CRM for startups
Founder-led selling works, and then one day it does not scale. This is about building the sales engine while it is still small enough to change — so the second salesperson inherits a system instead of a story.
Per user, per month · GST extra
The week founder-led selling stops working
It rarely announces itself. It shows up as four things that were fine last quarter.
The pipeline is a conversation, not a record
The founder can describe every open deal from memory, accurately. That works right up until someone else needs to answer the same question, or until the founder disappears into a fundraise for six weeks.
Every deal is run slightly differently
Improvising the sequence is an advantage while you are still learning what works. It becomes a liability the moment you want to know which approach converts, because no two deals are comparable.
The learning stays with one person
Which objection keeps recurring, which segment closes faster, which pricing conversation goes badly — all of it lives with whoever had the calls, and none of it transfers by explaining it once.
The first hire has nothing to copy
You hire someone to sell and discover the process was never written down. Onboarding becomes shadowing, which does not scale past the second person and rarely survives the first quarter.
What repeatable means once it is in the product
The word gets used loosely. These are the four things that have to be true before a process can be handed to anyone.
Stages that describe a customer state
A stage should name what has become true — the requirement is scoped, the price has been seen — not what your team did. Stages you define, each carrying a probability, so the pipeline number means something.
Qualification recorded, not recalled
Custom fields hold the two or three things that genuinely qualify a lead in your market. Written down while the definition is still changing, they become the evidence for whether the definition was right.
Ownership as a rule
Assignment on arrival — round-robin, load-balanced, or to a named person. The rule is the part that survives a new hire; a habit of whoever-sees-it-first does not.
A stage clock, not a stage label
Stages can carry an SLA in hours and every stage change is recorded, so time-in-stage becomes measurable. That is how you find the step where the process leaks rather than where you assume it does.
Building the engine, one step at a time
- 01
Capture without a decision each time
Website forms, WhatsApp through your own Cloud API credentials, bulk imports and your own systems via the API all land in one queue, deduplicated on arrival. At ten leads a week the habit is what you are building, not the throughput.
- 02
Assign by rule from day one
Even when the rule is everything-to-the-founder, encoding it means the day it becomes split-with-the-new-hire is a settings change rather than a new convention nobody remembers.
- 03
Qualify against written criteria
Custom fields turn qualification into a recorded answer. When the criteria change next quarter — and they will — you can still see which deals were judged under which definition.
- 04
Convert to a deal with a number on it
Value and probability against a stage, carrying activities, notes and history across. This is where a pipeline stops being a list of interested people.
- 05
Move it through stages that mean something
Stage history accumulates from the very first deal. Six months in, that history is the only honest evidence you have about where deals stall and how long a real cycle takes.
- 06
Make the follow-up structural
Workflows trigger on lead created, stage changed or follow-up due, creating the task, assigning the owner, updating the record or raising a notification — so execution does not depend on whose week is calm.
- 07
Mine the calls while the pitch is still moving
Early sales calls carry the product learning. Uploaded recordings are transcribed and read for objections, commitments and buying signals, then written back onto the deal as tasks.
- 08
Keep the commercial trail on the record
Quotes raised against a deal convert to invoices, with payments recorded alongside — so the paperwork does not fragment into a founder's inbox as volume grows.
- 09
Report from the records, not a spreadsheet
Saved report definitions and dashboards read the same rows the team works in, which is what makes the board number and the operating number the same number.
Stage discipline is what makes growth measurable
The highest-leverage thing a young company can do in a CRM is define stages that describe the customer's state rather than the team's activity, then leave them alone long enough to accumulate history. Demo booked is an activity. Requirement scoped and budget confirmed is a state, and only the second one predicts anything.
Once stages mean something, the machinery underneath starts paying: a probability against each stage turns open pipeline into a forecast, an SLA in hours turns a slow stage into an alert, and stage history turns deals seem to die after the proposal into a measured claim you can act on.
This is also the part that is genuinely hard to retrofit. Data you did not capture in month four cannot be recovered in month sixteen, and the first time an investor asks about conversion by stage is a poor moment to discover every deal was recorded differently.
Context that survives the team doubling
The record is the briefing
Email, WhatsApp, calls, notes and tasks sit on one timeline against the customer. Handing a deal over becomes reading it, rather than booking time with the person who used to own it.
Permissions arrive when they matter
Role-based access decides what each person can read and change. You do not need it on day one with three people; you will need it the quarter commercial data starts mattering, and it is already there.
Seats carry their own AI allowance
Each user you add brings the same monthly Ira credit allowance the plan carries per seat, so AI cost tracks headcount instead of arriving as an upgrade decision mid-quarter.
Capacity is published before you reach it
Leads, contacts, deals, storage, automations and reports have limits stated on the pricing page and enforced by the platform. Growth should not produce a surprise in the middle of a hiring month.
