The enquiries come in. Someone reads them when they get a minute. The good ones get answered fast, the ambiguous ones sit, and the ones that arrived on a busy Tuesday never get answered at all. A year later there are hundreds of names in a system nobody has spoken to, and no way to tell which of them were worth speaking to.
So the question comes up: should we automate this? The fear underneath that question is usually the same one. That a machine will send something wrong, to a real customer, in your name, and you will find out afterwards.
That fear is correct. It is also solvable, and the answer is not a better model. It is a gate. Three things go wrong without one, and each of them costs money:
None of that is fixed by a chatbot or another dashboard. It needs a system that does the reading, the ranking and the drafting, and then stops and waits for a person.
Not against a generic template. Against the signals that actually predict a customer in your business, learned from the enquiries you have already won and lost. Ranked HIGH, MEDIUM or LOW, so the day starts with a priority order instead of a flat list.
Leads that went quiet are the cheapest pipeline you have, because you already paid to acquire them. The system works the backlog, finds the ones worth another approach, and writes the approach.
Rules you set about what can be offered, to whom, and when. The system cannot quote outside them. This is where automation quietly costs money, because a tool that discounts to close will discount forever.
Every message is written and queued. None of them send on their own. A person sees exactly what is about to go out, to whom, and in whose name, and presses send.
This is the part that matters, and it is the part most vendors skip. A person presses send, and that person can see exactly what is about to go out, to whom, and in whose name.
We build it this way for a plain reason. When an automated message lands badly, “the model decided” is not an answer you can give a customer, a regulator, or your own team. Someone has to be able to say why it went out. So the system is built so that someone always can.
It also makes the whole thing auditable. Every score, every draft, every approval and every send is logged. If somebody asks what happened on a Tuesday in March, you can show them.
No new platform for your team to learn. No migration. No second system to keep in sync with the first. Your CRM, your inbox, your existing tools, and we build into them.
That is deliberate. Most automation projects fail at adoption, not at build. A system your team has to open a separate tab for is a system they stop opening.
Not a subscription to something we host while you rent access. If you stop working with us tomorrow, the system keeps running and someone else can maintain it.
Our first reactivation templates were too generic. We wrote them to cover every scenario, which is exactly why they read like marketing email. The lesson we now apply from the start: these have to be written in the operator's own voice, not ours, or the reply rate suffers for it.
We should have built the guardrail override path on day one. A senior person sometimes has a legitimate reason to go outside the approved range, and our first version gave them no fast route to do it. A guardrail with no sanctioned exception just teaches people to work around the system.
The scoring model was the easy part. The workflow around it is what decides whether a sales team actually uses the thing, and that is where the extra iteration goes.
Have more questions? Book a free 30-minute call and we will answer them honestly.
Book a call →AI lead scoring uses machine learning to evaluate and prioritise leads based on engagement history, behaviour patterns, and conversion likelihood — replacing manual review while keeping human oversight on high-value decisions. The model learns from your specific data and improves over time.
The system picks the moment to re-approach a lead that has gone quiet, then sends a message that references what they originally asked about rather than a generic blast. Across the leads this system has scored it runs at roughly a 12 percent reply rate, on contacts who were otherwise receiving nothing at all.
Pricing guardrails are hardcoded rules that prevent the system from generating or sending quotes outside approved price ranges. Unlike soft warnings, these rules physically cannot be overridden by the AI, the sales rep, or anyone else. Product specifications, discount tiers, and regional adjustments are all validated before a quote can exist.
It works wherever there is a lead pipeline and a record of what converted. The model is trained on your own conversion history rather than generic patterns, so the signals it weighs are the ones that predict a sale in your business. The more history you have, the sharper the scoring gets.
We do not publish a return figure. It depends on your lead volume, your margin and how long your sales cycle runs, and anyone quoting you a multiple without seeing those three things is guessing. The 21-Day AI Pilot exists to measure it on your own data before you commit to anything larger.
Through the KORIX 21-Day AI Pilot, a governed lead scoring system can be live within 3 weeks — including CRM integration, scoring model, approval workflows, and audit trail. Learn about the Pilot →
No. KORIX integrates with your existing CRM — we don't require you to switch platforms. The AI layer sits on top of your current system, pulling lead data and pushing scores back. No rip-and-replace.
