The platform · part of Automation and AI
Your own AI sales platform.
The AI sales platform is one internal app on top of your CRM. It copies every system your revenue team uses into a database you own, gives each deal and each account one screen, and runs AI through a layer that puts a cost on every call and waits for a person to approve anything that gets sent. We build it in your accounts, with your engineer, and hand it over. There are no per-seat fees, because it is yours.
It is the largest thing we build. Most teams should start smaller, and we say so on the call when that is true.
What it replaces
Six tabs before every call. Then one.
A rep getting ready for a call opens the CRM, the call recorder, billing, the help desk, the product dashboard and a spreadsheet. The platform copies all of them into one place on a schedule, so the screen a rep opens is already complete.
Your own sales platform on top of HubSpot, built with your engineer so you own it. No per-seat pricing, no vendor deciding your roadmap.
- CRMAccounts, contacts and deals
- CallsRecordings and notes
- BillingInvoices and what was paid
- SupportTickets and replies
- ProductWho uses what, and how much
- ContractsTerms and renewal dates
What your team sees
Four screens, drawn with invented data.
These are drawings of the kind of screen we build, filled in with a made-up company. Yours are designed around how your team works and built on your real records.
- Their contract with the current provider ends 31 March. From the call on 12 February.
- One objection is still open: the link to their dispatch software. Raised on 12 February, not answered yet.
- Their finance lead has not been on a call. Three of the four people who can say no have.
| Who can say no | Met |
|---|---|
| Operations director | Yes |
| Dispatch manager | Yes |
| Procurement | Yes |
| Finance lead | Not yet |
Read from the CRM, calls, billing, support and product usage. Last copied 6 minutes ago.
- Everything about the deal in one place, read from your own copy of the data
- The brief says where each line came from
- The eight moves of the method are on the record, filled or empty
| Account | A year | Renews | How sure is the date | Risk | Next step |
|---|---|---|---|---|---|
| Example Carrier Co | $48,000 | 31 March | Signed contract on file | Low | Renewal call booked |
| Sample Freight Ltd | $36,000 | 15 April | From an email, not confirmed | High | Save plan, owner named |
| Demo Distribution Inc | $62,000 | 2 May | Signed contract on file | Medium | Usage review this week |
| Placeholder Logistics | $21,000 | 20 May | Guessed from the start date | Medium | Confirm the date first |
Ranked by how much is at risk and how sure the date is, so leadership can see how much of the book is fact and how much is guess.
- Every renewal with how sure the date is
- Risk next to the money it puts at risk
- A named next step on every account
| Job | Runs | Cost a run | This month | Quality check | Switch |
|---|---|---|---|---|---|
| Brief before each call | 410 | $0.29 | $118.90 | 96% passed review | On |
| Call notes into the CRM | 1,850 | $0.02 | $37.00 | 94% passed review | On |
| Contract terms read out | 120 | $0.21 | $25.20 | 91% passed review | On |
| Lead score, first version | 0 | n/a | $0.00 | Failed its back-test | Off |
Every job has a cost for each run, a check on a sample of its work, a budget and an off switch. One that cannot beat doing it by hand gets switched off.
- A cost for each run, straight off the ledger
- A check on a sample of the work, every week
- An off switch on every job
| System | What is held | Last copied | State |
|---|---|---|---|
| CRM | 48,210 records | 6 minutes ago | In step |
| Billing | 3,940 invoices | 1 hour ago | In step |
| Support | 12,600 tickets | 20 minutes ago | In step |
| Calls | 5,120 recordings | 2 hours ago | In step |
| Product | 310 accounts | 1 day ago | Late. Retry queued |
When a copy is late the page still opens, and says when it was last copied.
- When each system was last copied
- The CRM checked against billing
- Duplicates counted, not guessed
What is in it
Six parts. You own every one.
Blueprint
An architecture for your platform: data mirror, API, workspace, AI router, integrations. Sized to your team, not to ours.
Revenue data layer
CRM, billing, support, contracts, calls, product usage, mirrored into Postgres on a schedule you control. One source per number.
Deal and account screens
One screen that replaces the six tabs a rep opens before a call. Linked straight from the HubSpot record.
AI layer
A central router with a key vault, per-user budgets, a cost ledger, versioned prompts, evals, and a page that shows every prompt in use.
Revenue tooling
ROI and pricing calculators, renewal and churn risk, pipeline analytics, task hygiene, partner attribution.
Build with you, then hand off
We build alongside your engineer or admin, leave a design doc and session notes, and stay on monthly support while you take over.
How it is built
Boring on purpose.
Four layers, one of everything. Two rules matter most: no outside system sits between a rep and a page, and exactly one layer holds the key to your data.
- The browser. The screens your team uses, behind your own sign-in.
- The edge. Checks who is asking and attaches the server-side key, so no credential ever reaches a browser.
- The server. Your rules, your schedules, and the AI router with its budgets and its cost ledger. The only layer that holds the privileged key to the database.
- The database and the mirrors. Every outside system is copied in on a schedule you control. A page reads the copy, never the vendor, so nothing waits on someone else's software.
What your team needs to own it
One engineer, or a technical RevOps person, who is comfortable with TypeScript and SQL. The parts are ordinary: React, Node, Postgres, edge functions and standard model APIs. We build together from the first week, so the handover is a formality.
Hosting and AI model spend are billed to you at cost. We do not mark them up.