What the calls are telling you while you can still act on it
At this stage the pitch, the pricing and sometimes the product are still moving, and the evidence for all three sits in conversations that currently evaporate. You provide the recording — the product does not place or record calls — and it is transcribed, speaker-separated and read once in context for the summary, intent, objections, commitments and buying signals.
What makes it usable as evidence rather than as a transcript is how it is stored. Nothing arrives as a bare assertion: each finding is marked explicit, inferred or absent, with its supporting line of transcript attached. Figures keep their qualifiers: a budget hedged as around eight, could stretch to ten depending on scope is recorded as eight with a ceiling of ten and the condition attached — which matters a great deal when those numbers are about to become a forecast you present to someone.
The output lands as work, not as a document: open tasks against the deal, the recording itself on the customer timeline, and a follow-up graded by how firmly it was promised — an explicit time is scheduled, a day without a time waits for you to confirm, and anything ambiguous stays a suggestion. Call Flagging AI reads the same transcript for phrases you define — useful the moment someone other than a founder is talking to customers.
The complete pipeline is on the Call Recording Intelligence page.
Automating before headcount multiplies the work
Automation early is not about volume — you do not have volume yet. It is about deciding the behaviour once, while there is one person to agree with, so that it is already true when there are six.
Workflows trigger on real events: a lead created, a stage changed, a follow-up date passing. The resulting task is created, ownership set, the record advanced, or a notification raised for whoever is meant to act. Customer-facing steps are two distinct choices at design time — one dispatches the message, the other leaves it in drafts for review — so the behaviour is settled the day the rule is written. Provider-backed channels such as WhatsApp and SMS remain inactive until connected.
Ira sits alongside that rather than inside it: it answers questions from your own records and drafts the message or the next action, and a person sends. As the team grows that division stays legible — the workflow does the mechanical part, the AI does the drafting, and a human keeps the customer relationship.
When it is too early for this
You have not sold anything yet
If you are still finding out whether anyone wants this, a CRM formalises a process you have not discovered. Have the conversations first — structure is worth building once there is something to make repeatable.
One founder, a few deals, no handover in sight
Plans start at three users. If nothing passes between people and nothing is being forecast, you would be buying coordination you do not yet need.
You want a platform to configure, not a process to run
There is no scripting layer and no bespoke module system here, and that is a deliberate limit rather than a gap. A genuinely unusual process may need a broader, more configurable platform instead.
Common questions
Our sales process changes every month. Will we outgrow the setup?
Stages, custom fields and routing rules are configured from the interface and can be changed as you learn, without a migration or a consultant. That is the argument for standardising early rather than against it: you record consistently while the shape is still moving, so when the definition changes you can still see which deals were judged under the old one.
What actually changes when the first sales hire arrives?
Less than you would expect, if the system existed before they did. Assignment becomes a rule that includes them, the deals they inherit arrive with their full timeline attached rather than a verbal briefing, and the stage history already there tells them how long a real cycle takes. The alternative is shadowing, which does not survive a second hire.
How does the AI help a two-person sales effort?
Ira answers from your own records inside the asker's permissions — which deals have gone quiet, what is overdue, what a customer's history is — and drafts follow-ups, emails and messages for you to send. It cannot send anything itself, and its write actions are limited to creating a lead or task, scheduling a meeting and assigning a lead. When a founder is selling and operating at once, the value is not reconstructing context before every call.
Why upload call recordings this early?
Because early calls are where the product learning is, and almost none of it survives in a note written afterwards. A recording is transcribed and read for objections, commitments and buying signals, with each extracted fact tagged as explicitly stated, inferred or missing and carrying its quote — so a hedged budget stays hedged instead of quietly becoming a number in a forecast.
Can it replace the separate tools we are using?
For lead capture, pipeline, follow-ups, quotes, invoices, payments and reporting, yes — those are one workspace here. Email and WhatsApp connect through your own accounts and credentials rather than being resold. What it is not: it does not place phone calls, and it does not cover finance, HR or support.
What does it cost as we add people?
AED 49 on Essential, AED 79 on Growth and AED 99 on AI Pro, per user per month and VAT may apply. Each additional user adds the same monthly Ira credit allowance the plan carries per seat — 1,000, 2,000 or 5,000. Workspace capacity limits are published per plan. There is no free trial.
Explore Insta AI 365
Insta AI 365 CRM
The one workspace all of this runs inside.
Sales Pipeline
Stages, probability, stage SLAs and the history behind them.
AI-powered Automation
Triggers on real events, and the actions you choose between.
Ira AI
Answers grounded in your own workspace records.
Pricing
What it costs per seat as the team grows.
Define the stages before the second salesperson arrives
Set the four or five stages you actually sell through, encode one assignment rule, and let the stage history start accumulating now. It is the one thing you cannot backfill later.
Per user, per month · GST extra