The marketing department, as a tool you run yourself.
It writes, you decide, it publishes, and every rupee is accounted for. A product with its own users, its own pricing model and its own launch question — and the launch question is still open.
Two kinds of customer in one column of the database.
A workspace is either a business or an agency, and the distinction changes the product rather than just the price.
A business workspace is a person doing their own marketing: guided mode on by default, every screen in plain words, a helper that points while the person presses, and a plan that meters what the platform's keys spend on their behalf. An agency workspace runs other people's brands on its own vendor keys, so nothing of the platform's is metered; it gets staff logins with roles, client portals and higher autonomy. The portal shows each client their own spend, leads and cost per lead, and never the agency's cost base.
- Relationship to the work
- Executed
- Sold as
- Software you run. Having it run for you is a separate arrangement, month to month
- Billing
- Not built. Plans are changed from the machine, never from a screen a customer can reach
- Customers
- None recorded
Five plans, and one place the numbers live.
The plans screen, the refusal messages and the arithmetic behind the caps all read the same row. A price that appears in three places is a price that eventually disagrees with itself.
| Plan | For | ₹ a month | Workspaces | Logins | Portals | Autopilot | Compute |
|---|---|---|---|---|---|---|---|
| Solo | a business doing its own marketing | 1,999 | 1 | 1 | — | level 1 | platform, metered |
| Solo Plus | the same, where short video is the channel | 3,499 | 1 | 2 | — | level 1 | platform, metered |
| Silver | a 2 to 5 person studio | 9,999 | 10 | 5 | 10 | to level 2 | own keys |
| Gold | an 8 to 15 person agency | 19,999 | 50 | 10 | 50 | to level 3 | own keys |
| Platinum | a multi-office agency | 39,999 | 200 | no cap | no cap | to level 4 | own keys |
Every refusal names the plan, the number and the way out. A limit a customer hits without being told which limit it was is indistinguishable from a bug.
Per workspace, the price falls by a factor of five.
An extra workspace costs the plan's own per-workspace rate, and the allowance runs out exactly where buying the next plan outright costs the same in total. So at every cap the next rung is a few rupees away rather than a cliff.
carries export, cancel, delete, own connections and own keys
carries the approval queue, the stop control, the audit log and compliance
does moving down delete, archive or hide anything
The most consequential number has not been set.
Metered, with a rupee ceiling behind the counts
- Counts for drafting turns, images and voiceovers in front; a rupee ceiling behind all of them, so an unusual month cannot turn a fixed-price plan into an unbounded cost.
- Video is capped separately per workspace, because a per-generation ceiling cannot see eighteen reasonable clips in a row.
The agency's meters, not the platform's
- The variable cost leaves the model entirely: the plan price becomes software rent rather than rent plus a pass-through.
- A workspace on its own keys with none on file is refused with the reason, never quietly moved onto the platform's.
Both models are built. The handover records, plainly, that neither has been priced. A costing bug found along the way is worth knowing: prompt-cache tokens are not priced like ordinary input tokens, and costing them at the base rate overstated AI spend roughly tenfold on the calls this platform makes most — an error in the direction of charging too much, which nobody would have noticed, because the margin would have looked fine.
Three ways to sell it, and an argument about fit rather than readiness.
- Before
- Three options, recorded as a genuine fork so the next person does not have to rediscover the trade-offs: self-serve software; an internal tool behind the agency; or a productised service where clients get a scoped login into their own project with view-and-approve rights only.
- What forced it
- The third needs the least new work by a distance — client accounts scoped to one project, no billing, no onboarding, no self-serve key management — and the existing tenancy was built for exactly that shape.
- The decision
- The case against self-serve is not readiness. It is fit: everything that makes this good is invisible in a ten-minute trial and obvious by month three. The disciplines, the craft layer, deterministic budget enforcement, the honest empty states — several of them look like the product doing less than a competitor that would have shown you an invented number.
- What it changed
- That is the wrong shape for a market won in the first session, and the right shape for delivery somebody is already paying for. The same restraint governs autonomy: raising the default level is named as a product decision, not an engineering one, and should follow evidence from real campaigns rather than confidence.
Each rung names a different bottleneck, not a bigger version of the same one.
Silver — writing capacity
A 2 to 5 person studio or a senior freelancer with 6 to 12 retainers. The plan's own description: it replaces the junior who writes first drafts. Naming the role it displaces is a harder claim to make and much easier for a buyer to price against a salary.
Gold — approval throughput
An 8 to 15 person agency with 20 to 50 retainers and an approval queue that has become the bottleneck. Autonomy to level 3 is what that buys: the routine work runs itself.
Platinum — cost visibility
A multi-office agency with a management layer that wants cost per client — which the platform can answer, because the finance module computes one definition of it that every screen reads.
The product can be used. It cannot yet be sold.
- Export, cancel, delete, own connections, own keys, the approval queue, the stop control, the audit log and compliance are promised on every plan
- What a higher rung buys is seats, autonomy, video and a lower price per workspace — not the right to leave
- Self-serve sign-up and billing are not built, and nothing has ever run against a real account
- Per-tenant stop and pause flags: today one flag covers the whole process, so stopping one workspace stops everyone
- A per-tenant autonomy level under the plan ceiling: today it is a single environment value, so the levels the plan table sells are not yet separable per workspace
- Whole-account export and delete account are promised on every plan and are on the to-build list
- Neither compute model has been priced
- No user count, revenue, retention or churn. The three bottlenecks the ladder is built around are positioning nobody outside the studio has confirmed
- The plans table prints the two business plans as tax-inclusive while Indizilla's claims register records that it is not registered and no tax line may appear; the two do not agree and neither is asserted here
Read the full study.
Ten panels, ending on what could not be verified.