and the e-sign step keeps the close inside the same artifact.
Selling globally from one checkout
Agencies and their billing live in different regions, so pricing is region-aware across INR and USD, and Razorpay supports both one-time orders and recurring subscriptions — one billing layer covering both the trial-purchase and the ongoing-subscription motion.
The result
- Proposals move from AI-assisted draft to e-signed close inside one connected flow.
- Engagement analytics replace "did they even open it?" with concrete follow-up signal.
- Per-agency multi-tenant isolation keeps each customer’s pipeline private by design.
- Region-aware INR/USD pricing and dual one-time + subscription billing serve a global base.
Lessons we'd bring to your build
- Bake multi-tenancy into the data model from day one — retrofitting isolation onto a single-tenant schema is far riskier than designing for it.
- Use AI to kill the repetitive 80% of B2B content while keeping the human in control of voice and the final 20%.
- Instrument the artifact that does the selling; engagement signal turns follow-up from guessing into timing.
- For global B2B, treat currency and billing mode (one-time vs subscription) as configuration, not as separate code paths.
Planning something like this?
We bring the same production discipline to client builds. Tell us the problem you're solving and we'll map an approach.
Book a discovery call