What comes in
- Every system you runCRM, billing, support, product, calls
- Your engineerOr ours, until you have one
- How reps workThe six tools open on a Thursday afternoon
What we build
- BlueprintArchitecture and phased roadmap
- Data mirrorPostgres, one source for every number
- Auth, API, AI layerOne credential, metered model calls
- Rep workspaceOne screen per deal
What you get
- One view of every accountAcross four departments
- No per-seat feesHosting and model spend at cost
- You own itCode, docs, credentials, roadmap
How you get there
Four steps. Two of them are sizes you can buy.
The blueprint is a document you keep, whoever does the building. The foundation ends with something your reps use.
- 1Discovery and blueprintFour weeks. Map the systems, talk to every team, write the architecture and a phased roadmap with effort on each piece.
- 2FoundationData mirror, auth, API, first workspace. Reps get something useful inside eight weeks.
- 3AI and analyticsRouter and cost meter first, then the first AI features behind gates, then the analytics screens.
- 4TransferYour team ships the last features with us reviewing. We hand over the design doc and step back to monthly support.
Platform blueprint
4 weeksone fixed quote, in writing
You are thinking about your own go-to-market (GTM) platform, the system your sales and service teams work in, and want the plan on paper before anyone builds.
- The architecture for your own GTM platform: which systems feed it, where the data lives and how the parts connect
- A phased roadmap: what gets built first, what follows, and what each phase gives your team
- Written so an owner can follow it and a technology leader can check it
- A document you keep, whoever does the building
Platform foundation
8 weeksone fixed quote, in writing
You have decided to own the platform and want the base built properly, with something your reps use at the end of it.
- Data mirror: your own copy of the data from each system, such as your CRM (customer relationship management system, where accounts and deals are recorded)
- Auth: sign-in, so each person sees what they should
- API: the interface other software uses to read and write the platform's data
- The first workspace your reps use
Both are sizes of Automation and AI, one of our five services. Each is quoted as one fixed number, in writing, after the free 30-minute audit.
Where it comes from
We have built one, and run it.
Our co-founder Neel designed, built and ran the reference platform as Revenue Operations Manager at a logistics software company. It started as an ROI calculator for the sales team. Fifteen months later sales, account management, marketing and finance worked in it every day.
- database tables
- 173
- routes
- 128
- versioned prompts
- 94
- systems connected
- 12+
- AI calls a year
- 157K+
- departments on it
- 4
- No third-party API on any read path. Nothing waits on a vendor when a page loads.
- One privileged credential, in one place. Writes go through a field allowlist.
- Renewal dates ranked by how sure we are of them, so leadership can see how much of the book is fact and how much is guess.
That platform stays with the company it was built for. What we bring to yours is the design, the order of work, and the mistakes already made once. Read the case study.
An honest answer
When not to build one.
Owning a platform is not always the right call. When it is not, we say so on the audit call.
Buy a product instead when
Your team is small, your process fits a product as it comes, and nobody on your side can own code. At small seat counts buying is usually cheaper, and we will say so on the call.
Start smaller when
The pain is one job, such as call notes or a brief before each call. One AI workflow fixes that in weeks, and it is the right first step for most teams.
Build your own when
Your data sits in several systems, your reps work across all of them, and you have an engineer or a technical RevOps person who can own it.
Put your own numbers into the build or buy calculator.
Questions about the platform
What is an AI GTM platform?
An internal app on top of your CRM. It mirrors every revenue system into one database, gives sales, account management and marketing one workspace per deal and account, and runs AI features like briefs, scoring and renewal risk through a metered, checked layer. It replaces a pile of point tools with one system you own.
Why build in-house instead of buying a sales AI tool?
Point tools each hold a slice of the deal and charge per seat. An in-house platform holds all of it, costs what hosting and model calls cost, and changes the day your process changes. The one our co-founder Neel built runs a whole revenue org on hosting plus metered model spend.
What does our team need to own it?
One engineer or a technical RevOps lead who's fine with TypeScript and SQL. The stack is deliberately boring: React, Node, Postgres on Supabase, edge functions, standard model APIs. We build together from week one, so the hand-off is a formality.
How long does a platform take?
A usable first workspace in eight weeks. Deals, accounts, briefs and analytics in four to six months, with your team shipping features by month three.
Does the platform work with HubSpot?
Yes. The reference platform is built on HubSpot: reps come in from a HubSpot record, the platform reads from a scheduled mirror, and writes go back through an allowlisted sync. Salesforce works the same way.
Is this software we license from you?
No. There is no license and no per-seat fee. We build the platform in your own cloud and database accounts, and the code, the documentation and the credentials are yours from day one. If you stop working with us, it keeps running.
What does a platform cost?
It is quoted as one fixed number after the free 30-minute audit, like everything we build. It comes in two sizes: the platform blueprint, which takes 4 weeks, then the platform foundation, which takes 8 weeks. After that you carry on with us month by month, or build the rest yourselves.
Where does our data live?
In your own database and your own cloud accounts. We work inside your systems, under your NDA or ours, and our access ends when the work does.
Does the AI act on its own?
No. AI prepares, drafts and checks. A named person approves anything that gets sent, and every job has a cost for each run, a quality check and an off switch.
Is the platform a separate service?
No. It is the two largest sizes of Automation and AI, one of our five services. Most teams start with a smaller size of the same service.
Ask about the platform
Tell us what your team opens before a call.
List the tools, say who would own the platform on your side, and what you would want on the first screen. We reply within one business day, and the free 30-minute audit tells you whether to build, buy or start smaller.