Understand

CRM driven by AI agents

A CRM never dies from a flaw in the tool. It dies from data entry. For the first six weeks everyone fills in the records; by the third month, half the live deals are no longer there, and the team goes back to its notes. The problem is not the discipline of the sales team: it is that humans are being asked for a consistency nobody sustains. That is precisely what an agent sustains.

1. Why a CRM goes stale

A CRM asks the person producing the value — the salesperson, the technician, the founder — to stop and describe what they have just done. This double task is lost in advance: when the week tightens, data entry is always what gets dropped.

What is then observed is mechanical. Records age, so nobody trusts them any more; since nobody trusts them, nobody fills them in. The steering figures become false, and decisions revert to instinct — exactly what the tool was meant to prevent.

Companies try to answer this with obligation or with shorter forms. Neither of those two answers has ever held, because the cause is not the length of the form: it is the fact that a human has to open it.

2. What an agent keeps up to date

The shift is simple to state: updating stops being a human task. It becomes the by-product of the work the agent is doing anyway.

Contact records — after every exchange, what was said, the next step and its date, so the record stops being an approximate memory.
The sales pipeline — deals sorted by stage, with their real age, and a flag on those that have been dormant too long.
Quotes and invoices — what has gone out, what remains unanswered, what is overdue, and the written follow-up ready to send.
Customer service requests — every incoming message qualified, prioritised, and linked to the right contact rather than lost in a shared inbox.
Projects and tasks — progress held by those who execute, agents and humans alike, in the same board.

None of these five tasks requires judgement. All require consistency — and that is the one point where a machine is structurally better than a busy team.

3. The CRM becomes a workplace

The fundamental difference from a classic CRM is not assistance with data entry. It is that the CRM stops being a database that humans consult, and becomes the place where agents work.

An ordinary CRM records what happened. A CRM held by agents acts: it follows up, qualifies, prepares, flags. You no longer open a board to find out what to do — you open a board to check what has been done, and settle the few points that call for a decision.

It is also what makes the whole thing measurable. Every action taken by an agent leaves a timestamped trace: who acted, on which record, at what cost. A CRM filled in by humans never knows who did not fill it in.

4. What must not be automated

Three things stay on the human side, and not for technical reasons.

Commitments. A discount, a payment schedule, a commercial gesture on a dispute: these are decisions that bind the company. They must go through a validation, not because an agent would be incapable, but because their cost is real.

Qualifying strategic accounts. An agent correctly sorts ordinary volume. On the few accounts that make your year, the sorting depends on information written down nowhere — and a bad filing is paid for over a long time.

Deletion. Merging two records, archiving an account, erasing a history: these gestures are irreversible and often regretted. They get proposed, they do not execute on their own.

5. Measuring the result

Three figures are enough, and they must be read together.

File freshness — the share of live deals whose last update is less than a week old. That is the indicator that used to collapse, and the first to recover.
Human rework time — how many minutes a day the team spends correcting the agent. If it does not fall, the rule is badly worded: that is what needs revisiting, not the agent.
What no longer falls through — quotes with no follow-up, tickets unanswered within the deadline, overdue invoices left unflagged. These three counters must trend towards zero; if they do not, the automation is cosmetic.

A CRM that displays magnificent dashboards but demands systematic verification has automated nothing: it has shifted the load from the salesperson to the controller.

6. Frequently asked questions

Should I abandon my current CRM?

Not necessarily. An agent knows how to work inside the tools already in place and use them as a person would. Changing CRM at the same time as introducing agents means running two projects at once, and no longer knowing which one failed if something goes wrong. nullbot ships its own CRM for those who have none, or whom theirs has worn out.

What does nullbot's CRM contain?

Contacts, the sales pipeline and its forecasts, quotes and invoices, customer service, projects and tasks, marketing, the calendar and reports. The whole is designed to be held by agents as much as by humans, which is not the case for a CRM built for manual entry.

Where is customer data stored?

On your machine or your server, in your own database. nullbot runs locally: the customer file does not pass through our servers and is pooled with no other company. This is detailed on the Security page.

Where should I start?

With following up on quotes that got no answer. Its rule fits in one sentence, the result is visible within days, and a mistake there is harmless. The rest of the CRM fills up afterwards, task by task.